基于YOLOv5s的配电台区施工多目标检测方法OACSTPCD

There are many construction projects in the distribution platform area,and the standardization and standardization of the construction personnel are low. The use of object detection algorithm to control the construction process can effectively ensure the quality of the project. Common object detection algorithms require high storage and computing power of devices,so how to deploy lightweight algorithms to edge devices has become the focus of research. In order to improve the detection accuracy of equipment construction identification in distribution station area and consider the demand of model lightweight,this paper proposes a multi-object detection algorithm based on YOLOv5s. Firstly,the bottle2neck module of improved Res2Net network was used to extract fine particles and multi-scale features to achieve multi-scale image feature extraction,ensuring model accuracy and lightweight. Secondly,based on bottle2neck module,a B4-Cat optimization model with higher detection accuracy is proposed. Finally,the advantages of the model are verified by the data of distribution station construction provided by certain region. The results show that compared with the existing algorithms,the model parameters and calculation amount of the proposed method are reduced by more than 25%,and the mAP index is more than 81%,which is better than the commonly used depth separable convolutional lightweight method,and is conducive to improving the intelligent management and control level of distribution station construction.