Research On The Edge Detection Of Gear Image And Bearing Image Based On Image Enhancement

With the improvement of China’s economic strength and science and technology,image processing technology has gradually entered a stage of rapid development.Common image segmentation,image denoising and image enhancement technologies include image segmentation,image processing and so on.Among them,image edge detection is the most advantageous feature information to describe the underlying features of the image.At the same time,as a key step of industrial vision measurement,it has an important impact on the application of image deep-seated information analysis and subsequent processing technology.Image enhancement is widely used in low illumination images,infrared images,remote sensing images and fog images.It can enhance the detail information that people need in the image,so as to improve the effect of image processing.Therefore,using image edge detection and image enhancement technology to study the related problems in industrial images has practical significance.There are many kinds of industrial images today.This paper mainly realizes the gear image enhancement and edge detection processing based on the relevant theories such as nonlinear filter,distance formula,power change,Retinex algorithm,wavelet modulus maximum and Canny edge detection.The contents are as follows:（1）Preprocessing the industrial image can lay the foundation for the subsequent deep detection.Therefore,this paper will first introduce the common image denoising algorithms and image color model,focusing on the denoising in the preprocessing stage of industrial gear image and bearing,and propose an adaptive median filter combined with Markov distance to denoise it,It lays a foundation for the subsequent research on detail enhancement and edge detection.（2）In order to highlight the details of gear image and bearing image,an improved Retinex image enhancement algorithm is proposed in this paper.Firstly,the image enhancement evaluation index information entropy is used to improve the power transform,so that it can adaptively select the power index according to the image features,and then it is applied to the enhancement algorithm based on Retinex theory,which effectively improves the overall contrast of the image.Secondly,in order to make the local detail information of the bearing image more prominent,the model is constructed according to the local mean and local variance of the image to get the final image.The experimental results show that the improved Retinex algorithm has a certain improvement in human vision than histogram equalization and adaptive histogram equalization,and the details are more prominent.（3）For the edge detection of gear image and bearing image after image enhancement,this paper uses the proposed improved adaptive median filter for image preprocessing,uses the improved Retinex algorithm to enhance image details,and finally improves the threshold selection method in the process of wavelet modulus maximum edge detection to extract the sub-pixel edge of noisy gear image.The experiment shows that the edge continuity of gear image is improved to a certain extent,and the influence of noise on the edge is reduced.