Research On Texture Extraction And Recognition Of Rice Seeds Based On Morphology

In China, rice is one of chief crops in grain production. The quality of seed is an important factor influencing rice yield. Seed purity is the most important target of seed quality and have a significantly influence on crop yields. At present, the identification of rice seed variety mainly depends on chemical method and paddy field method, can come true the purpose of determining purity of seed, but all of them areachieved by manual determining, and some of them not only cost a lot of financial and material resources but also waste too much time. Therefore, finding a method that can determine rapidly and precisely is very important to country, agriculture and fanners as well as is necessary to achieving increasing production and harvest and agricultural modernization. Rice seeds texture shown genetic characteristics appear to significant different in rice populations. Effectively extracting and identifying rice seeds texture feature make sense of rice seeds individual recognition and researching on rice seeds texture inheritance. The paper selects t welve rice seed varieties as the objects for study, using these implements of computer, CMOS, Optical lens, electric control platform and automatically adjustable light source, synthetic applying image processing technology and multivariate statistical methods. Texture extraction and recognition of rice seeds based on morphology were studied.The main content and conclusions of this research were as follows:(1) The machine vision system for texture extraction and recognition of rice seeds based on morphology was established,which was composed of the HDL-I-type cold light source, the digital camera, the monocular microscope with Continuous zoom,the electric control platform and the computer. The whole system was portable and practical, and the good original images could be taken by it.(2) According as texture image is noisy and uneven illumination,the thesis discussed a variety of image pre-processing algorithms to improve image quality, combined with multi-focus image fusion of rice texture and presented several common image processing methods that can highlight rice seeds texture feature. Rice seeds texture feature was separated from the texture image by binary segmentation. The rice seeds texture can be easily observed and measured after the image was processed by mathematical morphology method.(3) Three traditional texture extracting method based on Dual-Tree Complex Wavelet Transform(DT-CWT), Gabor transform and Gray-Level Co-Occurrence Matrix(GLCM) were analyzed. At the same time, By comparing with the conventional LBP algorithm, LQP algorithm, ELBP algorithm and EQP algorithm for the texture feature of rice seeds the actual results were identified in this work.(4) Finally, an image processing analysis system of rice seed texture based on Visual C++platform was developed. The main function of the system included pre-processing single static rice seed image, extracting several rice seed texture features of entropy, energy, moment of inertia, local stationarity and the distribution of texture features by EQP, automatically recognizing varieties in rice seeds with the SVM-RBF classifiers. Experiments shown that it is feasible to establish rice texture recognition and is a new method of the use of surface texture information to identify rice seed varieties.Texture feature extraction and expression, Image fusion and recognition and Mathematical Morphology and so on are comprehensively applied, and varieties of rice texture feature recognition system are established. These works presents a new method of rice varieties according to rice seeds texture image.