Research And Application Of Dairy Cow Image Edge Extraction Based On CNN

Dairy cow's digital image edge extraction is a prerequisite of a dairy cow type linear appraisal, which is one of the most pressing problems. One of the most basic features of the image is the edge. It is not only the basis of extracted characteristic parameters and object recognition , but also the most significant changes in the the local image brightness changes. Edge extensively exists in the targets to targets, targets to background, region to region and color to colors. The edge is the junction of more than two different regions. Due to widely spreading uncertainty and a large number of mutation information, it becomes an area containing the richest information. First of all, three CCD cameras are used to gather dairy cows'ahead, side and rear image as dairy cows edge extraction collection targets. And digital input computer through the image acquisition card; Secondly, preprocessing the collected the dairy cow's digital image (dairy cows'side ) and then filtering and contrast enhancing the digital image (linear gray level transformation, histogram equalization).In order to extracting accurate dairy cow's image edge and improving the level of dairy cow type linear appraisal, in this paper cellular neural networks (Cellular Neural Networks, referred to as CNN) were applied to the dairy cow's digital image edge extraction. Basing on studying Classic Algorithm such as Robert operator, Sobel operator , Prewitt operator, Log operator and Canny operator, edge detection operator with more dairy cows were digital images of the simulation. It is best summed up when the edge detection threshold of each operator and the standard deviation; Then, the cellular neural network is applied to the dairy cow's digital image processing. According to the structure characteristics of cellular neural networks, adjust the Grayscale pixel values to make it meet the input range. According to the practical problems of digital image, determine the feedback template A, control template B and the threshold value of Z of CNN algorithm. For the time convolution of CNN algorithm for the boundary cells A and B, some elements of the template will not find the boundary cells corresponding to cells, so this paper presents A "increase law" and the "narrow method" to solve the problem. CNN algorithm Again Finally, respectively based on CNN's dairy cows binary and gray image edge extraction algorithm, MATLAB language programming is realized to extract binary image edge and gray image edge.Comparing the dairy cow's image edge extracted by the classical algorithm , such as Robert operator, Sobel operator, Prewitt operator, Log operator and Canny operator, to the edge extracted by CNN algorithm, it is concluded that cellular neural network digital image edge extraction dairy cows continuity is better, more clear and higher precision positioning. It also lays a good, solid foundation for the automation dairy cows linear assess work basing on the machine vision.