An Algorithmic Approach for Identification, Conditon ...

8 downloads 0 Views 545KB Size Report
[53] Boris Zhukov et al; Unmixing-based Multisensor Multiresolution Image. Fusion- IEEE Transactions on Geoscience and Remote Sensing, Vol. 37,. No.3, May ...
An Algorithmic Approach for Identification, Conditon Assessment and Productivity Estimation of Agriculture Crop from Satellite Based High Resolution Imagery: A Review Anil B. Gavade

Vijay S Rajpurohit

Electronics and Communication Department KLS Gogte Institute of Technology, Belgaum, India [email protected]

Computer Science Department KLS Gogte Institute of Technology, Belgaum, India [email protected] system. India is bestowed with wide-ranging agro-climate, which is highly favorable for growing a large number of horticultural crops such as nuts, fruits, flowers, vegetables including root tuber an ornamental crops, medicinal and aromatic plants, spices and plantation crops like arecanut, cashew and cocoa etc. Fruits are the major sources of vitamins and minerals. In today’s agriculture world India contribute to major production of horticulture crop, and considered as the world's second largest horticulture producer. During the year 2012-13 India’s annual production of horticultural crops in the country was about 240 million tones, wherein the fruits production accounts for 72 million tones and vegetables production is for 134 million tones. As per the records of Indian agriculture ministry, horticulture crop helped the country to generate a foreign exchange of Rs 14,000 crore. Agriculture is major sector in developing India, 55% of its population and providing 16.5% of its annual GDP. The agriculture industry as a whole generates US $ 17.5 million alone in export. Karnataka state contributes a major share of horticulture fruit crops. The climatic condition and soil in Dharwad and Belgaum districts of Karnataka is suitable for the growth of horticulture fruit crops like mango, Banana, Lemon, Guava, Sapota, Pomegranate, Jack fruit, Papaya, Custard apple, Grapes and many more. The market is growing at fast pace and there is need to increase the crops. In this review study is emphasized on horticulture fruit crops. Changing markets, technological innovation and organizational progress in recent years have increased the intensity and scale of agricultural land use. Producer deal with very broad and rapidly expanding range of technology, in areas such as weed control, direct seeding and varieties selection, all of which are required to optimize productivity, protect the environment and maintain or improve the profitability. In the past, the main focus of agronomic research has been on crop production only but recently, profitable crop production, the quality of the environment has become an important issue that agricultural producer must address. So the agricultural manager requires strategies for optimising the profitability of the crop production while maintaining soil quality and minimizing environmental degradation. Solution of these new challenges requires consideration of how numerous components interact to effect

Abstract - Agriculture sector is greatly influenced by

Information Technology (IT); linking IT with agriculture is one way of balancing employment and crop productivity and at the same time reducing poverty. Modern technology on agriculture is heavily dependent on natural resources. Crop growth and productivity are determined by a large number of weather, soil and management variables. The Crop Modelling and Management (CMM) can be used in a wide range of agriculture activities including field crop production, dairy farming and forest management. Crop growth simulation has made valuable contributions to yield forecasting, proto-typing crop varieties, generation of input-output coefficients for improved agricultural production. Remote Sensing (RS) data, acquired repetitively over agricultural land help in identification and mapping of crops and also in assessing crop vitality. The aim of this research is to employ information technology in agriculture using remote sensing, data mining and softcomputing for identification, condition assessment and productivity estimation of crops. The proposed model attempts to simulate the way in which a crop responds to its environment. Mapping crop simulation model to Geographic Information System (GIS) will benefit “demand and supply chain management”, and crop insurance organizations. Keywords—Crop Modelling and Managment; Remote Sensing; crop vitality;Crop yield; Geographic Information System

I. INTRODUCTION Agriculture not only provides food and raw material but also employment opportunities to a very large proportion of population. Agriculture and farming were synonymous so long as farming was not commercialized. But as the process of economic development accelerated, many more other occupations allied to farming came to be recognized as a part of agriculture. Agriculture plays a crucial role in the life of an economy. Agriculture is the backbone of our economic

1

plant growth. With shrinking land and scarce natural resources the path ahead for the Indian horticulture should be to diversify into hi-tech horticulture to improve the productivity and quality and meet the demand. To achieve this goal, future agriculture research will require more effort and resource than present research activity [1- 3]. The application of aerial or satellite imaging is the first step towards the successful application of CMM for horticulture crops [4]. The use of CMM for horticultural crops has potential for increasing net returns and optimizing resource use. The delineation of orchards and spatial analysis using geospatial technology can provide additional information for management decision making, such as the determination of horticultural crop yield, the quantification and scheduling of precise and proper fertilizer, irrigation needs, and the application of pesticides for pest and disease management. Ultimately, it will improve profits for producers [5]. Remote sensing has become a universal tool for general detection of health condition of orchards on a larger scale. Digital imaging technology increasingly being used for intensive site-specific management of orchards as well, for instance: estimating the amount of fruits on individual trees, fruit quality, and also Leaf Area Index (LAI) or crown cover [5, 6]. II.

Fig. 2. Crop simulation model showing the interaction between enviromental factors, crop growth and development, and mangment.

The Digital Number (DN) of a pixel in a spectral band represents the amount of radiation received at the sensor, which is determined primarily by the capability of the ground object in reflecting and emitting energy. The amount of energy reaching the sensor is a function of the wavelength of the radiation. Thus, pixel value varies from band to band. A spectral class is defined as a cluster of pixels that are characterized by a common similarity in their DNs in the multispectral space. Whether a group of pixels can be regarded as one cluster is subjective, dependent on the specification of spectral distance among these pixels. If the distance between a pixel and a group of pixels falls within the specified threshold, this pixel is considered a part of that cluster. An information class, on the other hand, is a category of ground features retained in the classification results. Every category included in the enacted classification scheme represents an information class. It corresponds to a specific type of ground cover or feature to be extracted from remotely sensed data. The purpose of image classification is to map the input data into these information classes as accurately and reasonably as possible [10]. There is a complex relationship between information classes and spectral classes. Rarely, they correspond to one another. In other words, it is almost impossible for one information class to be linked uniquely with a spectral class. On the contrary, one information class may exhibit a wide range of variations in its spectral value. Consequently, it can correspond to a number of spectral classes that are formed out of many slightly different but significant variations in appearance caused by the status, spatial composition (e.g., varying density), and the environmental settings. For instance, the appearance of a typical forest in a satellite image is affected by its age, and the differing proportions of mixture with trees of other species and topography (Figure 3). The task of image classification is to merge these different and numerous spectral clusters rationally to form meaningful information classes [9].

