ID localization and recognition for railway oil tank wagon in the industrial scene

This paper presents a localization and recognition method for railway oil tank wagon ID in the industrial scene. Firstly, we use the maximum stable extremal regions (MSER) algorithm to obtain the extreme regions, and extract effective areas from the extracted extremal regions. Then the effective areas of the regions are merged into triplets region sequence as an ID candidate region, and we select the ID region through linear distance estimation method. For overcoming the interference from camera installation angles, we make use of four-point correction method to correct ID areas, and an improved projection method to segment characters. Finally, we train Tesseract-OCR text recognition classifier to recognize ID characters. Experimental results show that the proposed method has a good localization and recognition performance for railway oil tank wagon ID in the complex industrial scene.