The Research For Rice Appearance Quality Judgment Based On Machine Vision

Along with the rising living standard of the people, Rice quality plays an important role in people's diet, so it is more important for the detection and judgment of rice quality. In order to improve the efficiency of the monitoring system and the accuracy of recognition, machine vision technology has been widely applied to agricultural products industry. Therefore, machine vision technology is used to study the appearance quality of rice and the corresponding hardware and software system are established.In this paper, the basic principle of the machine vision technology and related references on rice quality are studied. Then the hardware and software of the rice appearance quality detection system are designed separately. In the part of hardware, we select the appropriate light source, the image sensor and background board, etc in order to get the image of high quality, little affected by external factors. In the part of software, at first we use both C++ and Open CV to realize the process including image acquisition, image preprocessing and image recognition, and then the PCA-BP algorithm is adopted to realize the judgment of rice quality.In order to obtain high quality rice images and extract feature data effectively, it is necessary to preprocess the image, including image enhancement, noise processing, background segmentation, chalkiness segmentation, etc. At the same time in order to mark a single grain of rice, if there is adhesion between two grain of rice, it needs to carry out the separation. According to the characteristics of rice image,the different algorithms are adopted to complete the image preprocessing, meanwhile the performance of different algorithms are analyzed, this lay good foundation for the back of the judgment.It is key to obtain the data of rice on the basis of image preprocessing. The parameters including shape, size and color are extracted from the rice image to detect the appearance quality. In order to reduce the complexity of the system and improve the efficiency of recognition, the combination between principal component analysis and BP neural network is proposed to build the recognition system. First, we use the principal component analysis method to extract the principal components of rice. And then the BP neural network is trained with the principal component as its input, Finally, the BP neural network fixed is used to recognition different types of rice. The two different algorithms are simulated by the MATLAB software. The simulation results show that the PCA-BP algorithm is greatly improved in accuracy and realtime in comparison with the BP algorithm.