A Visual Identification Method of Analog Instrument Panel Based on Faster R-CNN

Intelligent inspection robots are being widely used in unmanned gas stations. However, the complicated background environment is a disturbance for most of the current visual identification method of analog instrument panel. It is challenging to extract pointer centerline and surface area. The existing methods cannot satisfy the actual demand of gas stations to get the current problems solved. In this work, we presents an object detection algorithm based on faster R-CNN method and designed a two-stage automatic identification method of analog instrument panel combined with conventional computer vision algorithms. First, the improved faster R-CNN is utilized for the detection of the target dashboard area. Then, the image is preprocessed based on some image processing methods. Moreover, the pointer’s centerline is detected by the algorithm based on Hough transformation. Finally, the meter identification is calculated by the angle calculation method. The performance verification and analysis of the algorithm demonstrates that the method proposed in this work is reliable and efficient for automatic identification of instrument panel in the working condition of gas stations and provides a quite practical method for target detection and identification of instrument panel.