A System For Feature Selection And Recognition Of Apples Based On Machine Vision Technology

Manual grading method is the most primitive method of apple external quality grading.Not only great labour intensity but also the lack of objectivity and accuracy affect the efficiency and accuracy of the classification.Computer image processing and machine vision technique provide an automatic,nondestructive,high efficient and objective method for fruit external quality detection,which pushes forward the process of fruit grading research.The important features,like shape and color ,are important appearance quantities of apple, and they are also one of the important standards of grading. This grading system in this paper is studied through shape and color.Basic study done in this paper is based on the methods of automatic classification of apples, bringing forward a set of basic pretreatment, forming and selecting features as well as classification criteria. Most work has been focused on the feature extraction and selection. The selected feature parameters act on apples classification.According to the former experience of apples classification, combining with the former work in feature extraction and considering the transmission belt to place the status of Apple, quantificational described shape parameters, such as shape and size of the special status in apple image are extracted on the basis of apple segmentation. Following work has been done in color parameters extraction.Considering the complexity of different parameters, a step-by-step strategy is adopted in feature selection of apples to get optimalized feature parameters. Judging whether the apple's surface has stem or calyx or not, we determine their status in the transmission belt firstly.And then a feature selection method , named MPSO (Modified Particle Swarm Optimization), is proposed on the basis of Discrete Binary PSO. The original group velocity of a particle swarm was changed into two separate,different and parallel particle swarm velocity, which was effectively and quickly applied to the shape and color feature extraction of the optimum samples respectively. Then the least squares support vector machine classifier is used to verify the feasibility of this method.Multi-class classifier models of LSSVM(Least Squares Support Vector Machine) are constructed for apples recognition. With a small quantity of samples, our proposed methods achieve a decent performance of apples classification. This apple quality grading system, which selects DM643 as the main processor, is feasible for apple automatic grading in the detection results and has a wide range of applications.