汽车乘员类型视觉检测算法

首先根据不同乘员类型体型轮廓的差异性,建立了基于低成本摄像头的乘员类型视觉检测系统;根据乘员图像样本数据信息量大而复杂的特点,利用Legendre矩阵对原始图像中乘员体型轮廓特征进行了描述;然后针对实验样本类型多、模式类型少的特点,引入智能模式识别中支持向量机的方法。通过样本数据的训练和参数的调整,建立了乘员类型识别器。最后进行了样本数据的测试。结果表明,用此种方法建立的识别算法能够以较高的准确率对乘员类型进行分类,为智能安全气囊系统提供准确的乘员类型信息。

Vision detection algorithm for vehicle occupant classification

A vision detection system based on the low-cost CMOS camera is established according to differences of outlines of occupants.Principal character of occupant outline in the original image is described using Legendre matrix for the original image contains much more information than the detection algorithm actually needs.Support vector machine is introduced as the classifier after being trained and tuned using sample data considering the characteristics of the original data.The testing results indicate that the vision detection algorithm in this paper can effectively classify different kinds of occupants,providing for the intelligent restraint system with robust and accurate occupant classification information.