MSER (Maximally Stable Extremal Region) and genetic optimization SVM (Support Vector Machine)-based TSR (Traffic Sign Recognition) method

Abstract

The invention provides an MSER (Maximally Stable Extremal Region) and genetic optimization SVM (Support Vector Machine)-based TSR (Traffic Sign Recognition) method, and belongs to the technical fieldof image processing. By adopting a method for carrying out edge detection and image segmentation on a to-be-recognized region through a feature vector of a blocking HOG (Histogram of Oriented Gradients), influence brought by translation and rotation can be inhibited to a certain degree, and the interference to an image due to variation of illumination intensity can be reduced; meanwhile, comparedwith a traditional HOG, the blocking HOG has the advantages that the dimensionality is greatly reduced, and the computing efficiency is increased; during a classification and recognition phase, an optimal SVM classifier parameter is obtained by computation by applying an adaptive crossover and mutation-based improved genetic optimization optimal parameter searching algorithm, fallibility of manualmarking and large-amount time consuming of machine training can be avoided, the requirements on accuracy and instantaneity are well balanced by combining the advantages of all methods, and automaticdetection and recognition of traffic signs can be realized; according to the MSER and genetic optimization SVM-based TSR method provided by the invention, testing images in a Germany traffic sign detection standard database are recognized, and a better effect is obtained.

Description

technical field [0001] The invention belongs to the technical field of image processing and is applied in intelligent traffic scenes. Using the traffic sign extraction algorithm based on the MSER maximum stable extremum region to segment and extract RGB images containing traffic signs from natural scenes. Afterwards, the features of traffic signs are extracted by using block HOG gradient direction histogram. Using the improved genetic optimization optimal parameter search algorithm based on adaptive cross-variation, find the optimal penalty parameter g and kernel function parameter c in the SVM classifier to construct the optimal SVM classifier for this problem to identify images traffic signs in. Background technique [0002] Image processing technology is an important part of the field of intelligent transportation. Efficient and accurate automatic traffic sign recognition can guide traffic participants to participate in traffic behavior in a standardized manner, reduce ...

AI Technical Summary

Problems solved by technology

Therefore, in view of the disadvantages of color recognition in complex environments, the error-proneness of manual marking and the time-consuming machine training, and the real-time requirements of the system, there is no algorithm with more balanced performance and more stable effects to achieve Intelligent Recognition of Traffic Signs

Images