A Product Sorting Technology Based On Color Feature

In the process of commercialization, products need to detection of classification and artificial detection is the main method. However, there are many disadvantages, such as strong subjective factors、 low efficiency and high testing cost. This paper based on such a background, a product sorting system based on color feature is designed. In order to accurately 、 rapidly and efficiently grading different grades of products by applying this system to the control structure of color sorter.The main several aspects work of this paper as follows:(1) Analysed color sorter internal parts: camera and light, and chose CCD camera and LED light to ensure that collect high quality products images.(2) The collected products’ images tend to exist noise and highlight areas,and they will affect the latter processing procedure. To reducing noise and removing highlight areas effectively, this paper puts forward an improved neighborhood average membership degree of fuzzy clustering algorithm to process the collected product images.(3) To experimental analyse the surface color information in product’s histogram after pretreatment. The results showed that product’s classification depend on a variety of surface color proportion together.(4) Using the product samples’ data to train back-propagation neural network and support vector machine based on binary tree separately. Then use them to classify products experiments respectively. The results showed that support vector machine based on binary tree has higher precision. In the end, did classification experiments to other products. And the experiments results showed that the system has high accuracy and universality researched in this paper.