The daily detection of highspeed electric multiple units (EMU) body is very important for China railway maintenance system. This paper proposes a new method based on machine vision to detect bolts and switches on EMU side skirt, which aims to replace manpower. Yolov3 network is used to identify and locate bolts and switches. After positioning, the status of bolts is detected through Alexnet network. Experiments show that the processing method can achieve the detection work efficiently and accurately. It is suitable for many cases such as insufficient illumination and non-vertical shooting of image samples, which has high robustness.

No abstract

No abstract