CROP MODEL - PROBLEM DEFINATION

A model is a programme (and the science behind it) that simulates (predicts) the behaviour (output) of a plant (e.g. yield) from environmental conditions (inputs, incl. management) and variables describing the plant's ecophysiology (parameters) Figure 1 shows the system flowchart of a comprehensive image analysis procedure, Figure 2 schematically illustrates the operation of a typical crop model [ 8 ].

Fig. 1. Comprehensive image anslysis procedure

2

Parameters considered for Horticulture crop: • Current Horticulture status of study area, Current and future demands and trends for horticulture crop business in India and Abroad • Climatic Weather and atmospheric conditions of study area for horticulture crop production • Diseases and pests - Diversity and management, Classification of horticulture crop density based on the verity • Atmospheric condition which may have impact on horticulture crop of Belgaum and Dharwad • Horticulture crop pattern of study area (Statistics of 3 to 5 years duration, which will collected from government organization)

Fig. 3. Relationship between an information class and spectral classes.

II. STUDY AREA For the study of horticulture crop review we have considered two district of the north Karnataka namely Belgaum and Dharwad region.

III.

CROP MODELLING

A. Crop Modelling: Crop models are tools of systems research which helps in solving problems related to crop production. A MODEL attempts to simulate the way in which a crop responds to its environment. Model outputs are usually valueadded parameters that are more closely linked to crop yield than the inputs. Crop modeling enables scientist to define projections of future agricultural yields and restate research priorities. Researchers can estimate the importance and effect of various parameters and arrive at conclusion about which factors should be studied more in future to increase the understanding of system. Crop growth is a very complex phenomenon and a product of a series of interactions of water, soil, plant and weather [11]. Thus, Crop model must possess potential to help understand basic interactions in soil-plantatmosphere system. In India, we still predominantly use traditional techniques such as field based Crop Cutting Experiments (CCE) to assess the crop yield and acreage. It is worthwhile to note that in India all crop exports and import decisions are still based on historical production data (previous year’s production records), a more scientific way of assessing the yield and acreage in advance way is done by using remote sensing and GIS techniques. The ramification of taking crucial export and import decisions is based on historical data that is based on perceived shortage or surplus. Agricultural data is currently generated by multiple agencies in multifarious ways; both conventional field surveys based as well as advance information technology based [12].

Fig. 4. Study area Belgaum and Dharwad, Karnataka, India

TABLE I.

STUDY AREA

Districts of North Karnataka, India Districts

Latitudes (North)

Longitudes (East)

covers an area

Dharwad

15°19’ and 15°14’

74°43’ and 75°15’

4,263 sq. km

Belgaum

15°00’ and 17°00’

74°00 and 75°30’

13,444 sq. km

V. Radha Krishna Murthy [13] discusses various crop growth modeling approaches viz. Statistical, Mechanistic, Deterministic, Stochastic, Dynamic, Static and Simulation etc. Role of climate change in crop modeling and applications of crop growth models in agricultural meteorology are also discussed. A few successfully used crop growth models in agrometeorology are discussed in detail. V.K. Dadhwal [14] Crop growth and productivity are determined by a large number of weather, soil and management variables, which vary significantly across space. Remote Sensing (RS) data, acquired repetitively over agricultural land help in identification and mapping of crops and also in assessing crop vigour. As RS data and techniques

Fig. 5. LISS III satellite Images of Study area Courtsey (KSRSAC: IRS RESORCSAT)

3

have improved, the initial efforts that directly related RSderived Vegetation Indices (VI) to crop yield have been replaced by approaches that involve retrieved biophysical quantities from RS data. Thus, Crop Simulation Models (CSM) that have been successful in field-scale applications are being adapted in a GIS framework to model and monitor crop growth with remote sensing inputs making assessments sensitive to seasonal weather factors, local variability and crop management signals. The RS data can provide information of crop environment, crop distribution, Leaf Area Index, and crop phenology [15]. This information is integrated in CSM, in a number of ways such as use as direct forcing variable, use for re-calibrating specific parameters, or use simulationobservation differences in a variable to correct yield prediction. A number of case studies that demonstrated such use of RS data and demonstrated applications of CSM-RS linkage are presented. Advantage of CMM: The aim of constructing crop models is to obtain an estimate of the harvestable (economic) yield. According to the amount of data and knowledge that is available, within a particular field, models with different levels of complexity are developed. The advantages of CMM in general include the ability to provide a framework for understanding a system and for investigating how manipulating it affects its various components, assess longterm impact of particular interventions, make available an analysis of the risks involved in adopting a particular strategy and provide answers quicker and more cheaply than is possible with traditional experimentation.

profitability, productivity, sustainability, crop quality, environmental protection, on - farm quality of life, food safety, and rural economic development. Refinement and wider application of PA technologies in India can help in lowering production costs, enhancing higher productivity and environmental benefits, and better utilization of natural resources. When PA technologies judiciously implemented, farmers could be benefited in many ways. In the short term, growers can use forecast based on remote sensing and alleviate problems such as water stress, nutrient deficiency and pests/diseases more affectivity. Database - building benefits will be in the form of accurate farm document keeping for effective management of inputs, property, machinery and labor, and efficient monitoring of environmental quality through recording the amounts and location of input through applying at exact locations that produce maximum profit margins. PA technologies also increase opportunities for skilled employment in farming, and provide new tools for evaluating multifunctional character (including non - market functions) or agriculture and land [1619]. IV.

