Research About Quality Grading Of Corn Seeds Based On Digital Image Processing

Quality grading of corn seeds is an important process of post-harvest processing. Corn seed quality can be graded by using the machine vision technology in a damage-free,high-accuracy,and high-speed manner,which can replace large amount of repeated manual work.This paper analyzes the corn seed grading method based on digital image processing,with corn seed quality grading as the objective and corn seeds as the analysis target.Considering the limitations in traditional corn seed quality testing,this paper extracts six eigenvalues for statistics analysis,including circumference,area,circular degree,rectangular degree,stretch degree,and hue.Approach seeds grading from size, shape,saturation,color and other multi-angle,and uses MATLAB software to grade seeds and validates the grading via experiments,then proves the feasibility through the use of computer vision to replace human to achieve the seed corns' automatic classificationDue to limitations of experimental conditions,during image acquisition,black is used as the photography background color,which varies greatly from the color of corn seeds,to avoid errors in the selected eigenvalues.10 pictures are acquired for each seed to get the average value of the eigenvalues.Totally 2820 pictures are acquired,covering 282 corn seeds.During image bottom information processing,the image of the seeds is enhanced according to the grayscale distribution of corn seed images.The enhanced image is compared with the effect after binarization,to lay emphasis on the significance of image enhancement.As for image smoothing,after a comparison analysis with the neighborhood averaging method,the median filter algorithm is used for de-noising.As for image segmentation,the edge detection algorithm is abandoned due to its poor effect while the thresholding segmentation is used.Of the multiple thresholding algorithms,the maximum entropy and corrosion segmentation algorithm are used,to achieve better effect,after a comparison with discrimination analysis method.After that,contour tracking is conducted and the effects are satisfactory.This paper uses the corn seed quality grading methods based on membership function and BP neutral network(quality grades are classified into good,moderate, and poor).After a comparison between the experiment process and experiment results, it is found that the corn seed quality grading based on BP neutral network is more favorable for real-time grading,to be better used in industries.In the core seed quality grading system based on BP neutral network,the eigenvalues of 30 seeds are taken as samples for network experimenting and 100 seeds are tested.System testing results achieve 97%accuracy in quality grading.The quality grading results can accurately describe the quality of the seeds.