基于谱回归特征降维与后向传播神经网络的识别方法研究

采用后向传播(BP)神经网络对空间目标进行识别时，高维的输入特征导致网络结构复杂，识别性能降低。针对上述难点，该文提出一种基于谱回归(SR)特征降维与BP神经网络的识别方法。该方法首先对空间目标进行HOG特征提取，然后将提取的高维HOG特征进行SR降维，最后把降维后的数据通过BP分类器进行训练识别。实验结果表明:该方法的降维和识别特性优于传统降维方法PCA, KPAC, LPP, KLPP等，能够兼顾实时性和准确性，提高了识别性能。

Research on Recognition Method Based on Spectral Regression and Back Propagation Neural Network

When using Back Propagation (BP) neural network to recognize the spatial target, the high dimensional input features induce the complexity of the network structure and the poor performance of the recognition. In this paper, a new recognition method based on Spectral Regression (SR) feature dimension reduction and BP neural network is proposed for the above difficulties. Firstly, the HOG features are extracted from the spatial object, and then the feature dimensions are reduced by SR. Finally, the BP classifier is used to train the data. Experimental results show that the proposed method is better than the traditional dimension reduction methods such as PCA, KPCA, LPP, KLPP in dimension reduction and recognition, which can juggle real-time and accuracy, thus improving the recognition performance.