Train Detection And Applied Research In Varying Illumination

The Ministry of Railways promotes two sets of automatic safety detection system: Trouble of Moving Freight Car Detection System(TFDS) and Trouble of moving EMU Detection System(TEDS). The two systems are designed to improve the efficiency of the train dynamic fault detection by automatic detection of the computer. Trains run complex environment in which illumination variations greatly affected the train detection. This paper focuses on the target positioning, segmentation, feature extraction and recognition of the train image in varying illumination, and makes the following algorithms:For the truck brake detection under the array capturing device and the spot light source, first based on the discrete-point sampling and the significant edge features to preliminary position the target region; then proposed the image segmentation algorithm based on edge and regional structure to segment target region, and finally recognize target based on edge feature and regional feature.For the character recognition under the linear capturing device and the linear light source, first preliminary position the characters region based on gray gradient; then segment characters region based on the local edge information; next divide the character sequence according to the number of characters, the distance in class and the distance between classes, and normalize the character image, finally study and design the character features and classifier, and compare different classifiers, to recognize the character.Experimental results show that the above method can effectively solve the truck brake detection under the array device and the spot light source, and the character recognition under the linear array device and the linear light source.