FPGA-Based Design Of A Mature Tomato Machine Vision Recognition System

With the development of new agricultural production models and new technologies,agricultural picking robots have gradually become a new trend in the development of agricultural production.Automatic detection and identification of crops through machine vision technology has become one of the key technologies in the design of picking robots,which directly determines the picking effect of robots and the economic efficiency of farms.The FPGA has high speed parallel computing capability,low power consumption and compact structure,which can meet the high real-time requirements,so this paper uses FPGA core processor to implement the design of machine vision recognition system.This paper designs a machine vision recognition system with mature tomatoes as the research object.The main research contents are as follows.First,the visual detection and recognition algorithm of ripe tomatoes is studied,and color image segmentation,motion target detection,and red-ripe tomato recognition are performed on tomato images.In order to separate the red-ripe tomatoes from the background in the actual scene,the extra-red and extra-green differential index are proposed as the color image segmentation algorithm through experiments.To simulate the motion of the picking robot,the motion target detection of the segmented tomatoes is performed by the inter-frame differential method,and a suitable threshold value is determined by simulation.The detected motion targets are framed through the minimum border to achieve automatic recognition of red-ripe tomatoes for subsequent operations by the picking robot.Next,a machine vision recognition system for ripe tomatoes was designed with FPGA as the main control chip.The tomato image information is captured by OV5640 camera and configured using Verilog language to output image data in 30 fps frame rate and RGB565 format.For engineering implementation,the extra-red and extra-green differential index are simulated and verified on Modelsim,and the extra-red and extra-green differential index and inter-frame differential method are implemented on FPGA by implementing ping-pong operation on SDRAM for image data caching.For the false detection of moving targets caused by lighting,the noise of non-targets is removed using open operation,the target tomatoes in the red-ripening stage are framed by the minimum border,and the recognized results are displayed on the VGA display in the display mode of 640*480@60 to achieve automatic detection and recognition of ripe tomatoes.Finally,the recognition effects of super-red and super-green differential index segmentation effect,inter-frame differential method with different thresholds,and minimum enclosing frame with video stream superposition method are verified after simulation and practical tests.And the automatic detection and recognition of ripe tomatoes in unobstructed and obscured environments verifies the correctness of the system design.