相似车型分类识别算法研究与仿真

Classification and Recognition Algorithm Research and Simulation for Similar Car Models

研究车辆车型种类的分类识别问题,提高车型分类的准确率。针对汽车车体型号分类多是根据提取车辆的高度、宽度等几何特征作为对车体型号判定的依据,当不同型号的车辆高度、宽度等特征相差不大的时候,单纯依靠车辆几何特征进行分类的方法容易出现特征混淆,造成分类准确性不高。为了解决上述问题,提出了一种神经网络算法的车型识别算法。通过提取车型特征作为训练样本,经过BP神经网络训练,多层细分,得到车体型号准确特征,避免了仅对体积特征提取的弊端。实验证明,改进识别算法实现简单,识别准确率高,取得了满意的效果。

Research the classification of cars to improve the accuracy of classification and recognition. This paper proposed an car models identification algorithm based on neural network. Through feature extraction of car models as the training samples, and after training the BP neural network and, multilayer subdivision, accurate car body charac- teristics were obtained, to avoid only obtaining volume feature. Experiments show that the recognition algorithm is simple, of high identification accuracy and good results.