Research on Recognition Method of Automobile Collision Warning Icons Based on Machine Vision
Aiming at the problem of time-consuming data processing of automobile collision warning test,machine vision is innovatively applied to the data processing of automobile collision warning system test.The collision warning system test is divided into a testing phase and an offline data processing phase.A method based on the local minimum of color histogram similarity is constructed in the testing phase to extract key frames.During the test,the inter-frame color histogram similarity sequence is firstly acquired in real time,then the similarity sequence is smoothed using the Hanning window,and finally the key frames are acquired based on the smoothed inter-frame similarity local minima.In the data processing stage,an image containing the warning icon is manually selected to obtain the position of the icon in the image.Feature extraction and matching are performed on the local region according to scale-invariant feature transformation feature extraction and fastest neighborhood matching.The test results prove that the method can realize fast and effective recognition of automobile collision warning icons and output key frames containing time information.The method can improve the efficiency of collision warning system test data processing and enhance the degree of automation.
万方数据
维普
