The Study Of Automatic Classification Algorithm For Rice Based On Machine Vision

Although the rice industry is the important to our country agricultural industrialization, now, most the measuring of appearance quality of rice in our country still relies on naked eyes, which lacks objectivity and repetitiveness. With the development of computer and the improving of image processing algorithms, machine vision technology has become an effective method to the detection for the appearance quality of rice.Automatic detection algorithm of broken rice rate,chalky rice rate and the identification algorithm for different kinds of rice is researched based on machine vision and National standards of rice.1)In order to optimize image acquisition conditions, light intensity and shooting angle which are the principal factors to the quality of rice image are studied, and the ratio of the rice in the segmentation image and the actual number in the rice image were used as the evaluation standards. An analysis of the collection of rice images in various of light intensity and shooting angle has been made . The results show that, in a black background, at the 90°of the shooting angle and the range of 40~50(Lux) of the light intensity can obtain the better effect.2)A novel method with touching rice grain based on erosion and dilation process is proposed in order to obtain each single rice. This method can obtain a better effect.3) After obtaining the single rice, the number of broken rice are accumulated based on the feature of area. This method can accurately determine the existence of the broken rice in the samples and obtain the rate of broken rice. The experimental results showed that in the state of different rate of broken rice and the connection between the rice, the novel method can achieve higher detection accuracy.4) Using double threshold method segment the gray-scale image in order to obtain the number of rice and chalky rice. The experimental results showed that in the state of different rate of chalky rice and the connection between the rice, the novel method can achieve higher detection accuracy.5)According to national standards, the grade of appearance quality of rice has been evaluated by using the results of the broken rice rate and chalky rice rate. The experimental results showed that the novel method can achieve the purpose of the detection.6) Removing the broken rice from the image using the feature of area, and extracting the appearance characteristics of the rice without broken rice. At last using the method of neural network optimized by fuzzy algorithm to identify four kind of the rice, the result of detection is 100%,83.33%,91.65%,100% respectively.