基于改进YOLOv5s和Canny模型的货车物料超限检测方法研究

道路上不断出现货车超限的情况,对人们生命财产安全造成很大影响。而检测货车超限的前提是准确识别货车,为了提高检测货车的速度与精度并判别装载物料是否超限,该文提出一种基于改进YOLOv5s和Canny模型的货车物料超限检测方法。采用轻量级提取网络ShuffleNetV2作为YOLOv5s网络的主干部分,以此来减少模型的计算量和参数量;用EIoU损失函数优化边框回归,精确货车目标定位;为检测对货车图像进行灰度化和边缘检测处理,再与标准超限体比较,可判断货车是否超限。实验结果表明,在自制货车数据集上,改进YOLOv5s的平均精度可达95.8%,比原模型提高了9.2%,参数量和计算量为原模型的45.4%、44.4%。最后将处理后图像与标准超限体对比,可检测货车是否超限,有利于消除道路交通安全隐患,为城市治超管理提供方法。

In view of the fact that trucks continue to exceed the limit on the road,the safety of people's lives and property has been greatly affected,while the premise of determine whether the truck exceeds the limit or not is accurately recognizing the truck loading,this paper proposes a truck material out-of-limit detection method based on improved YOLOv5s and Canny algorithm,in order to improve the speed and accuracy of truck detection.The lightweight extraction network ShuffleNetV2 is used as the backbone of the YOLOv5s network to reduce the amount of calculation and parameters of the model;the frame regression is optimized by EIoU loss function to accurately locate the truck target;the detected truck image is grayed out and edge detection is processed,and then compared with the standard out-of-limit body,whether the truck is out of limit can be judged.The experimental results show that on the self-made truck data set,the average accuracy of the improved YOLOv5s is 95.8%,which is 9.2%higher than the original algorithm,and the number of parameters and the amount of calculation are 45.4%and 44.4%of the original algorithm.Finally,the processed image is compared with the standard out-of-limit body,which can detect whether the truck is out of limit,which is helpful to eliminate the hidden danger of road traffic safety and provide a method for urban management.