为正确判断车厢连接处钩头是否为死钩,避免由于翻车机所翻卸火车厢两端均为死钩时造成的重大事故.采用图像特征提取的方法,将车皮特定区域的特征提取出来,经过适当筛选,将所得特征值作为神经网络的输入,通过训练好的网络,进而得出车皮活钩、死钩情况.通过现场采集一定数量的车皮图像作为实验数据,选取其中具有代表性的图像作为神经网络的训练集对BP网络进行训练,利用现场采集到的另外一组图像数据作为验证集,通过实验证实方法的可行性和准确率.%In order to accurately identify whether the train coach junction is fixed-hook or not, and avoid the accident causeed when dumping both sides of the train coach are fixed-hooks. This paper adopts image feature extraction method, which to extract special region feature of the wagon images and select. After appropriate screening, the eigen value is regarded as neural networks input, through the trained network, and a conclusion is reached. Collect a certain number of wagon images as experimental data. Those typical images are as the BP neural network training sets . Use another group images as confirm sets. It is confirmed from experiments that this method is feasible and accuracy.
展开▼