This paper presents a new neuroimmune-based method for block recognition of railway rolling stock inventory numbers. The advantage of such approach is classification without using of negative samples. Developed technique combines segmentation and classification that allows to achieve higher noise robustness, segmentation possibility of fuzzy combined digits which have different fonts and typeface, and invariance of existing numbers to scale changes. Proposed method allows to constantly increase the training set for the improvement in classification accuracy by new committee classifiers statistics due to the data reduction property achieved by using the immune clustering mechanism. Research results were implemented in the software system of automatic recognition of cars numbers (ARNV), which is operated on the JSC Russian Railways.
Keywords: Method for letters block recognition, the committee neuroimmune classification model, identification, automatic recognition car number, duplicate number
Identification of railway vehicles is relevant for the conversion from automatic control systems with manual data input to automatic modeling environment of train and vehicle. The most effective and economic inventory number recognition is optical recognition. But there is question about veracity in such technology. This paper represent the qualitatively new approach for optical recognition based on building integral robust constructive features of vehicles and allowed to significantly increase level of recognition veracity. Proposed technique investigated in introduction subject of automatic vehicle number recognition system (ARNV). Computational experiments demonstrated relevance of proposed technique for using in optical recognition of vehicle numbers.
Keywords: Automatic systems, number identification, vehicle, optical recognition, ARNV, robust features