GEOSPATIAL TECHNOLOGY

Remote Sensing: Remote sensing can be defined as a technology that is employed to acquire information about an object by detecting the energy reflected or emitted by that object when the distance between the object and the sensor is far greater than any linear dimension of the sensor. Remote sensing techniques have been used for land cover mapping activities for over 30 years, even before the term was coined, and they played an important role in many developments that we enjoy today. “Mapping forest vegetation from aerial photographs was first attempted in the 1850s using a camera carried aloft in a hot air balloon” and has been used for the identification and monitoring of agricultural land use targets since the late 19th century. The backscatter signature from each crop in a microwave image using active sensors varies according to the target’s characteristics, such as leaf moisture, plant separation and number of leaves per square meter or canopy over the ground. The crop growth models predict the characteristics of different crops at given times, and provide inputs to the radar models, which estimate the backscatter from each crop of interest. In this context, the use of microwave images may be an alternative to direct field measurement when many of the fields become inaccessible due to the wet season conditions [20-22].

B. Precision Agriculture(PA): Precision agriculture is a technology provides three basic requirements for precise and sustainable agricultural management. Those are ability to identify precise location of field, gather and analysis information on spatio - temporal variability of soil and crop conditions at field level, and adjust input use and farming practices to maximize benefits from each field location. Precision Agriculture also known as precision farming, Variable Rate Technology (VRT) and site specific agriculture – is one of the nurturing agriculture technologies of 21st century, as it symbolizes a better balance between reliance on traditional knowledge and information and management - intensive technologies. Precision agriculture changes an agriculturalist viewpoint, because focus of attention changes from average field conditions to the variation of those conditions. Precision farming takes advantages of Geospatial technology, and is a combination of essential tools: Remote Sensing, Geographic Information Systems, Global Positioning Systems (GPS), Information Technology Variable Rate Technology (VRT), Crop Models, Yield Monitors and Precision Irrigation. Agricultural sustainability in India can be achieved only if the natural resource base upon which it relies is well managed. Modern frontier technologies involving system approach towards efficient crop and input management, and scientific land and water use planning, is thereby the need of this century sustainable agricultural management.

A. Spectral imaging for remote sensing of terrestrial features: Spectral imaging for remote sensing of terrestrial features and objects arose as an alternative to high-spatialresolution, large-aperture satellite imaging systems. Early applications of spectral imaging were oriented toward groundcover classification, mineral exploration, and agricultural assessment, employing a small number of carefully chosen spectral bands spread across the visible and infrared regions of the electromagnetic spectrum. Improved versions of these early multispectral imaging sensors continue in use today. A new class of sensor, the hyperspectral imager, has also

Advantages of PA: PA promises to revolutionize management as it offers a variety of potential benefits in

4

emerged, employing hundreds of contiguous bands to detect and identify a variety of natural and man-made materials. Electro-optical remote sensing involves the acquisition of information about an object or scene without coming into physical contact with that object or scene. Panchromatic (i.e., grayscale) and color (i.e., red, green, blue) imaging systems have dominated electro-optical sensing in the visible region of the electromagnetic spectrum. Longwave infrared (LWIR) imaging, which is akin to panchromatic imaging, relies on thermal emission of the objects in a scene, rather than reflected light, to create an image. More recently, passive imaging has evolved to include not just one panchromatic band or three color bands covering the visible spectrum, but many bands—several hundred or more— encompassing the visible spectrum and the near-infrared (NIR) and shortwave infrared (SWIR) bands. This evolution in passive imaging is enabled by advances in focal-plane technology, and is aimed at exploiting the fact that the materials comprising the various objects in a scene reflect, scatter, absorb, and emit electromagnetic radiation in ways characteristic of their molecular composition and their macroscopic scale and shape. If the radiation arriving at the sensor is measured at many wavelengths, over a sufficiently broad spectral band, the resulting spectral signature, or simply spectrum, can be used to identify the materials in a scene and discriminate among different classes of material. While measurements of the radiation at many wavelengths can provide more information about the materials in a scene, the resulting imagery does not lend itself to simple visual assessment. Sophisticated processing of the imagery is required to extract all of the relevant information contained in the multitude of spectral bands. The measurement, analysis, and interpretation of electro-optical spectra is known as spectroscopy. Combining spectroscopy with methods to acquire spectral information over large areas is known as imaging spectroscopy. Figure 9 illustrates the concept of imaging spectroscopy in the case of satellite remote sensing [23-25].

• Multispectral remote sensing involves the acquisition of visible, near infrared, and short-wave infrared images in several broad wavelength bands and Hyperspectral imaging systems acquire images in over one hundred contiguous spectral bands. • While multispectral imagery is useful to discriminate land surface features and landscape patterns, hyperspectral imagery allows for identification and characterization of materials. In addition to mapping distribution of materials, assessment of individual pixels is often useful for detecting unique objects in the scene. C. Image processing of remotely sensed data: Image processing is an aspect of the computer vision area that involves two steps: low level and high level as described in Rao and Jain [26]. Low-level vision is based on the extraction of features, resulting in a segmented image with labelled different regions, where the shapes, spatial interrelationships, and surfaces of objects may be described. The high-level vision consistently attempts to interpret the labels obtained from low-level processing using a priori information about the scene’s domain. Thus, they considered the two main steps in image processing: segmentation and interpretation. The basic steps involved in general image processing are also discussed in Gonzalez and Wood [27] as acquisition (capturing image), pre-processing (quality improvement), segmentation, and interpretation. D. Remote Sensing image classification: In remote sensing image classification, techniques are most generally applied to the spectral data of a single-date image or to the varying spectral data of a series of multispectral or multi-date images. The complexity of image classification techniques can range from the use of a simple threshold value for a single spectral band to complex decision rules that operate on multivariate data. Numerous classification approaches have been used with varying degrees of success. Despite the considerable recent developments, the accuracy with which thematic maps may be derived from remotely sensed data is still often judged too low for operational use [28-31]. E. Remote Sensing in Horticulture Crop CMM: Remotely sensed images are a quick and inexpensive means to differentiate agriculture crops or orchards from other land-uses. The remotely sensed images are the best possible resource to differentiate the spatial variation of crops, including horticultural crops. Several studies have been conducted over the years to delineate or classify forests or shrubs, e.g., horticultural plants, based on satellite and aerial images for precision agriculture decision making. Many fruit and nut orchards are close to, within, or adjacent to, forested land cover [32]. Differentiating the plants from forest trees is a difficult task, but it can be achieved with advanced image processing. The total reflectance of a forest canopy is the combination of illuminated and shaded components of the tree crown as well as the background because of the bidirectional behavior of forest canopy. With advanced image processing techniques and use of high resolution multispectral and hyperpsectral imagery, most fruit and nut plants can be

