基于超声波传感器阵列的车辆周围目标物识别
Target recognition around a vehicle based on an ultrasonic sensor array

为了对车辆环境信息进行探测与识别，提高道路交通安全，该文研究了利用超声波传感器阵列识别车辆环境目标类型的方法：以分类器为核心，搭载目标物区分算法，实现对目标物的实时探测与区分。首先根据车辆周围环境信息提取了具有一定形状特征的典型目标物，分别为平板、圆柱和角形3种类型，以支持向量机（support vector machine，SVM）为分类器建立分类模型，根据目标物形状不同构造与优选分类特征指标集；依托分类结果建立目标物的类型区分算法，得到每一时刻各类目标物的概率值。仿真结果表明：分类模型的分类准确率达到91.5%；同时该方法对各种类型的目标物进行区分时，均达到较好的识别效果。在实际道路试验中，采集了汽车、行人和自行车3类目标物，采用该方法达到了良好的识别效果，具有可行性。
Abstract：Road traffic safety can be improved by detection and recognition of vehicle environment information. This paper describes a method for recognizing the environment target type by an ultrasonic sensor array. A high performance classifier is used in an object type discrimination algorithm for real-time target detection and identification. Three general shapes are identified as planes, cylinders and triangles by a support vector machine (SVM) classification model. The target shape is then extended to more specific features. The classification algorithm also gives the probability of each object at that moment. Simulations show that the classification models are 91.5% accurate which demonstrates that the object type discrimination algorithm can recognize environmental objects near the vehicle. In road test, the experimental platform collected data on road objects, such as vehicles, pedestrians and cyclists with good recognition accuracy.