Edge Detection Of The Slope Vegetation Root Image Based On Wavelet Transform

In recent years, the method of vegetation in the slope is widely used, and this method has played dual role in greening and reinforcement for the slope. This thesis is originated from the national natural science funds project that research on the mechanism of vegetation in the slope is based on the analysis of the dynamic model of root microstructure. This project makes use of array distribution endoscope to obtain image information of vegetation in the slope, such as root form, structure of soil, particle composition and fracture of soil. Then we need to analyze and process the image information to construct a dynamic model and analyze it to research the mechanism so as to realize the quantitative analysis of the function.This paper uses image processing technology to detect and process the parameter closely related with the root form and do a great help for constructing the model, the implementation and completion of the project funded.Edge one important characteristics of image, is discontinuous performance of the image, namely a set of the singularity. In the image processing, edge detection plays a very important role because the edge of usually contains lots of valuable information which constitutes the outline features of the image. The outline of root image just is the edge that needs extracting during the project implementation process.Wavelet transform a branch of mathematics developing based on Fourier transform has good localization characteristics between time domain and frequency domain. It can conduct multi-scale analysis to a signal through scalability and translation operations and especially has great advantages in signal singularity detection. Therefore this paper adopts wavelet transform to detect the edges of the collected vegetation image in the slope.This paper firstly adopts the traditional differential edge detection operators to detect the edge of the slope vegetation root image. Test results show that the operator is more sensitive to noise and can not detect the detail information well in a single test scale. Then based on the multi-scale characteristics, we choose the partial derivative of gauss function as a wavelet function, and use the method of modulus maxima to detect the edge under the multi-scale images. Test results show that the multi-scale edge detection can better restrain the noise in image edge, and have better detecting effect on the edge details and the edge integrity. Finally, the article selects the cubic B-spline wavelet function to conduct multi-scale edge detection. According to the Canny criteria, cubic B-spline wavelet is proved to be approximate optimal, then getting high-pass and lower-pass filter coefficient by calculating. Results show that the B-spline wavelet is also good at restraining noise and has incomparable advantage in the details detection. Then we give the lifting algorithm of the B-spline wavelet transform and the lifting wavelet transform can achieve the same detecting effect, but has simplified the detection algorithm and improved the calculation speed so as to make it superior in slope vegetation root image edge detection.