Fig. 6. Imaging spect spectroscopy, remote sensing, the graphs in the figure illustrate the spectral variation in reflectance for soil, water, vegetation

B. Multispectral and Hyperspectral Remotely Sensed Images: • Multispectral or hyperpsectral images are preferred for accurate classification of different crops in CMM. • These different techniques operate at different wavelengths.

5

duration to capture Cartosat 1/2 around once in 40day and LIS III images can be obtained once in 20days.

distinguished from the mixed vegetation of tall forest trees and dwarf grass or other small shrubs if their unique spectral characteristics are known, especially with respect to the land covers such as forests or grasses [ 33]. Spectral reflectance curves are useful for researchers engaged in developing CMM for individual fruit and nut crops through the use of remote sensing. The curves would guide researchers to decide which band images they need to procure to obtain distinguishable digital information. Again, these curves also point towards obtaining the digital number range (0–255) for a crop at a particular band using the reflectance percentage (Y-axis info of the curves) with respect to the wavelengths (X-axis info of the curves) [34-37] .

H. Resolution of Satellite Image: Spatial and temporal resolution of satellite image is important for accurate analysis, classification region of interest, today many techniques are available but super resolution reconstruction and/or image fusion techniques give promising results. The spatial resolution of Multispectral or hyperpsectral images can be improved to a greater extent, and this help in accurate analysis and detection of any crop from the satellite image. I.

F. Classification of crops:

Super Resolution based Image Reconstruction (SR):

Super-resolution is the task of obtaining a highresolution image of a scene given low resolution image(s) of the scene. Applications of super-resolution include satellite, forensic, medical imaging, surveillance et al. Obtaining highresolution images directly via better hardware (better image sensors, larger chip size) is quite costly. Image interpolation methods are not considered as super-resolution methods since even the ideal sinc interpolation cannot recover the high frequency components that are lost in the low-resolution sampling process. Most of the super-resolution approaches work on the principle of combining multiple slightly-shifted low-resolution images of the scene. This involves image registration, interpolation and deblurring as the basic operations. However, this technique is numerically limited to small increases in resolution. There also exist methods based on techniques like gradient profile priors and sparse representation of images in an over-complete dictionary. A recent approach based on a single image was proposed in that exploits the redundancy of patches within the image and combines this information with example-based techniques such as [41-43]. Super-resolution mapping is a relatively new field in remote sensing whereby classification is undertaken at a finer spatial resolution than that of the input remotely sensed multiple-waveband imagery. A variety of different methods for super-resolution mapping have been proposed, including spatial pixel swapping, spatial simulated annealing and Hopfield neural networks, feed-forward back-propagation neural networks and geostatistical methods [48]. There is, therefore, a need for greater inter-comparison between the various methods available, and a super-resolution intercomparison study would be a welcome step towards this goal. A geostatistical optimization algorithm is proposed for super-resolution land cover classification from remotely sensed imagery. The algorithm requires as input, a soft classification of land cover obtained from a remotely sensed image. A super-resolution (sub-pixel scale) grid is defined. The soft land cover proportions (pixel scale) are then transformed into a hard classification (subpixel scale) by allocating hard classes randomly to the sub-pixels. The number allocated per pixel is determined in proportion to the original land cover proportion per pixel. The algorithm optimizes the match between a target and current realization of the two-point histogram by swapping sub-pixel classes within pixels such that the original class proportions defined per pixel are maintained. The algorithm is demonstrated for two simple

Remote sensing image classification involves the grouping of all or selected land cover features into summary categories that allow the interpretation of the earth’s surface imaged by the imaging sensor. In remote sensing image classification, techniques are most generally applied to the spectral data of a single-date image or to the varying spectral data of a series of multi-spectral or multi-date images. The complexity of image classification techniques can range from the use of a simple threshold value for a single spectral band to complex decision rules that operate on multivariate data. Numerous classification approaches have been used with varying degrees of success. Despite the considerable recent developments, the accuracy with which thematic maps may be derived from remotely sensed data is still often judged too low for operational use. Martin and Heilman [38] conducted special measurement of several rice varieties with different morphological characteristics using a portable radiometer in Thematic Mapper spectral bands. They observed that vegetation indices were highly correlated with LAI; however, vegetation indices were insensitive to morphological differences among varieties. Berg and Gregorie [39] adopted remote sensing techniques for area estimation of rice in West Africa. Area estimation was done at two stages of crop development; three months and one month befire harvest. The methodology was based on visual interpretation as well as digital analysis of two Landsat images. In chickpea, Vyas et al. [40] reported remote sensing based acrage estimate using IRS-1A LISS-I data. The relative deviation of acreage estimate was 1.89 per cent to that of Official estimate by Bureau of Economics and statistics (BES) in Hamirpur district of U.P. G. Available IRS satellite for crop modelling: India has its own satellites like Indian Remote Sensing Satellite (IRS) series - ResourceSat, Cartosat, satellite data sets supports the CMM application for Forecasting Agricultural yield. IRS have different satellites operating over our study region, namely LIS III with resolution of 23.5m, Cartosat 1 with resolution 2.5m and Cartosat 2 with resolution of 1m are available, CMM application requires collection of satellite data for every 15days or 20day duration LISS-III will supports the purpose but we need high temporal resolution, than Cartosat 1/2 aerial images will be the best, but the

