The Research Of Color Sorting Machinetechnology Based On DSP

With the improvement of the people living standard, demands on quality of riceare gradually upgrading. Machine vision technology is applied in agriculturalindustry to improve the quality and efficient of the rice detection. Therefore, thecolor sorting is created for this. The domestic and international color sortingtechnology is used for reference and the visual inspection portion in color sorting isstudied and designed in this paper.Firstly, the hardware platform of visual inspection and processing system isbuilt in this paper. It includs optical part, FPGA hardware part, the driver part oflinear array CCD, signal processing part, DSP haredware part, power supply part.FPGA is chose to be the main hardware controller and to realize image preprocessing.DSP is used to finish the image processing. The advantages of both are used toachieve the quick detection purpose of the unqualified rice.After space image filtering algorithms are compared, median filtering algorithmis used to solve the noise problem during the process of capturing image with CCD.The image is blurred due to rice relative CCD in motion. Fast uniform linear motionrestoration algorithm suitable for discrete image and FPGA hardware design ispresented in this paper. The restore effect is efficeint with this restoration algorithmafter MATLAB simulation. Restoration algorithm and improved median filteringalgorithm is coded with Verilog language and simulated with Modelsim-Altera. Then,the function of restoration and filter is realized with FPGA.The data after prepropoessing are transmitted to DSP to extract the objectiveand detect four edges and find the belonging rice for each pixel background with thedifferential edge detection algorithm sutiable for the detection of dynamic discreteimage. Finally, thresholding algorithm is applied inside DSP to identify theunqualified rice. The rice is identified as the unqualified rice while the area exceeds the standard. Then the position information is calculated and sent out. The algorithmis programmed with C language and simulated with CCS3.3software. The resultshows that the unqualified rice can be identified efficiently and quickly.