摘要：The science of image processing and computer vision holds that edge detection is crucial. Discontinuities in a digital image may be categorized into three varieties: point, line, and edge. Spatial masks are the most popular method of finding discontinuities. The more isolated points and lines that can recognize edges, the better picture segmentation will perform. Edge detection is a way of finding and accentuating edges in a picture, so as to increase the apparent picture quality in some circumstances. Edge detection is a key method used in many image processing applications, including feature extraction and object segmentation, to get information from the frames as a preprocessing step. An edge-detection operator calculates the degree of variation between individual pixels by performing a matrix-area gradient operation. A matrix that centers on a pixel chosen as the center of the matrix region is used to calculate the edge-detection operator. The center pixel is considered as an edge if the value of this matrix region is over a certain threshold. Gradient-based edge detectors such as the Roberts, Prewitt, and Sobel operators are some of the most frequently-used edge detectors. Differential operators help image gradients provide edge detection information. Actually, to simulate either the first derivative or the second derivative of a picture, traditional edge detection algorithms such as Roberts, Sobel, Prewitt, and Laplace filters use small convolution masks that mimic either the first derivative or the second derivative of the picture; for example, they emphasize the improvement of edges by using little smoothing or no smoothing at all. To do this, the output of these filters is passed through a threshold. Later edge detectors allow better control over smoothing and edge localization, although these filters have a little speed advantage. These filters are therefore particularly sensitive to noise. A method is described that details the building of a fuzzy inference system that determines the gray-scale values of pixels which are neighbors, discovering differences in the light intensity of related pixels. The technique described in this work scans an input image using a 2×2 mask, and then uses fuzzy logic to locate vehicle edges in the picture. Fuzzy logic was created using the recently developed set theory, a form of mathematical logic that includes approximate results. Fuzzy set theory is famous for its aptitude in characterizing all types of uncertainty and ambiguity. A technique for applying trapezoidal and triangle membership functions to four inputs and one output of mamdani type FIS was presented, using these parameters: two fuzzy sets and one fuzzy set. A Gaussian filter is used to achieve better outcomes. The method's experimental success suggests it can identify thin edges in vehicle images.