6

simulated images. The advantages of the approach are its ability to recreate any target spatial distribution and to work with features that are both large and small relative to the pixel size, in combination [44-47]. J.

image fusion method has its own suitability for a specific feature extraction. For any kind of remote sensing image, its spatial resolution and spectral resolution are contradictory factors. For a given signal to noise ratio, a higher spectral resolution is often achieved at the cost of lower spatial resolution. Image fusion techniques are therefore useful in integrating a higher spectral with higher spatial resolution image. From the available literatures all the fusion techniques improve the resolution and the spectral result. The HPF keeps exactly the statistical parameters of the original images. The statistical analysis of the spectral characteristics of the data indicate that the results generated by HPF with principal component spectral transform is least distorted and the low frequency noise problem can be remove automatically, colours are well preserved in the fused image, but unrealistic artifacts for spatial improvement are observed, for vegetation extraction the HPF fused image found better [49-54].

Image Fusion:

Image fusion refers to the acquisition, processing and synergistic combination of information provided by various sensors or by the same sensor in many measuring contexts. Information Fusion is a naturally occurring phenomenon in most biological systems. Data from various sources are merged in order to make optimal decisions. International society of information fusion aptly defines it as – “Information fusion encompasses the theory, techniques and tools conceived and employed for exploiting the synergy in the information acquired from multiple sources such that the resulting decision or action is in some sense better than would be possible if any of these sources were used individually without such synergy exploitation” Image Fusion is a similarly inspired effort to merge relevant visual data sets which are dependent and yet have disparity to certain extent in order to come up with a smaller data set apt for a better semantic interpretation of data for a given application. In the field of remote sensing, satellite images are captured in various frequency bands with different spatial, temporal and spectral resolutions. The data acquired from all these sources have a disparity in nature and images of same locations have spatial dependency. This situation is exploited by image fusion algorithms to come up with merged images which are more detailed and information-rich than any of the individual images. Image Fusion in Remote Sensing: If the resolution of the captured data is too low or the distortion in the data is too high, the data may be of no use in any application. Unfortunately in real-world conditions the resolution and distortion are generally worse than what the original hardware was designed for or what is required of a given applications. Image fusion plays a very important role in such cases. It is able to convert a combination of such data sets with low information content to a single data set of higher information content. “Higher information content” is only meant in a semantic sense for a particular application. The objective of image fusion is to combine the high spatial and the multispectral information to form a fused multispectral image that retains the spatial information from the high resolution panchromatic image and the spectral characteristics of the lower resolution multispectral image [39]. Generally, image fusion methods can be differentiated into three levels: pixel level (iconic), feature level (symbolic) and knowledge or decision level. Of highest relevance for remote sensing are techniques for iconic image fusion, for which many different methods have been developed. Many researchers have focused on how to fuse high resolution panchromatic images with lower resolution multispectral data to obtain high resolution multispectral imagery while retaining the spectral characteristics of the multispectral data and various image fusion methods have been developed. The most popular image-fusion methods are Brovey, Ehlers, HPF (High Pass Filter), Modified HIS, Multiplicative, Principal component, and Discrete wavelet Transform (DWT). Each

Indian satellites such as IRS1C, IRS1D, IRS-P5, and IRS-P6 provide panchromatic images at a higher spatial resolution than in their multispectral mode. IRS 1C or 1D dataset, both multispectral Linear Imaging and Self-Scanning sensor (LISS III or LISS IV multispectral (3 band) image with spatial resolution of 5.8 ms / 23 ms) and Panchromatic (PAN) (Cartosat I/II spatial resolution of 2.5 m / 0.5 m). After comparing all spectral and statistical parameter, normalised difference vegetation index (NDVI) value of each fused image is evaluated and finally suggested the suitability of the fused method for vegetation identification. Along with statistical parameters and NDVI values of each fused image, it is observed that the Ehler’s method is suitable for vegetation mapping is concerned. K. Data mining, soft computing tools: Soft computing is a set of “inexact” computing techniques, which are able to model and analyze very complex problems. For these complex problems, more conventional methods have not been able to produce cost-effective, analytical, or complete solutions. Soft computing has been extensively studied and applied in the last three decades for scientific research and engineering computing. In agricultural and biological engineering, researchers and engineers have developed methods of fuzzy logic, artificial neural networks, genetic algorithms, decision trees, and support vector machines to study soil and water regimes related to crop growth, analyze the operation of food processing, and support decision-making in precision farming. With the concepts and methods, applications of soft computing in the field of agricultural and biological engineering are presented, especially in the soil and water context for crop management and decision support in precision agriculture [55]. Thomson and Ross developed a coupled sensor- and model based irrigation scheduling method. Based on this work Thomson [56] described a concept by which readings from sensors could be used to re-parameterize or adjust inputs to a crop simulation model. Readings from sensors and the modeled results were used synergistically in an adaptive learning scheme. One of the most recent advances in the field of remote sensing has been the adaptation of Artificial Neural Networks (ANN) in a wide range of applications and image

7

GPGPU along with CUDA programs will increase the speed 10X time faster. This will help to perform CMM for large agriculture area [62- 64].

analysis and the number of reported application has been steadily increasing. The literature shows that the use of different types of neural networks has been found for various applications of remote sensing data. The most commonly encountered ANNs in remote sensing are Multilayer perceptron (MLP) type. Among others, applications are found employing the SOM, LVQ, and Hopfield neural networks methods for the classification of remote sensing images. Several researchers also compared the results from neural networks with the results from different conventional classifiers [57]. In the field of remote sensing, one of the main applications of neural networks is the image classification, including: supervised classification, unsupervised classification [28]; and image segmentation. Some of other uses of neural networks in remote sensing are also found, such as in geometric correction and image compression etc. Different types of neural networks have been used on a variety of remotely sensed data including optical high (Landsat TM, SPOT) and low (NOAA-AVHRR) resolution multi-spectral imagery, data from Imaging Spectrometers (AVIRIS) and SAR data (ERS-1, RADARSAT). Although early experiments made use of single source data, recent research has demonstrated the flexibility of neural networks in the fusion of multi-source data for improved land use and land cover classification L.

