Research On The Design Of Appearance Quality Parameters Select System Of Soybean Based On DSP

The twenty-first century,Chinese soybean productionin the world share decreased year by year, the main reason is that,domestic soybean acreage is relatively small,and soybean quality is relatively poor; "Hybrid, mixed income, mixed storage, mixing sell" phenomenon has become the most prominent problem.Tiered storage can’t be based on the quality of soybeans,Seriously affect the economic development of the soybean industry.So accordingto the quality of soybeans how to grade,which is the important issues. Implementing nondestructive testing to the quality of appearance of soybean which the main problem lies in the quality of soybean extraction, so the appearance parameters collection becomes the key.Visual inspection of agricultural products have been an important direction of research in the field of machine vision,traditional soy appearance grading work mainly through artificial means and cleaning the screen,among the artificial way is that staff grading process with the naked eye,constrained by physical and mental condition,the efficiency is low;To some extent to reduce the workers’ working pressure impurity screen appears.However, this approach can only be a single soybean impurities removed, and does not recognize soybean varieties and diseases.In recent years, Scholars at home and abroadonly rely on the PC and Matlab software visual inspection of agricultural research in the field,using digital cameras to capture images,further to its analysis and processing,this mode of operation can’t meet the needs of the automation hierarchy.From "static checking" to "dynamic detection" has become an inevitable trend of soybean appearance quality detection.DSP processor has low power consumption, fast processing speed, high precision operation advantages, this study uses TMS320DM6437 processor,builds a peripheral circuit.As the third-generation digital multimedia processor among TI company’s TMS320DM6437, With enhanced DMA controller, Convenient image data transfer and format conversion information, Some corresponding peripheral also includes Ethernet MAC、UART、I2C、SPI、GPIO、McASP and three PWM and so on. Powerful peripheralsenough to meet the system research applications. In this study, TMS320DM6437 processor as the core.Under a certain lighting conditions, Soybean image acquires by CCD camera,Soybean image preprocessing, and finally complete the acquisition of soybeans appearance quality characteristic parameters.In order to reduce external environmental disturbances, soybeans image acquisition process carried out in the black box.According to the focal length CCD camera, adjusting shooting distance, and designing size camera obscura. 10 varieties of soybean images were collected,which were normal soybean varieties, moldy soybeans, gray leaf spot of soybean, soybean broken, worm-eaten soy,did large number of experiments,Obtained sufficient experimental data, provided data support for the variety and quality recognition PC.In the experiment processing,Soybean image filtering section,image after gradation processing uses Wiener filter and morphological filtering, respectively.After the level of the filtered image by wavelet transform,andfusion process complete denoising work.The experiments show that,this filtering approach compared to the more traditional single filter,a better place for the noise, meet the system requirements.Experiment with National Soybean Engineering Technology Research Center from the Northeast Agricultural University offers 10 varieties of soybeans for the study.Respectively, for the normal soybeans, soybean diseases(ash spot disease, and insect Eclipse grain, and mildew disease, and broken grain) of geometry features(area, and long axis, and short axis, and perimeter, and round degrees, and partial heart rate, and rectangle degrees), and color features(R component are value, and G component are value, and B component are value, and R component three order moments, and G component three order moments, and B component three order moments, and R component marked poor, and G component marked poor, and B component marked poor, and H component are value, and S component are value, and V component are value, and H component three order moments, and S component three order moments, and V component three order moments, and H component marked poor, and S component marked poor, and V component marked poor), and texture features(are value, and three order moments, and standard poor, and consistency, and smooth degrees, and entropy)total 31 features parameter for extraction, for upper machine of varieties recognition and disease recognition provides full powerful of data based.