EMBEDDED ELETRONIC (FPGA) APPLIED TO AGRICULTURAL IMAGE PROCESSING Rodrigo das Neves Bernardi (institucional studentship) /
[email protected] Advisor: Mário Luiz Tronco / Colaborator: Luciano Cássio Lulio Computer Science - University of São Paulo (EESC/ICMC) - São Carlos – SP KeyWords: Agricultural Robots, Image Processing, VHDL.
Introduction
Image 1. From left to right: original Image, filter with Gaussian efect, and Canny edge segmentation.
The design presents na solution embedded processing of agricultural images by means of techniques and algorithms implemented in an FPGA VHDL DE2-70 modification module, so that the response time is optimized, while the data rates are acquired in greater quantity and with many original features, unlike applications involving high-level processing with times of ups and unwanted response to image segmentation platforms. Are also used to Raspberry Pi platform and the DSP module controller card and image acquisition Blackfin (Surveyor SRV-1 Robot) for image processing in Python and use the OpenCV library dedicated to the Raspberry Pi GPIO control and specific programming languages to the Blackfin DSP, for testing and research for interconnection and performance between these devices.
The current tests are directed to implement algorithms in Python for the Raspberry Pi, which allows manipulation of the GPIO signals and motor drives the Surveyor SRV-1 robot, image capture (OpenCV) for microcomputer integrated camera and also via Blackfin board with the correspondences of image processing algorithms prototyping by VHDL.
Objetives Develop techniques for image processing in agricultural environments to the problem of autonomous mobile robot navigation in these environments, implemented on an architecture of FPGA hardware and embedded computing platforms.
Material and Methods The methods used are implementations of algorithms in VHDL and Python / OpenCV. The goals are stereo cameras modeling, preprocessing of the images and maps generating disparities; Prototyping in VHDL preprocessing, segmentation and image filtering; interconnection between devices as Raspberry Pi board, Surveyor SRV-1 Mobile Robot and DSP Blackfin camera modules; simulation and implementation of the embedded routines for testing in the rows system.
Results and Discussion In the current state, were customized edges segmentation techniques, as Canny in OpenCV, and their correspondence in VHDL.
XXVI Congress of Scientific Initiation
Image 2. SRV-1 robot (left) and Raspberry Pi board (right).
Conclusions Through the pre-images of the results obtained and processed, we observe the attempts of separation of segments / regions, which applies the locomotion of the mobile robot. With the improvement of integration techniques of image capture, preprocessing and filtering in Python / OpenCV and image segmentation in VHDL / FPGA, the project will meet the intended goals of autonomous locomotion, low response time, making alternatives application of digital image processing in embedded solutions and real-time in a practicable manner.
Thankfulness To advise and colaborator, for the commitment and assistence dedicated to elaboration of this work, and to the institution for the oportunity. ____________________ ¹ Idoeta, Ivan V.; Capuano, Francisco G. Digital electronic elements. 40.ed. São. Paulo: Érica, 2008 ² Brown, Stephen D.; Zvonko Vranesic, Fundamentals of digital logic with VHDL design. 3rd ed. United States: 2009. Appendix A, pg. 779. ³ Siegwart, R.; Nourbakhsh, I., Introduction to autonomous mobile robots. Massachusetts: 2004
