The Study Of Shiitake Classification Tecnology Based On Machine Vision

With the development of science and technology, image technology which is a nondestructive testing technique has drawn increasing attention. Shiitake were taken as research objects, this paper systematicallly studied the principle and method of classification technology of shiitake based on machine vision, constructed a machine vision system which were applied to shiitake sorting. This paper focuses on the image processing system. The image processing system was divided into several modules, which was the computing of the size of shiitake cap, the recongnition of shiitake shape, the analysis of shiitake cap texture, the recongnition of shiitake broken.The main result of research are as follows.(1)the reseach of recongnition of the front and back of shiitakes. Shiitake cap is divided into two parts by the methods which are ostu threshold and external polygon.10texture parameters are extracted from the two parts by probability density function. Then,minimum distance classifier are constructed. Experiments show that the accuracy of final the model reach up to98%.(2) the reseach of recognition of shiitake shape and size. Linear interpolation of edge curve of the shiitake mushroom handle removed region is performed under polar coordinate, and the shape of pileus is reconstructed, then nine shape parameters and one parameter of size are distracted based on the curve reconstructed. The shape parameters are reduced to3with principal component analysis. The minimum distance classifier can be constructed as the sorting model with the3main shape parameters as input. Considering with the identification of stem of shiitake mushroom, the final identification of the shiitake was determined on both of the shape and the size classification. Experiments shows that the accuracy of shape sorting can be greatly improved by identification of stem of shiitake mushroom, and the precision ratio of shiitake mushroom shape grading was95.6%.(3) The research of shiitake handle recogintion. With shiitake mushroom as a case study, the boundary is tracked and the curvature is calculated by tracking the boundary and then the concave-convex quality of boundary can be distinguished with the curvature, then the possible shiitake mushroom region is found and the actual position of the shiitake mushroom can be found with the class radius of curve.(4)The research of the automatic grading method of shiitake based on texture analysis.To achieve the design of automatic shiitake grading system, the images of four varieties such as Tian pai-hua Shiitake, Pai-hua Shiitake, Tsa-hua Shiitake and Smooth Cap Shiitake were taken as research objects. Shiitake texture was a vital indicator of shiitake quality.The more white texture shiitake pileus, the higher its price. Shiitake grading was mainly processed by manual operation for a long time.The grading operation was heavy workload, inefficient and not conducive to automatic production. So shiitake market was eager for shiitake grading equipments.This study selected three models to describe pileus texture.The first texuture model was derived from gray histogram and grey level co-occurrence matrix. The second model was called gauss makov random field.The third model was defined by fractal dimention. Shiitake grading process was described as follows. Firstly, the texture analysis region was intercepted from shiitake pileus by appropriate rectangle. Five texture feature parameters were extracted from the texture analysis region according to the gray histogram; another five texture feature parameters were extracted according to grey level co-occurrence matrix; twelve texture feature parameters were extracted according to gauss makov random field; the fractal dimension extracted from fractal model was the last texture feature parameters. Three texture models could describe texture information from different perspective. Each texture feature expressed specific physical meanings. But it was relevant among texture features in most cases. This study chose sequential feature selection algorithm to eliminate the defect. Sequential features selection algorithm could remove the correlation among features and six effective feaures were selected after the correlation-removal operation. Finally, the K-nearest neighbors classifier was constructed as the shiitake species classifier, than the test shiitake samples could be classified with the six effective features mentioned above by the K-nearest neighbors classifier. Experimental results showed that the final accuracy reached to91%, which could meet the requirement of production.(5) The research of the Application of Machine Vision in the detection of broken shiitake. In order to detect the broken shiitakes, a automatic detection system of shiitake with the practical corresponding algorithm was developed based on machine vision. Identification algorithms based on analysis of curve and shiitake edge grayscale are presented in this paper. First, remove the background of shiitake images, track the edge of shiitake and then obtain the coordinates of shiitake boundary. A closed curve is composed of these coordinates. Two initial curves can be generated from the closed curve of internal and external respectively.Two final curves are born evolved on these two intial curves, which meet the condition of specific termination criterion. Two parameters(Nin, Nout) are extracted from the difference of final curves and initial curve. These parameters can determine condition of shiitake; shiitake edge region are sampled with the method of morphology.Then4parameters can be extracted from the sequence of the gray scale of sampled regions.These parameters are mean(μ), variance(p), average width of peaks ((?))and maximum width of peaks(Lmax) respectively. Processing the4parameters with the method of pattern recognition and obtaining the conditon of shiitake from the processing results. The final discrimination result is given with mixed results of both curve anylysis and grayscale anylysis. Experiments show that the accuracy of final shiitake discrimnal model reach up to88.3%, Which provide a technical foundation for shiitake automatic sorting...