Research Of Key Technologies Of The Silkworm Cocoon Autofilter Counting System

With the development of computer vision technology, image processing technology is more and morewidely used in agriculture, industry, medicine engineering and other fields. Currently light for the textilesector, the counting method under the thread of each cocoon slot’s silkworm cocoon—manual counting cannot meet the productivity and quality of raw silk demand for the textile sector, this paper proposes a silkwormcocoon autofilter counting method based on threshold segmentation which bases on connected componentslabeling and another a silkworm cocoon autofilter counting method which based on K-means clusteringalgorithm. These counting methods have counting fast, counting accurate, counting stability and someotheradvantage, so they have good application prospects.This paper through introduced by the use of silkworm cocoon average area, maximum area as the prioriknowledge, using silkworm cocoon average area as the reference area, silkworm cocoon adhesion zone andseparate zone points of processing, proposes the threshold segmentation method which based on connectedcomponents labeling. Through using the method of the maximum distance find the cluster centers improvesthe K-means clustering algorithm, proposed adhesions silkworm cocoon segmentation and counting methodwhich based on improved K-means clustering algorithm. The key technologies of silkworm cocoon autofiltercounting system include silkworm cocoon image acquisition, image processing and autofilter counting threemodules. This paper first discusses the image preprocessing techniques of the silkworm cocoon autofiltercounting system, which include image file format, color space model, image color space model conversion,image enhancement, image denoising, etc, proposed a method which combines median filter method,mean-shift drifting method and Fourier transform, this method not only can effectively filter the image noise,but also can enhance the brightness of the image, enhance the image contrast of the silkworm cocoon targetand background, lay foundation for the subsequent image processing. Then the image segmentation andcounting methods of silkworm cocoon autofilter counting system have been researched, in this chapter wediscussed the theory and application of threshold segmentation algorithm, edge detection segmentationalgorithm, morphology corrosion and expansion segmentation method and K-means clustering segmentationalgorithm, meanwhile we compare and analysis these algorithms by doing experiments, in this paper, we usethe segmentation method which combines threshold segmentation algorithm, corrosion and expansion segmentation algorithm and K-means clustering segmentation algorithm, and it can receive a satifactorysegmentation result. Researches show that for the silkworm cocoon image segmentation, this method hasgood robustness. Finally, the paper designs a silkworm cocoon autofilter counting system based on PC andbased on Android platform to verify the key technologies of this article, and verify the counting accuracy,robustness and counting efficiency of the system by a human-computer interation interface.The results shows that the proposed silkworm cocoon counting method which combines thresholdsegmentation algorithm, corrosion and expansion segmentation algorithm with K-means clustering algorithmhas good robustness, mainly can meet the needs of factory assembly automatic counting, the averageaccuracy rate of automatic counting is more than96%. And this method can overcome background reflective,silkworm cocoon shadows and other phenomena of silkworm cocoon image, solve the problems of thetraditional manual counting slowly, low efficiency and low accuracy, achieve high efficiency and goodrobustness of silkworm cocoon counting.