| After color screening,waste plastic bottles can be made into colored plastic particles by crushing,high-temperature melting and cooling pelletizing.These plastic particles can be recycled as raw materials according to their different colors,which has high economic benefits.The plastic bottle recycling mainly rely on artificial recycling in China,this way of recycling efficiency is low,harm to human body health,based on machine vision for this industry are actively developing automatic sorting equipment of plastic,but the existing equipment of waste plastic bottle color classification accuracy is low,cannot meet the old plastic recycling reproduction industrial equipment for the pursuit of efficiency and accuracy.The research shows that after industrial compression treatment,waste plastic bottles are prone to produce sticky overlapping on the recycling and sorting conveyor belt,which is the main reason for the low accuracy of color classification of waste plastic bottles.In view of this,this article in order to improve the color classification accuracy and efficiency for the purpose of waste plastic,in view of the image obtained by industrial camera,in image denoising,shape,color matching and classification three aspects study,put forward a set of waste plastic bottles based on machine vision recognition classification of visual design,solve the sticky plastic bottle overlapped recognition problem,plastic color classification method was optimized and improve the detection efficiency.The main research contents of this thesis are as follows:1.This thesis proposes a image processing based on double threshold segmentation and morphology interested region extraction method,select the background image after image enhancement,color for the background image modeling,determine the high and low threshold value of image segmentation,threshold segmentation on binary image processing,morphological filtering to remove noise existing in the threshold segmentation,get interested in target image area.2.In order to solve the interference caused by the sticky and overlapping phenomenon on the color classification of waste plastic bottles,a shape descriptor based on the spatial structure characteristics of the set of pixels in the polar coordinate system was proposed in this thesis.The contour tracking algorithm was used to extract the edge point set of the plastic bottle,and the extracted contour edge point set was transformed into the polar coordinate system.The Angle and scale were normalized.The coordinate system was divided into multiple regions,and the proportion of pixel points in each region was used as the shape descriptor.Finally,this kind of shape feature is trained and matched with support vector data description method,so as to complete the recognition of plastic bottle body sticking and overlapping based on shape feature.3.In order to reduce the influence of surface reflection of compressed plastic bottles on color feature extraction,this thesis proposes a color classification method based on improved k-means color clustering.Under the Lab color space by color clustering method to filter the image after color quantization,establish classification standard color palette,through calculating the bottle each pixel in the image with the standard palette color difference between pixels and the standard color mapping,the final statistical bottle frequency as the color features of each pixel color for the color of the plastic bottles classification. |
