Image processing method based on belief propagation

Abstract

The invention provides an image processing method based on belief propagation. The method comprises the following steps of: (1) acquiring a group of images via a camera; (2) extracting the features of the group of images through a feature extraction algorithm; (3) executing preliminary estimation on the belief of the features; (4) according to the characteristics of the problem to be solved, establishing a constraint relation of the features for solving the problem; (5) taking the constraint relation as a condition, calculating the belief of the features in the step (4) through a belief propagation algorithm; and (6) updating the features in the step (4) with the belief calculated in the step (5), and ending the process. The image processing method based on belief propagation is used in an automobile safety electronic system, has high accuracy rate in feature extraction, and can improve the identification accuracy rate of the automobile safety electronic system.

Description

Technical field: [0001] The invention relates to the technical field of image processing, in particular to an image processing method based on belief propagation used in automotive safety electronic systems. Background technique: [0002] In automotive safety electronic systems, one or more cameras are generally used to collect video signals of roads and vehicles in front or around, and then feature extraction algorithms are used to extract features (such as lane lines, contours, edges, etc.), and then judged based on the features The driving environment of the car. [0003] The current image processing method is generally to do technical processing on a single image, and the video recorded by the car while driving is affected by the environment (such as light, shadows of surrounding objects, weather, etc.), the picture is polluted by noise, The extracted features may contain false and wrong information, such as edges that do not correspond to any objects, etc. Even useful ...

AI Technical Summary

Problems solved by technology

[0003] The current image processing method is generally to do technical processing on a single image, and the video recorded by the car while driving is affected by the environment (such as light, shadows of surrounding objects, weather, etc.), the picture is polluted by noise, The extracted features may contain false and wrong information, such as edges that do not correspond to any objects, etc. Even useful features, such as the edges of lane lines, may be incomplete and blurred due to factors such as lighting, so , only relying on the information of a single image, the accuracy of feature extraction is low, which seriously affects the subsequent processing in the automotive safety electronic system, affects the recognition accuracy of the automotive safety electronic system and the accuracy of judging the driving environment of the car

Images