Study On Granular Crops Color Sorting System Based On Machine Vision

Machine Vision Sorting System is a smart grouping equipment concerned with light, electric, machine and image processing. In general, the system can fall into two categories, namely the Black-and-white Sorting System and the Color Sorting System. Nowadays, the Color Sorting System generally adopts computer as the kernel of image processing, but still has flaws such as big size and low selection efficiency. However, with the development of national economy, people are indicated to have increasingly high requirements about the resolution ability, identification accuracy, and selection efficiency of Color Segmentation System. Fortunately, the Embedded System is featured by high integration, good real-time performance, low power consumption and high reliability. This thesis is aimed to study the Embedded Color Segmentation System based on Machine Vision. Applied in the Color Segmentation System, the Embedded System will unfold wide prospects.This study takes color linear array CCD camera of DALSA as sensor and FPGA XC3S500 E and DSP TMS320DM6437 as processing core to design the image acquisition unit, image processing unit, image representation unit and selection operation unit, thus establishing the embedded platform for Color Segmentation System. Furthermore, the FPGA IP kernel technique is adopted to construct dual-port RAM, realizing separate control on the storage and read of image data. At the same time, the EMIF communication interface between FPGA and DSP is constructed and their data can be exchanged with the help of VHDL programming. Finally, the VGA display circuit driver is programmed to display the color image acquired by the camera at resolution of 640 ? 480.In this study, the author takes the broken peanut as subject, to study image processing algorithm by the application of simulation tools, including identifying the broken area and peanut body by statistics and analysis of different pixel value in the image through RGB color model, suppressing pulse noise in the image by studying different image filtering algorithm and improving the median filtering algorithm, removing the point and hole pulse and smoothing the edge of subjects through morphological algorithm, inspecting the edge of peanut to identify the peanut body and calculate the broken area of each peanut by eigh t connected domain chain code algorithm, and at last counting other geometrical parameters of the peanut such as its perimeter, area, major and minor axis and centroid.This study transplants and tests image processing algorithm in Embedded Machine Vision Sorting System. The author does not only establish a CCS project, but also configures the project files. Meanwhile, C is adopted into the program of broken area inspection, median filtering, morphological algorithm and the track of image outline. The program is also used to process the image collected by CCD camera and each process has been valued by its effect and time. Finally, the author implements two algorithms to inspect breakage and analyzes the applicable scope according to the operation results.