Document not found! Please try again

Implementation of Line Tracking Algorithm using Raspberry Pi in

0 downloads 0 Views 537KB Size Report
the transmission lines and proper navigation, we have used a masking technique .... the Number Of Frames per seconds (FPS) of raspberry pi is approximately 5 ...
Implementation of Line Tracking Algorithm using Raspberry Pi in Marine Environment Samreen Amir,

Ali Akbar Siddiqui, Nimrah Ahmed

Associate Professor, Hamdard Institute of Engineering & Technology, Hamdard University Karachi, Pakistan [email protected]

Electronic Engineering Department Sir Syed University of Engineering & Technology Karachi, Pakistan [email protected], [email protected]

Bhawani Shankar Chowdhry Dean Elect. Electronic & Computer Engineering Mehran University of Engineering & Technology Jamshoro, Pakistan [email protected] Abstract—In recent days it is necessary to maintain continuous surveillance of underwater transmission lines or oil pipelines. For such purpose, we require an underwater vehicle rover capable of tracking these wires or pipelines and detect the fault if it occurs. For this purpose we have designed an intelligent quad leg rover. Image processing as a key deployed for tracking and tracing the fault or damage. Hough’s Transformation is used for the detection of the wire, and threshold levels were also set of the underwater environment for vehicle to focus only on wire. For tracking the transmission lines and proper navigation, we have used a masking technique. The system is implemented on Raspberry pi (Broadcom BC2835) as well as Intel Core Processor T7250 to improve and analyze the performance in terms of size and mobility. The results presented in this paper are simulated on Intel T7250 processor and on raspberry pi. It helps in evaluating the response time of the raspberry pi when compared to any other processors in terms of computation and robustness.

Keywords- Hough Transformation, Raspberry pi, masking technique, thresholding

I.

INTRODUCTION

In modern era, underwater surveillance is important to keep a complete track of underwater pipeline, transmission lines, etc. For this purpose we require a rover proficient of tracking and tracing out the fault or damage on underwater transmission line and pipelines before they become cause of serious concerns. Rovers are commercially designed to perform survey, subsea inspection, construction and repair operations at modest (fewer than 100m) depths [1]. It is hard for the operator to recognize the pipeline from the images. Furthermore, the surveillance of underwater pipelines is monotonous and the operator can easily make mistakes due to the distraction and tiredness. Therefore, the study focuses on developing the automatic detection of pipelines by using image processing algorithms. Image processing is a growing area with applications such as

surveillance, medicine, pipeline detection, transmission line inspection, authentication, and many more areas. Applications such as these perform processes like object detection and image enhancement. The impact of image processing and the importance of their implementations on hardware are to attain enhanced performance. High performance can be obtained by various Image processing algorithms with high computation load in parallel processing capable co-processor such as FPGA [2]. Bing proposed an algorithm on predicting- verificationupdating, it can be implemented on real time lane detection on DSP based image processor, and it enhances the efficiency [3]. Bruce A Draper has implemented Image processing algorithm on to FPGA which can obtain 8 to 800 fold speed ups over Pentium [4]. Scientific community proposes to combine FPGA and DSP. In FPGA when computation becomes complex, no of gates increased, so DSP will be a wise choice for this purpose. Liu has combined DSP and FPGA for target position detection in real time [5]. Real-time image processing system is broadly used in numerous fields; it is required to have high speed. In order to satisfy image processing system structure based on DSP and FPGA, that is DSP is used as advanced image processing unit and FPGA as logic unit for image sampling and display [6]. DSP processors has less development cost, less power consumption and can implement complex algorithm easily rather than FPGA. FPGA are faster and they can hit DSP processors in terms of performance. Sometimes parallelism, fast speed or even less power consumption is not always the first priority, therefore we have performed image processing on Raspberry pi and T7250 Core processor to our proposed system. The Raspberry pi is a low cost single-board computer which has recently become very popular [7]. It is a new paradigm that takes advantage in some parameters like size and light weight. It is being used for children to teach computer programming and also can be used for many other purposes.

Previously, the detection and the real time tracking of submarine pipelines from a series of underwater images are performed by using a Kalman filter which improves the precision in the computation of the straight line equation with a little overload [9]. The inspection of underwater pipelines are performed by using the image processing method in which images are improved by Gabor filter, and then passed by an edge detector, and Hough transformation is used for the calculation of parameters. The Kalman filter is explored to predict the parameters of the pipeline on the next image [10]. Sonar based pipeline detection and tracking starts by the acquirement of the sonar image and further real time processing is performed. The path following method used is based on a combination of Lya- punov technique and PI controller acting over two horizontal thruster‟s dual torpedo AUV [11]. II.

