Vehicle-borne multi-obstacle classification device and method based on Bayes classifier

Abstract

The present invention relates to a vehicle-borne multi-obstacle classification device and method based on a Bayes classifier. The classification device comprises a camera and a PC which is connected to the camera. The PC comprises a Kalman filter module for carrying out Kalman filtering on the vehicle front video image collected by a camera and detecting an obstacle target, a characteristic extraction module which is used for carrying out characteristic extraction on the detected obstacle target, and a Bayes classification module which is used for using a Naive Bayes classifier to obtain the classification of the obstacle target according to the characteristics of the obstacle target, wherein the characteristics comprise a symmetry characteristic, a horizontal edge straightness characteristic and a length and width ratio characteristic, and the classification comprises a cyclist / motorcycle rider, a vehicle side face, a vehicle front side and pedestrians. Compared with the prior art, the device and the method have the advantages of high recognition accuracy, strong anti-interference ability, high efficiency, and good real-time performance.

Description

technical field [0001] The invention relates to the technical field of vehicle-mounted multi-obstacle classification, in particular to a vehicle-mounted multi-obstacle classification device and method based on a Bayesian classifier. Background technique [0002] Vehicle multi-obstacle classification is an important research topic of intelligent transportation and an important part of the field of intelligent vehicle surrounding environment perception technology. In recent years, in the field of intelligent transportation research at home and abroad, many methods have been proposed for the identification of vehicle-mounted obstacles, mainly including the following methods: [0003] Meng Huadong et al. proposed a vehicle type detection method based on multi-sensor information fusion combined with Bayesian networks. The height profile and plane profile of vehicles were obtained through microwave and video sensors, and Bayesian networks were used to classify 7 types of vehicles....

AI Technical Summary

Problems solved by technology

[0004] The main disadvantages of the existing technical solutions are: the current Bayesian method has the main problem that it is difficult to select training samples for inter-class differences, and in the Bayesian classification stage, the time complexity depends on the degree of dependence between the eigenvalues. The accuracy rate is low, and the real-time performance of the system is not strong

The main disadvantages of this method are the accumulation of errors and the problem of non-separable regions

The BP neural network classifier has the characteristics of parallelism and robustness, which can greatly reduce the calculation time of the classifier, but the error rate of the classifier is high

Images