The Method for Pointer Instrument Detection and Reading Recognition in Foggy Conditions
Aiming at the oilfield inspection robots'missed detection and low recognition accuracy of the pointer instrument in foggy weather,a method which combining FFA-Net dehazing network and Yolov5 de-tection algorithm was proposed.Firstly,it has a FFA-Net algorithm improved for the foggy images and has the multi-scale structure and the feature fusion residual block and optimization module based to effectively enhance the algorithm's performance in dehazing operation;and then,as for the instrument detection and reading recognition,it has a smaller detector head adopted to improve Yolov5's ability in detecting small targets,and has a spatial transformation module introduced to convert the dial's images detected into a front elevation in line with human perception;finally,it has an end-to-end framework created to tightly couple the meter component retrieval and meter reading recognition so as to improve the accuracy of meter readings.The experimental results show that,the proposed method boasts good robustness in the oilfield foggy environment,and it improves both accuracy of detection and reading recognition of the pointer meter in foggy environment.
万方数据
维普
