The Study Of Algorithm Of Potato Surface Defect Detection Based On Machine Vision

In recent years,we advocate digital agriculture, the shadow of "digitization" is widely found in the farmand farmyard of country. Automation of agricultural production is not only bring high-yield growers surprise,but also bring direct economic benefits for consumers. Potato automated detection of surface defects is animportant agricultural production automation components. In this study, the possible defects (green grouper,decay, mechanical damage) of potato in harvest is our target. In the based on study of the recognition algorithmnow, carried out on three kinds of defect recognition algorithm, more suitable detection algorithms areexplored. Its purpose is maximize the reduction of the human factor in the process of potato quality testingerror. Contents and results of this study are as follows:(1) Collecting and preprocessing potato image. The image acquisition system include Canon A490digitalcamera and the stand of camera and a white background plate bracket is used for image acquisition, the imagewould be used in the study that the RGB color image of normal potato and the potato with defects will be got.(2) Extract Green grouper and decay and mechanical damage defects characteristic factor. Analysecharacteristics of defects of green grouper and decay and mechanical damage, these characteristicparametersthe are chosed: selected channel image of R and G values of all pixels, and the ratio R/G, G and Bchannels of all image pixel values and the ratio G/B as the characteristics of defects potato livedo parameters;selection in21to60and61to80within two grayscale pixel distribution probability P1, P2as thecharacteristic decay defect detection operator; choose defective parts minimum bounding rectangle shapecontour length and width of the DK DC ratio P as mechanical damage defects characteristic operator.(3) Identificaate the defect of green grouper, rot and mechanical damage. Using Fisher classifier, thecharacteristic values of defect of green grouper and decay defect as input samples, obtained discriminantfunction. When the same characteristic values is put into the function, as long as the set of characteristicswhich determine the magnitude relationship between the0value, it will recognize the characteristics of thisgroup of which the potato is a potato or with healthy green spots and potato rot defects. The use of mechanicaldamage defects eigenvalues and a comparison of the relationship, you can determine whether a potato with amechanical damage defects.Experimental results show that the algorithm for the three defect recognition ratewas93%,90%and87.5%, reaching agricultural production automation requirements. The identificationalgorithm has the simple, efficient features, can solve the existing problems of classification algorithm,providing technical support for the quality of products of the food processing industry.