采用模糊支持向量机算法的前车识别系统

Vehicle ahead recognition system based on fuzzy support vector machine algorithm

针对辅助驾驶汽车在行驶过程中对前方车辆识别的实时性差、效率低等问题,提出基于方向梯度直方图(histogram of oriented gradient,HOG)特征提取与模糊支持向量机结合的一种前车识别系统。对汽车行驶过程中的图像灰度化、二值化等预处理后,进行HOG特征提取。对模糊支持向量机算法进行研究,通过增加模糊度变量的优化来选择最优分类决策面,使其对每个训练的正、负样本集的区域特征进行分类后识别。实验结果显示:模糊支持向量机算法不仅能够降低训练时的噪声,与支持向量机相比增强了支持向量,而且提高了训练时间与准确率;在能见度低的情况下识别效果较好,满足前车实时识别。

Aiming at solving the problems of poor real-time and low efficiency towards a vehicle ahead recognition in the process of assisted driving,a vehicle ahead recognition system is proposed based on the combination of directional gradient histogram(HOG)feature extraction and fuzzy support vector machine.Firstly,after the preprocessing of image graying and Binarization in the process of vehicle driving,the directional gradient histogram(HOG)feature is extracted.Then,the fuzzy support vector machine algorithm is studied,and the optimal classification decision surface is selected by increasing the optimization of fuzzy variables so that it can classify and recognize the regional features of each positive and negative sample set.The experimental results show that the fuzzy support vector machine algorithm can not only reduce the noise during training,but also enhance the role of support vector compared with support vector machine,and improve the training time and accuracy.In the case of low visibility,the recognition effect is better,which can meet the real-time recognition of the vehicle ahead.