Study In Apple Grading Technology Of Image-Based Multiple Indice Fusion

With the improvement of agricultural automation and the pursuit of high quality of life,automatic apple grading technology has become a necessary means to improve efficiency and quality in apple production.Fuji Apple is one of the main item both here and abroad,and its automatic classification technology has important value to production.At present,the existing automatic classification technology of Fuji Apple is mainly used to deal with one of its single features.However,the single feature does not accurately reflect the comprehensive quality of apple.In this paper,a joint multi feature apple grading method based on computer vision is used.Therefore,in this paper,computer vision,pattern recognition technologies are applied to research and practice of multiple indice fusion automatic grading technology for non-contact Fuji Apple based on the main indicators used in the industry.This paper designs and implements an apple grading software system based on multiple indice fusion of images.The system processes in sequence,including image acquisition,image preprocessing,surface multiple feature extraction,and apple classification based on SVM multiple feature integration.There are 4 steps.In the image acquisition stage,each apple is photographed with a top view image and three different parts of the side view image as a system input.In the image preprocessing stage,the first use of homomorphic filtering to eliminate image uneven illumination,and adjust the gray-scale range;then it proposes an improved opening operation algorithm,to achieve accurate segmentation of foreground image part of apple.On the surface of multi feature extraction stage,the reference in the actual production of commonly used color,texture and roundness index,respectively,using the corresponding algorithm to extract the corresponding feature,including: color were analyzed in terms of degree calculation by H channel on the foreground segmentation,Gray-level co-occurrence matrix is applied to the feature of texture extraction analysis and the roundness of the apple is represented by measuring the error between the contour of the top view of apple image and a perfect circle.Finally,a high dimensional feature representation vector under combined multiple feature information is designed.In order to take into account the learning and generalization ability of classification models,a nonlinear classification network based on an improved mixed kernel function support vector machine is constructed to achieve automatic grading of three grades of apple quality.The proposed algorithm is implemented,and an apple grading software system is established and a large number of experiments have been carried out.The experimental results reveals that the proposed system is accurate and robust on apple classification task,it can significantly improve the efficient of apple classification task and reduce the cost of labor.