Study On Varieties Identification And Detection Of Maize Seeds Based On Machine Vision

Study on varieties identification and detection of maize seeds based on machine vision is one part of the task"Study on automatic detection technology of agricultural products based on computer vision"which is a science and technology program, provided financial aid by Education Office of Shandong Province. The system of maize seeds varieties identification and detection based on machine vision was established. The hardware selection system was set up. The software system of maize seeds varieties identification and detection was developed in VC++. The system can collect and analyse the images of maize seeds, extracting the geometric characteristics and color features of multiple seeds objects in maize images, identifing and detecting varieties of maize seeds.The image acquisition hardware system of maize seeds was designed based on the morphology and optical features of maize seeds, the image of seeds which is up to the mustard can be got by the hardware system.The color information of maize seeds in color space was analysed, the transform of color space and the convert of image format were verified, prepared for the extraction of geometry and color features.A series of image preprocessing algorithm such as image format conversion, image transform, image histogram information statistics, image enhancement, image segmentation and morphological processing were verified. After comparing with the algorithms, the image preprocessing programme of maize seeds was designed. Experiments showed that the image preprocessing programme can provide us the target image of maize seeds.A multi-object contour extraction algorithm was proposed, the algorithm has high accuracy and adaptability to geometrical feature extraction of maize seeds. The problem that the maize seeds image must be segmented to get single seed image for extracting the contours of maize seeds was settled. The track and extraction of the contours to a mount of seeds can be done simultaneously, and the efficiency of the system was improved.All character parameters of maize seeds were defined, a new data structure and vector sign arithmetic were introduced in the multi-object contour extracting algorithm. Multi-object geometric feature and color feature extraction algorithm were proposed based on the multi-object contour extraction algorithm in the seeds image feature extraction .The difficulty that extracting of multiple objects feature of seeds image once has been solved effectively,the precision and speed of extracting character parameters of maize seeds has been improved.In the aspect of maize variety identification and detection, established maize seed features bank of four varieties: Nongda 108, Zhengdan 958, Ludan 981, and Wuyue 18. The test of varieties identification were done to four varieties maize seeds including Nongda 108, Ludan981, Zhengdan 958, Wuyue 18,the recognition accuracy rates were 96.23%, 95.33%, 98.13% and 100% separately.