Traffic light full-automatic positioning and recognizing method based on self-regulated learning

Abstract

The invention discloses a traffic light full-automatic positioning and recognizing method based on self-regulated learning. The method includes the steps of firstly, positioning traffic lights according to an inputted video stream in the aspects of the spatial domain, the frequency domain and the time domain so as to obtain final positioning results, and determining the final positions and the final sizes of the traffic lights according to the positioning result of the frequency domain and the positioning result of the time domain; secondly, determining the initial recognition states of the traffic lights according to spatial domain detection results of the traffic lights and the final positioning results of the traffic lights; thirdly, renewing a Gaussian distribution model of pixel values which are generated when the traffic lights are turned on and turned off, and calibrating the initial recognition results based on the renewed model; fourthly, outputting positioning and recognizing results of the traffic lights. By means of the traffic light full-automatic positioning and recognizing method, restrictions such as color restrictions and shape restrictions of the traffic lights are prevented from being excessively depended on, the common difficult problems of positioning and recognizing of the traffic lights in a natural scene such as the color fuzzy problem, the color cast problem, the exposure problem, the automatic white balance problem, the similar interferent problem, the shielding problem, the electronic device aging problem, the long-scale contrast problem and the video jittering problem can be well solved, and the traffic lights are accurately positioned and recognized.

Description

technical field

[0001] The invention relates to a method for automatic positioning and recognition of traffic signal lights based on autonomous learning, and involves technologies such as pattern recognition, machine learning and image processing. Background technique

[0002] In the field of intelligent transportation, traffic lights play an important role in improving the efficiency and safety of intersections and the application of intelligent transportation systems. Video-based automatic detection and recognition of signal lights can be used in electronic police, traffic information collection, unmanned vehicles, etc., and has broad application prospects in the field of intelligent transportation.

[0003] Most of the current signal light detection and recognition algorithms need to pre-designate the signal light area, and then recognize it based on the empirical color or shape. The signal light recognition algorithm based on empirical color can achieve better results w...

Images