V.

CONCLUSION

This paper provides a comprehensive review on the development and application CMM for horticulture crops. The review also confirmed that not many studies (compared to row crop agriculture) have been conducted so far on CMM for horticulture crops. This review paper provides detailed investigation of CMM for horticulture crops specifically related to India. The study provides a detailed application of geospatial and information technology for CMM in general. They can be replicated as described or modified based on local requirements and needs. This study also provides details of super resolution and image fusion techniques that can be used to improve the spatial resolution of satellite image for accurate analysis. Finally, this review paper would encourage researchers to advance the CMM for horticulture crops as is being done in other row crop cases. The exploration concludes with an innovative model for integration of CMM with Geographic Information System (GIS) which can benefit “demand and supply chain management”, and crop insurance. VI. ACKNOWLEDGMENT We would like to express our gratitude to Principal, KLS GIT, Belgaum and KLS Society for providing opportunity to do the research work. We would also like to express thanks to Dr. D K Prabhuraj, Director, Dr. K Ashoka Reddy, Scientist, Dr. B P Laxmikantha, Scientist, of KSRSAC, Bangalore and VTU Belgaum. We also thank Dr. TGK Murty Ex-Director, ISRO Bangalore. We would like to express deep gratitude to Mr. Girish S Pujar, Scientist, NRSC, Hyderabad for extending immense support.

The integration of GIS for image processing:

A GIS is a broad term which encompasses a large range of applications. These applications commonly fulfill the five M’s of GIS: mapping, measurement, monitoring, modeling, and management [58]. The integration of remote sensing and Geographic Information Systems (GISS) has received considerable attention in the literature P.L.N. Raju [ 59], The handling of spatial data usually involves processes of data acquisition, storage and maintenance, analysis and output. For many years, this has been done using analogue data sources and manual processing the introduction of modern technologies has led to an increased use of computers and information technology in all aspects of spatial data handling. The software technology used in this domain is Geographic Information Systems (GIS). GIS is being used by various disciplines as tools for spatial data handling in a geographic environment [60].

VII. REFERENCES [1]

[2]

[3]

M. Supply chain management and crop insurance: [4]

The CMM gives the appropriate scenario of crop yield this based on these parameters demand supply chain can be established. This will reduce the delay of reaching the crops to any geographic territory. CMM generated parameters can be used by a government or agriculture insurance agencies based on these parameters they can allocate their financial budget.

[5]

[6]

N. General purpose graphic processing unit: When any image processing applications run on normal computers they take huge time for processing the data, the provided data is enormous in magnitude. To increase the processing speed we need to adapt parallel processor architecture. The best solution is to use NVIDIA Tesla

[7]

8

Donald W Larson, Eugene Jones, R.S Pannu, R.S Sheokand, Instability in Indian agriculture - a challenge to the Green Revolution Technology, Food Policy, Volume 29, Issue 3, June 2004, Pages 257-273 Roddam Narasimha, Science, Technology and the Economy: An Indian Prespective, Technology in Society, Volume 30, Issues 3-4, August – November 2008, Pages 330-338 Katinka Weinberger, Thomas A. Lumpkin, Diversification into Horticulture and Poverty: A Reseach Agenda, World Development, Volume 35, Issue 8, August 2007, Pages 1464-1480 Todoroff P, De Robillard F, Laurent J B, Interconnection of a Crop Growth Model with Remote Sensing data to Estimate the Total Available Water Capacity of Soil, Geoscience and Remote Sensing Symposium, 2010, Proceedings, IGARS, 2010 IEEE International Pages 1641-1644 Sudhanshu S Panda, Gerrit Hoogeboom, Joel O. Paz, Remote Sensing and Geopatial Technological Applications for Site-Specu\ific Management of Fruit and Nut Crops: A Review, 2010 Remote Sensing Journal open access ISSN 2072-4292, Pages 1973-1997 Guerif M, Launay M, Duke C, Remote Sensing As A Tool Enabling The Spatial Use of Crop Models for Crop Diagnosis and Yield Prediction, Geoscience and Remote Sensing Symposium, 2000, Proceedings, IGARSS 2000 International Volume: 4, Pages 1477-1479 K. Usha, Bhupinder Singh, Potential Applications of Remote Sensing in Horticulture – A Review, Scientia Horticulture, Volume 153, 4 April 2013, Pages 71-83

[8]

[9] [10] [11]

[12]

[13]

[14]

[15]

[16]

[17]

[18]

[19]

[20] [21] [22] [23] [24] [25]

[26]

[27]