Figure 1. The straight track

SYSTEM HARDWARE MODEL

Rover is a quad leg structure made up of plexi glass. Its dimensions are 0.4826x0.3048x0.23688 m. It is based on three gear motors controlled through a raspberry or Intel Dual Core Processor T7250-2GHz. Worm gear motor works in right angle configurations, it also provides high starting torques and can tolerate high shocks load. The rover will be fixed with light and camera to perform the task. Additional sensors and tools can be fitted as required for specific tasks. In our proposed system, we have implemented raspberry pi as control device. Raspberry pi is basically a single chip computer capable of performing various tasks and can also communicate with other devices such as PC, Televisions, etc. It has Broadcom BCM2835 700MHz ARM1176JZFS processor[8] with FPU and Video core 4 GPU, GPU provides Open GL ES 2.0, hardwareaccelerated Open VG, and 1080p30 H.264 high-profile decode, GPU is capable of 1Gpixel/s, 1.5Gtexel/s or 24 GFLOPs of general purpose compute and features a bunch of texture filtering and DMA infrastructure, 512MB RAM, Boots from SD card, supports Linux distros such as Fedora, Debian and Arch Linux, 10/100 wired Ethernet, HDMI output, USB 2.0 interface x 2, etc [12]. To integrate the system we have created a virtual oceanic environment. It was basically a 5x5x5 m water tank. Fig.1, Fig.2 & Fig.3 are showing the possible trajectory of a pipelining on a sea bed as a straight, turning left and turning right.

Figure 2. The pipeline turning left

Figure 3. The pipeline turning right

III.

SOFTWARE MODEL

We use MATLAB environment for Intel T7250 Core processor; since it requires high processing speed which in the case of Raspberry Pi (Broadcom BC2835) is not available. First the Camera was initialized and frame from a live video stream is acquired. Once the frame is acquired, we require adjusting the threshold level according to the underwater environment. The marine environment requires a special adjustment of the threshold level. The

intensity of light is different in scenarios. The Threshold value is selected as 0.5. A frame is then converted to a Binary forma with some redundant information. So these unwanted blobs and objects are removed from the frame through filtration. Then Hough‟s transformation technique is applied for the precise detection of the pipeline or transmission line. The masking maps or stencils for all three possible pipeline positions are developed, shown in Fig.4, Fig.5 & Fig.6.

acquired frame. The entire logic depends on the value of correlated output. If the correlated output is unity then it means the wire is in straight direction, and hence our ROVER will move in the same direction. If the value of the correlated output is not unity, then it means the wire is either turning left or turning right. The whole process will then be repeated for masking maps Fig.2 & Fig.3.The whole software flow chart is shown in Fig.7.

Figure 4. The masking map for straight track.

Figure 7. Software flowchart Figure 5. The masking map for left turn.

The Raspberry pi has limited hardware configuration as compared to Intel Core Processor T7250. Python is a useful tool for developing software on Linux Operating System. We have used its power along with simple CV library to overcome the hardware limitations of pi. The tracing and tracking methodologies are kept same as that for MATLAB. IV.

Figure 6. The masking map for the right turn.

OR operation is then applied to these maps and acquired frame. The result is then correlated with the

RESULTS & DISCUSSION

Refer to Fig.1 required line for our rover to follow is the straight line. Fig.4 is used for masking of straight line. Fig.8 is the bit formats for the straight track cases. The filtration is done using area opening method in which, it removes the connected objects from the image within given frames. Filtered image is shown in Fig.9. It will eliminate the unwanted objects, and blobs from the image and made the desired line or path completely clear for the

rover to track or even detect the damage if it exists.

Figure 8. Bit formats for the left cases.

Figure 9. The filtered image.

Fig.10, Fig.11 & Fig .12 represents the actual Hough‟s line transform for the straight, turns right and left movement. It also shows the peaks that were detected in our image for further accuracy. These peaks are represented by the small squares.

Figure 11. The actual Hough‟s transformation for straight direction.

Figure 12. The Hough‟s transformation for the image Turing left.

Figure 13. The Hough‟s transformation of Image turning right.

Fig.14 represents the corner detection of the desired straight track. Corners and non uniform border lines are detection is implemented. When the transmission lines are turning then Hough transformation lines are not shown. In such cases it will completely depend on our masking techneque.

Fig ure 14. The edge detection and non uniform borders

V.

