基于深度学习的输电通道入侵物体识别方法研究

Research on Intrusion Object Recognition Method of Transmission Corridor Based on Deep Learning

针对输电通道在线监测过程中入侵物体大小差异巨大、部分图像对比度低等问题,结合异物图像的特征,提出了一种基于目标检测算法的输电通道入侵物体识别方法。采集输电通道入侵物体图像,利用Retinex算法对输入图像进行增强。在目标识别部分,采用改进的EfficientDet算法作为主体,对算法中锚框的长宽比采用K-means聚类算法进行优化,同时在损失函数中加入了梯度均衡机制。实验结果表明,改进后的算法将mAP值从83.72%提升至87.12%,在入侵物体识别任务上有着优异的性能。

In order to solve the problems of huge difference in size between intrusion objects and low contrast of some images in the process of online monitoring of transmission corridor,combining with the characteristics of foreign object images,a transmission line intrusion object recognition method based on object detection algorithm is proposed.Firstly,the intrusion object images of transmission corridor are collected,and the input images are enhanced by Retinex algorithm.In the part of object recognition,the improved EfficientDet algorithm is adopted as the main body,and the length-width ratio of the anchor frame is optimized by K-means clustering algorithm.Meanwhile,the gradient equalization mechanism is added into the loss function.Experimental results show that the mAP value of the improved algorithm increases from 83.72%to 87.12%,and it has excellent performance in the intrusion object recognition task.