The detection of power safety signs is an important component of the intelligent power safety operation control system.To improve the detection accuracy of safety signs in complex power scenarios,a power safety sign detection method based on improved YOLO was proposed.On the basis of YOLO,the network's prediction ability for small targets was improved by increasing the resolution of the prediction layer.In addition,the introduction of index pooling mechanism and the use of pooling masks to limit the introduction of useless information improved the accuracy of network classification and recognition.The experimental results show that the average accuracy of the improved detection network on the power sign test set reaches 75.2％,which is 3.2％higher than the conventional method.The proposed intelligent detection method can improve the detection and recognition ability of power signs,which is conducive to ensuring the safety of power production.
关键词
电力标识牌/深度学习/目标检测/索引池化/网络识别
Key words
power signboard/deep learning/object detection/index pooling/network identification
维普
万方数据