Guerif M, Launay M, Duke C, Remote Sensing As A Tool Enabling The Spatial Use of Crop Models for Crop Diagnosis and Yield Prediction, Geoscience and Remote Sensing Symposium, 2000, Proceedings, IGARSS 2000 International Volume: 4, Pages 1477-1479 Jay Gao, Digital Anaysis of Remotely Sensed Imagery; by The McGraw-Hill Companies ISBN: 978-0-07-160466-6, Pages7,255 Jian Guo Liu, Philippa J. Mason, Essential Image Processing and GIS for Remote Sensing, 2009 by John Wiley & Sons Ltd, Pages22-27 YanMin Shuai, DongHui Xie, Suhong Liu, Uncertainty Analysis of Spectra Simulated Using Cropmodels Driven by Remote Sensing Data Observed in the Field, Geoscience and Remote Sensing Symposium, 2004, Proceedings, IGARS, 2004 IEEE International Volume: 4, Pages 2389-2392 Ashok Mishra, Christian Siderius, Kenny Aberson, Martine Van der Ploeg, Jochen Froebrich; Short- Term Rainfall forecasts as a soft adaption to Climate Irrigation Management in North-East India, Agriculture Water Management, Volume 127, 2013, Pages 97-106 V. Radha Krishna Murthy, “Crop Growth Modeling And Its Applications In Agricultural Meteorology”, Satellite Remote Sensing and GIS Applications in Agricultural Meteorology Proceedings of the Training Workshop 7-11 July, 2003, Dehra Dun, India., page no 235-263. V.K. Dadhwal, ”Crop Growth And Productivity Monitoring And Simulation Using Remote Sensing And GIS”,, Satellite Remote Sensing and GIS Applications in Agricultural Meteorology Proceedings of the Training Workshop 7-11 July, 2003, Dehra Dun, India., page no 263-291. Heng Dong, Qiming Qin, Lin You, Xinxin Sui, Hongbo Jiang, Jinliang Wang, Haixia Feng, Hongmei Sun, Models for Estimating Leaf Area Index of Diffent Crops Using Hyperspectral Data, Geoscience and Remote Sensing Symposium(IGARS), 2010, Proceedings, IEEE International Pages 3283-3286 Pinaki Mondal, Manisha Basu, Adoption of precision agriculture technologies in India and in some developing countries: Scope, presentstatus and strategies, Progress in Natural Science, Volume 19, Issues 6, 10 June 2009, Pages 659-666 Pinaki Mondal and V.K. Tewari, Present Status of Precision Farming: A Review. International Journal of Agricultural Research, 2007. Pages 110, Naiqian Zhang, Maohua Wang, Ning Wang, Precision Agriculture – A Worldwide Overview, Computers and Electronics in Agriculture, November 2002, Volume 36, Issues 2-3, Pages 113 – 132 Santhosh K Seelan, Soizik Laguette, Grant M Casady, Geroge A Seielstad; Remote Sensing Application for Precision Agriculture: A Learning Community Approach, Remote Sensing of Enviroment, Volume 88, Issues 1-2, November 2003, Pages 157-169 Lillesand, T., M., and Kiefer, R., W., 1994, Remote Sensing and Image Interpretation, Wiley,3rd ed., ISBN: 0471577839. Campbell, J., B., 1996 Introduction to Remote Sensing, 2nd edition. Guilford Press, New York. Sabins, F., F., 1997, Remote Sensing: Principles and Interpretation, 3rd Ed. W.H. Freeman and Company, New York, 494 pp Ulaby, F., T., Moore, R., K. and Fung, A. D., 1981: Microwave Remote Sensing: Active and Passive, Volume 1, Microwave Remote Sensing Fundamentals and Radiometry, (Reading, MA: Addison-Wesley). D. Landgrebe, “Hyperspectral Image Data Analysis,” IEEE Signal Process. Mag. 19 (1), 2002, Pages 17–28. W.E. Easias, M.R. Abbott, I. Barton, O.B. Brown, J.W. Campbell, K.I. Carder, D.K. Clark, R.H. Evans, F.E. Hoge, H.R. Gordon, W.M. Balch, R. Letelier, and P.J. Minnett, “An Overview of MODIS Capabilities for Ocean Science Observations,” IEEE Trans. Geosci. Remote Sens. 35 (4), 1998, Pages 1250–1265. Rao, A. R., and Jain, R., 1998. Knowledge Representation And Control In Computer VisionSystems. IEEE Expert System and their Applications, pp 65-79. Gonzalez, R. C., and Woods, R. E., 1992. Digital Image Processing. Reading, Addison Wesley Publishing Co.

[28] Wilkinson, G., G., 1996a, Classification algorithms – where next? In E. Binaghi. P. A. Brivio and A. Rampani (eds). Soft Computing in Remote Sensing Data Analysis, Singapore: World Scientific, pp. 93-99 [29] Townshend, J., R., G., 1992, The European contribution to the application of remote sensing to land cover classification. International Journal of Remote Sensing 13, 1319-1328. [30] H. Bischof, W. Schneider, and A. J. Pinz, “Multispectral classification of Landsat-images using neural networks,” IEEE Transactions on Geoscience and Remote Sensing, vol. 30, no. 3, pp. 482-490, May, 1992. [31] Benediktsson, J. A., Sveinsson J. R., Briem G. J., 2003, Multiple Classifiers for Multisource Remote Sensing Data, Department of Electrical and Computer Engineering, University of Iceland. [32] Aschbacher, J., Pongsrihadulchai, A., Karnchanasutham, S., Rodprom, C., Paudyal, D., R., and Toan, T., L., 1995. Assessment of ERS-1 SAR data for rice crop mapping and monitoring, Proceedings of IGARSS ’95, Florence, Italy, pp. 2183-2185. [33] Trout TJ, Johnson LF, Gartung J (2008) Remote sensing of canopy cover in horticultural crops. Horticultural Science 43, 333-337. [34] Teillet P.M., Gauthier R.P., Chichagov, A. and Fedosejevs G., 2002, Towards integrated Earth Sensing: Advanced technologies for in situ sensing in the context of Earth Observation, Canadian Journal of Remote Sensing, 28(6), The Canadian Aeronautics and Space Institute, Ottawa. [35] Panda, S.S.; Hoogenboom, G.; Paz, J. Distinguishing blueberry bushes from mixed vegetation land-use using high resolution satellite imagery and geospatial techniques. Comput. Electron. Agr. 2009, 67, 51-59. [36] Izzawati, E.D.; Wallington, E.D.; Woodhouse, I. Forest height retrieval from commercial X-band SAR products. IEEE Trans. Geosci. Remote Sens. 2006, 44, 863-870. [37] Guerif M, Launay M, Duke C, Remote Sensing as a Tool Enabling the Spatial Use of Crop Models for Crop Diagnosis and Yield Prediction, Geoscience and Remote Sensing Symposium(IGARSS), 2000, Proceedings, IEEE International Volume 4, Pages 1477-1479 [38] Martin, R.D. and Heilman, J.L,1986,Spectral reflectance patterns of flooded rice. Photogrammetric Engg. And Rem.Sens.,:1885-1890. [39] Berg ,A. and and Gregorie,J.M. 1982, Use of remote sensing techniques for rice production forecasting in West Africa (Mali and Guinea:Niger-Bani Project) [40] Vyas, S.P.,Manjunath,K.R.,Kalubarme,M.H. and Gupta,P.C.,1993, Remote sensing of dry land farming pulse crop in Hamirpur district of . P.proc.Nat.Symp.Rem.Sens.App.Res.Management,2527,November,Guwahati, Pages314 - 320 [41] S.C. Park, M. K. Park, and M. G. Kang, “Super-resolution image reconstruction: a technical overview,” IEEE Signal Processing Magazine, vol. 20, no. 3, pp. 21-36, 2003. [42] Elad, M.; Feuer, A.; “Restoration of a single superresolution image from several blurred, noisy, and undersampled measured images,” Image Processing, IEEE Transactions on , vol.6, no.12, pp.1646-1658, Dec 1997. [43] Jian Sun; Zongben Xu; Heung-Yeung Shum; , “Image super-resolution using gradient profile prior,” Computer Vision and Pattern Recognition, 2008. CVPR 2008. IEEE Conference on , vol., no., pp.1-8, 23-28 June 2008. [44] J. Yang, J. Wright, T. Huang and Y. Ma “Image Super-Resolution via Sparse Representation” in IEEE Transactions on Image Processing, 2010. [45] D. Glasner, S. Bagon, and M. Irani. “Super-resolution from a single image,” In Proc. of ICCV '09. IEEE Computer Society Press, 2009. [46] William T. Freeman, Thouis R. Jones, Egon C Pasztor, “Example-Based Super-Resolution,” IEEE Computer Graphics and Applications, v.22 n.2, p.56-65, March 2002. [47] Peter M. Atkinson, “Issues of Uncertainty in Super-Resolution Mapping and the Design of an Inter-Comparison Study”, Proceedings of the 8th International Symposium on Spatial Accuracy Assessment in Natural Resources and Environmental Sciences Shanghai, P. R. China, June 2527, 2008, pp. 145-154

