| Along with the development of the computer technology and information technology, it becomes an inexorable trend to identify agricultural products automatically. using computer vision to automatic detection of the quality of the seed. Computer automatic recognition accuracy depends on the characteristics of parameters and identification model. This article based on domestic and international research results, choosing eight corn varieties such as Zhengdan 958, Liaodan 565, Jingke 25 and etc, these corn as the object of study on corn seed varieties based on image recognition technology. First of all, preprocessing the corn seed images. Using the method of median filter to remove noise, using the gray threshold segmentation method for background segmentation, boundary tracking and extraction method to extract single corn seed image, the results show that the algorithm can be very good to get the edge image of corn seed, then it laid a foundation for follow-up feature parameter extraction. Secondly, we extract the corn seeds whole and local feature. Analysis of corn seed overall external feature extraction, selection of the maize seed color parameters and configuration parameters. Extracting color parameters in RGB and HIS color space , morphological parameters including the size and geometrical parameters. Using independent component analysis algorithms to segment the endosperm and the crown portion of corn seed, and extracted the color features and area features from the endosperm portion and the crown portion. Then, optimizing the feature parameter . Using genetic algorithms to optimize the original characteristics of corn seed , using principal component analysis method to choose the characteristic parameters of the dimension reduction. Finally, using genetic algorithm and neural network to identificate the corn seeds. Selection of fitness function, Using genetic algorithm to determine the initial weights of the neural network , and using the neural network classifier to build corn varieties to improve the accurate rate of the recognition of corn varieties.This article based on Visual C++ 2008 development environment, combined with the MATLAB7.0 image processing function, designed and implemented a corn image recognition system, this system can complete the corn seed image denoising, segmentation, extraction of single corn seed, extraction corn seed characteristic parameters and corn breed recognition. |
