Image Processing Algorithm Research, Food (shrimp) Sorting System

The intelligent sorting system of food is a special device, which covers the technology of photoelectricity, mechanism, image processing, etc. In the food processing production line, the sorting system can classify materials automatically. This sorting system has a high degree of automation, security, accuracy and efficiency. It liberates labor from the traditional hand-sorting, and improves productivity greatly. At present, many domestic enterprises directly import foreign equipments. However, there is a wide variety of China's food industry, meantime, imported equipment can not meet the needs of domestic enterprises, and the equipments are relatively expensive. Under the circumstances, it urgently needs to develop our own food sorting system to meet the different needs of the food industry.This paper attempts to use methods of image processing and pattern recognition to achieve separation of the shrimp. The article first introduces the image segmentation algorithm, including the background color set based segmentation, adaptive threshold and contour-based segmentation algorithm. Then it presents the method of image feature extraction, the extraction of the features include:HSV color space based on color histogram and the gradient-based shape feature. Then we calculate the average sorting accuracy of two features to determine the characteristics of linear convergence factor. Next introduced the classifier design, k-nearest neighbor classifier is chosed to classify and the classification results are given in simulation platform. Then it comes to introduce the systematic application of sorting algorithm, which includes the hardware platform structure, the algorithm improvement based on hardware feature, and other works for system development. Finally, the paper gives out the practical application of the sorting performance, the genuine rate of genuine basket is 91.1%, the debris rate of debris basket is 77.6%. we can infer from the sorting results the algorithm is stable and reliable, it achieves the desired goal basically.