A fast traffic sign detection method based on kernel extreme learning machine

Abstract

The invention discloses a fast traffic sign detection method based on a nuclear ultra-limit learning machine, which belongs to the field of image signal processing and pattern recognition. First, read the original sample image, use the objectness method based on BING to generate the area that may contain traffic signs, extract the HOG features of the candidate area, and send it to the kernel extreme learning machine classifier to obtain the final detection result; The invention abandons the traditional sliding window scanning method and uses the BING algorithm to reduce the search space and improve the detection speed. The traditional ELM algorithm has a single hidden layer structure, which has great limitations in analyzing complex signals. The present invention adopts KELM for classification and detection, and the kernel extreme learning machine can make the learning model more stable, and the generalization performance is stronger, improving the Detection performance, and maintain the advantage of low time consumption of ELM.

Description

technical field [0001] The invention belongs to the field of image signal processing and pattern recognition, and relates to a method for fast traffic sign detection by using a binarization normative gradient (BING) and a nuclear ultra-limit learning machine. Background technique [0002] In today's society, with the development of the economy, people's living standards are improving day by day, and the number of private cars is increasing day by day. At the same time, the problem of urban road traffic safety is becoming more and more serious, so the intelligent transportation system emerges as the times require and develops rapidly. Intelligent transportation systems include car navigation systems, collision warning systems, traffic sign recognition systems (Traffic SignRecognition System, TSR) and other intelligent systems. The first two systems have been widely used, but TSR has not yet reached the level of practical application. Regardless of actual driving or unmanned...

AI Technical Summary

Problems solved by technology

However, the training cost of such algorithms based on machine learning and classifiers is very high, and it is difficult to meet the requirements of real-time traffic sign detection under natural conditions.

Moreover, this kind of method adopts the traditional exhaustive sliding window scanning method, which has a larger search space and a larger amount of computer, which greatly slows down the detection speed.

Images