PERFORMANCE ANALYSIS

Although both of our devices performed exactly according to our expetation, but due to the major difference between the capabilities of two processors (T7250 Processor and Pi processor i.e. BCM2835 700MHz ARM1176JZFS) is the difference of Frames per Second (FPS). Our T7250 is 2GHz Processor equipped with the 2Gbytes of RAM, whereas Raspberry Pi lacks such processing capabilities, it has 700MHz processor and is only equipped with 512Mbytes of RAM.The number of Frames of T7250 processor is approximately 30 FPS, and the Number Of Frames per seconds (FPS) of raspberry pi is approximately 5 Frames per Second. Although the Frame rate can be enhanced through more efficient programming. By computing the results, the FPS rate of T7250 Core Processor is much greater than that of Raspberry Pi (Broadcom BC2835). Although using Raspberry Pi will make our system much cheaper and compact, whereas T7250 Processor is much more bulky and costly. We require a compact hardware for our design for underwater system, so Raspberry Pi will be more suited for the task. Design features can vary depending on the hardware but FPS is a key feature for faster processing. With correlation results, we have examined that it will continue its direction, as its correlated output remains 1. For instance; if it is moving straight, left or right so it will be moving in that direction unless its value is changed by 1. Fig15 represents the correlated output with respect to time. Blue line represents correlated output of straight track, red for left and green for right. This plot represents that once straight line correlated output is unity the other correlated output of left and right track is not unity. We have implemented our proposed system for marine environment and compiled our results based on simulated output. Results were compiled with both T7250 processor and Raspberry Pi. The proposed system performed according to its expectation.

Figure 15. Correlated output

VI.

CONCLUSION

We have implemented our proposed system for marine environment and compiled our results based on simulated output. Results were compiled with both T7250 processor and Raspberry Pi. The proposed system performed according to its expectation. The Raspberry pi offers better size but less speed. Accuracy of both systems was similar even if the FPS rate is very different. Our algorithm can be implemented to almost any marine environment given the task for which it is designed for.

REFERENCES Lois l Whitcomb,” Underwater Robtics out of the research Into the Field,”, in International Conference on Robotics and Automation, IEEE 2000. [2] Viorela Simona, “VLSI architecture for motion estimation in underwater imaging,”PhD thesis, Universitat de Girona, June 2005 [3] Bing-Fei Wu, Chao-Jung Chen, Yi-Pin Hsu and Ming-Wei Chung, “A DSP-Based Lane Departure Warning System”, in Proc. of the 8th WSEAS‟06,2006, pp. 240-245 [4] Draper B.A., Beveridge J.R. , Bohm A.P.W. , Ross, C. , Chawathe M, “Accelerated Image Processing on FPGA,”.Image Processing 2003, IEEE Transactions ,Volume:12 , Issue: 12 . [5] Liu Yanjun ,Yan Haixia , He Xin , Wei Zhonghui ,“Digital Image Processing Based on DSP,” in Information Engineering and Computer Science ,ICIECS „10 ,2010, paper. [6] Duan Jinghong , Deng Yaling ,Liang Kun ,“Development of Image Processing System Based on DSP and FPGA ”.Electronic Measurement and Instruments, ICEMI '07. 8th International Conference. Page(s): 2-791 - 2-794 . [7] Sundaram G.S. , Patibandala B. , Santhanam H. , Gaddam S. , Alla V.K. , Prakash G.R. , Chandracha S.C.V. , Boppana S. , “ Bluetooth communication using a touchscreen interface with the Raspberry Pi” .Southeastcon, 2013 Proceedings of IEEE , Page(s): 1-4 . [8] (2013) BCM2835 Media Processor; Broadcom website. [Online] .Available http://www.broadcom.com/ [9] Tascini G. , Zingaretti P. , Conte G. , Zanoli S.M. ,“Perception of an underwater structure for inspection and guidance purpose .Advanced Mobile Robot”, 1996., Proceedings of the First Euromicro Workshop, Page(s): 24 – 28. [10] Calvo O. ,Sousa A. , Rozenfeld A. „“Smooth path planning for autonomous pipeline inspections” . Systems, Signals and Devices, 2009. SSD '09. 6th International Multi-Conference . [11] Jinbo Chen , Zhenbang Gong , Hengyu Li , Shaorong Xie ,“A detection method based on sonar image for underwater pipeline tracker” .Mechanic Automation and Control Engineering MACE, 2011 Page(s): 3766 - 3769 . [12] (2013) Raspberry Pi website. [Online] Available: http://www.raspberrypi.org/ [1]