为解决传统基于 Haar 特征和自组织映射概率神经网络(SOMPNN)的车辆检测算法中存在当 Haar 特征向量维数过大时决策时间缓慢和因平滑因子σ单一易导致分类错误的2个不足,提出了一种用低维的 Haar-NMF 特征代替 Haar 特征和平滑因子自适应修正的改进 SOMPNN (ISOMPNN)车辆检测算法。首先用非负矩阵分解对 Haar 特征进行降维,生成低维 Haar-NMF 特征;其次,以 SOM 输出层神经元的原型向量数作为修正因子,构建了指数函数形式的平滑因子修正函数,并以修正后的平滑因子训练 SOMPNN 分类器。实验结果表明,与传统的 Haar +SOMPNN 算法相比,采用 Haar-NMF 和 ISOMPNN 构建的车辆检测分类器在检测率、误检率和检测时间等性能指标上都获得明显提升。%The traditional vehicle detection algorithm based on Haar features and self-organized map-ping probability neural networks (SOMPNN)has two shortages:High-dimensional Haar features u-sually cause long decision time;the constant smooth factor σof SOMPNN often causes false classifi-cation.To solve these problems,low-dimensional Haar-NMF(non-negative matrix factorization) features instead of Haar features and an improved SOMPNN(ISOMPNN)with adaptive smooth fac-tor correction are adopted to build the vehicle detector.First,NMF is used to generate low-dimen-sional Haar-NMF features.Then,the neuron number of the output layer of SOM is set as a correc-tion factor to build the smoothing factor correction function in the form of the exponential function. The SOMPNN classifier is trained with the corrected smoothing factor.Experimental results demon-strate that the performance of the Haar-NMF +ISOMPNN-based vehicle detection classifier is im-proved in the detection rate,false detection rate and detection time compared with the traditional Haar +SOMPNN-based algorithm.
展开▼