Method for analyzing region of interest of traffic scene

Abstract

The invention discloses a method for analyzing a region of interest of a traffic scene. The method includes the steps of firstly performing feature extraction on an input color traffic image frame sequence, extracting relevant features including color, statistics, texture, structure and the like, and stacking all the extracted features, and finally obtaining a three-dimensional data cube, which isa feature image; vectorizing each dimension of the feature image to obtain a two-dimensional matrix V, subjecting the matrix to dimension reduction processing, and decomposing the matrix into the form of a product of two matrixes W and H by using a vertex component analysis method; for the matrix H, two-dimensionalizing each column to form an image with the same size as the input traffic image, aregion of interest being in each image. Therefore, by the way of feature extraction and then decomposition, the method obtains the region of interest of the traffic scene and solves the problem of unsupervised extraction of the region of interest of the traffic scene image.

Description

technical field [0001] The invention relates to an analysis method for exploring regions of interest in traffic scene images in the form of feature image decomposition, and belongs to the technical field of artificial intelligence. Background technique [0002] Common intelligent traffic image processing systems include four parts: segmentation, detection, recognition and tracking. For segmentation tasks, researchers often use thresholding methods based on visual experience or related priors. For detection and recognition tasks, researchers design different descriptors, such as edges, gradient orientation histograms, local binary patterns, and extended forms, to deal with them. For example, Ruta adopted the Canny edge detection operator and reported a 100% detection rate, but only for one traffic sign. Haar-like features are also widely used and have achieved certain results. Other features have also caught the researchers' attention. HOG is a typical feature. It forms f...

AI Technical Summary

Problems solved by technology

However, these features are designed according to the target, and more space and texture information are used, and there is less research on the more general and fundamental question of "what is the region of interest?"

Images