Research On Classification Algorithm Of Recycled Plastic Bottle Based On Color Feature

As a big plastic consuming country,the consumption of plastic bottles is also large.If plastic bottles aren't recycled,treatment of burning and burial will cause environmental pollution.Meanwhile it's also a waste of resources.When recycled,single-color plastic bottles can reduce the limitations of recycled plastic applications.Moreover,the classification of recycled plastic bottles is mainly carried out manually in China at present,but the classification efficiency of this method is low and the working environment is poor.Therefore,it is of great practical significance to study the color classification of recycled plastic bottles.Color image processing and pattern recognition technology are applied to the automatic recognition of color classification of recycled plastic bottles in this paper.In order to improve the recognition rate of automatic classification and change the status quo of low classification efficiency.The images of recycled plastic bottle are smoothed and sharpened by color image processing technology in this paper.Then,according to the characteristics of color space,a watershed image segmentation algorithm based on morphological gradient reconstruction is proposed in HSI model.In this algorithm,the multi-scale structural elements of the two-structure element are applied to the operation of morphology opening and closing,and the improved morphological gradient operator is obtained.After that,the reconstruction operation and watershed segmentation are performed to obtain the area of interest.In order to reduce the interference of overlapping plastic bottles,this paper proposes a method of identifying the overlapping of recycled plastic bottles based on contour shape characteristics.The image after segmentation is binarised,and the contour tracking algorithm is used to obtain the outline point set.An effective descriptor is obtained by normalizing the outline point set in polar coordinates,and the shape matching is carried out through the Manhattan distance to achieve the purpose of overlapping identification.Then,the sample matrix is obtained from color histogram of hue and saturation two components of plastic bottle images.The paper analyzes the principle of three feature extraction algorithms,namely histogram statistical characterization,fast principal component analysis and kernel principal component analysis.They are used separately for effective feature extraction on sample matrices.At last,according to the influence of the parameters of support vector machine on the classification performance and the advantages of particle swarm algorithm,this paper uses the particle swarm algorithm to optimize the parameters.Then the extracted features are classified and recognized by the optimized support vector machine.Through the simulation experiments of LabVIEW and MATLAB,the feasibility and effectiveness of the algorithm are verified.In this paper,the proposed segmentation algorithm has a clear edge and a relatively complete target area.And the proposed overlapping recognition method can quickly and effectively achieve the purpose of identification.The eigenvalues obtained by the three color feature extraction algorithms are used for color classification recognition respectively.The results show that the recognition rate is higher when the eigenvalues extracted by the kernel principal component analysis method are used for classification.At the same time,compared with the traditional SVM classification recognition results,the optimized SVM algorithm has a higher classification recognition rate.