An Improved Recognition Method of EMU Surface Apron Board Defects Based on Depth Learning

An Improved Recognition Method of EMU Surface Apron Board Defects Based on Depth Learning

EMUs play an important role in the rapid development of China’s economy, so ensuring the safety of high-speed EMUs has been a research issue for relevant experts and scholars. Because the defect change of the motor car apron will lead to the stability of the whole motor car running on the side, this paper takes the motor car apron defect as the recognition object, and proposes to add feature reasoning module to the typical convolutional network model to improve the ability to identify occlusion defects by solving the problem that the traditional recognition model cannot identify the features of occlusion and stacking and the status quo that cannot be identified under a small number of samples, Then the meta learning mechanism is incorporated into the model to realize the recognition under the condition of few samples in the dataset. Through experimental comparison, it is proved that the performance of the improved recognition network algorithm in recognition accuracy and recognition speed is improved as a whole.