9

[48] P. M. Atkinson, "Super-resolution target mapping from soft classified remotely sensed imagery," Photogrammetric Engineering and Remote Sensing, vol. 71, 2005: 839-846 [49] Ehlers, M.: Spectral Characteristics Preserving Image Fusion-based on Fourier Domain Filtering.In: Proc. SPIE, vol. 5574,2004, pp. 1–13 [50] Wald, L., Ranchin, T., Magolini, M.: Fusion of Satellite Images of Different Spatial Resolutions – Assessing the Quality of Resulting Images. Phot. Eng. & Rem. Sens. Vol.63, pp. 691–699(1997) [51] C.Pohl et al, - Multisensor Image Fusion in Remote Sensing: Concepts, Methods, And Applications --International Journal of Remote Sensing,1998, Vol.19,No.5,pp.823-854 [52] Xiong-Mei Zhang et al- Contourlet-Based Fusion Algorithm and Its Optimization Objective Image Quality Metrics- Proceedings of the 2007 International Conference on Wavelet Analysis and Pattern Recognition, Beijing, China, 2-4 Nov. 2007. [53] Boris Zhukov et al; Unmixing-based Multisensor Multiresolution Image Fusion- IEEE Transactions on Geoscience and Remote Sensing, Vol. 37, No.3, May 1999. [54] Yi Yang -A Novel Image Fusion Algorithm Based on IHS and Discrete Wavelet Transform - Proceedings of the IEEE International Conference on Automation and Logistics-August 18 - 21, 2007, Jinan, China. [55] Yanbo Huang, Yubin Lan, Steven J. Thomson, Alex Fang, Wesley C. Hoffmann, Ronald E. Lacey, “Development of soft computing and applications in agricultural and biological engineering”, Computers and Electronics in Agriculture, Volume 71, Issue 2, May 2010, Pages 107127 [56] Thomson, S.J., 1998. “Expert systems for self-adjusting process simulation”, In: Curry, R.B., Peart, R.M. (Eds.), Agricultural Systems Modeling and Simulation. Marcel D ekker, New York, pp. 157–195. [57] Rao, A. R., and Jain, R., 1998. Knowledge Representation And Control In Computer VisionSystems. IEEE Expert System and their Applications, pp 65-79. [58] Ehlers, M., G. Edwards, and Y. Bedard, 1989. Integration of remote sensing with geographic information system: A necessary evolution, Photogrammetric Engineering b Remote Sensing, 55(11):1619-1627. [59] P.L.N. Raju, ” Fundamentals Of Geographical Information System”,, Satellite Remote Sensing and GIS Applications in Agricultural Meteorology Proceedings of the Training Workshop 7-11 July, 2003, Dehra Dun, India., page no 103-121 [60] Longley, P. A., Goodchild, M. F., Maguire, D. J., and Rhind, D. W. (2005). Geographic Information Systems and Science (2nd ed.). Chichester, West Sussex, England: John Wiley and Sons. [61] Yi Yang -A Novel Image Fusion Algorithm Based on IHS and Discrete Wavelet Transform - Proceedings of the IEEE International Conference on Automation and Logistics-August 18 - 21, 2007, Jinan, China. [62] Hopson B, Benkrid K, Keymeulen D, Aranki N, Real-Time CCSDS Lossless Adaptive Hyperspectral Image Compression on Parallel GPGPU and Multicore Processor Syste, NASA/ESA Conference on Adaptive Hardware and Systems (AHS), Nuremberg, Germany, June 2528, 2012 Pages 107-114 [63] A. Joshi, A. Phansalkar, L. Eeckhout, and L. John, Measuring Benchmark Similarity using Inherent Program Characteristics, IEEE Transactions on Computers, Vol.55, No.6, 2006. [64] NVIDIA CUDA™ Programming Guide Version 2.3.1, Nvidia Corporation, 2000

10