基于多特征融合和BP-AdaBoost算法的列车关键零件故障自动识别

针对列车集尘器和安全链锁紧螺栓的故障检测，提出了一种基于多特征融合和BP-AdaBoost的故障自动识别算法。首先提取故障区域与非故障区域的局部二进制模式（LBP）、方向梯度直方图（HOG）和Haar-like三类特征；其次利用主成分分析（PCA）定义不同特征对故障识别准确率的贡献值，并据此对这三种特征进行降维和融合；再次利用融合特征来训练BP-AdaBoost分类器；最后用训练好的分类器结合不同的识别算法，对集尘器和安全链锁紧螺栓的故障进行定位和识别。实验结果表明，该算法能较好地识别两种不同故障，故障识别率高，误检率和漏检率低，对于光照不均和遮挡情况有一定的鲁棒性。

Automatic Fault Recognition for Key Parts of Train Based on Multi-feature Fusion and BP-AdaBoost Algorithm

An automatic fault recognition method was proposed for the fault detection of the fastening bolts and dust collectors based on multi-feature fusion and BP-AdaBoost algorithm. Firstly, the local binary pattern (LBP), histogram of oriented gradient (HOG) and Haar-like features of the faulty and non-faulty areas were extracted. Then, the principal component analysis (PCA) was used to define the contribution of different features to the fault recognition accuracy, the three features metioned above were fused, and the dimensionality reduction was conducted to the fusion feature. Then the BP-AdaBoost classifier was trained by the fusion features. Finally, the trained classifier and the recognition algorithm were used to detect the dust collector and fastening bolt faults. The experimental results show that the algorithm may adapt to the recognition of two different faults. High recognition accuracy rate, low false ratios and low omission ratios are obtained, and the algorithm is robust to light unevenness and occlusion.