基于改进YOLOv5的安全绳目标检测

Safe rope target detection based on improved YOLOv5

摘要:

在工业施工过程中, 工人安全已成为一个日益重要的问题, 佩戴安全绳等安全装备是保护工人在高处工作时生命安全的重要措施；在现代化生产施工过程中, 通过使用监控摄像设备结合人工智能算法的方式来检测工人佩戴安全绳等设备越发普遍, 但安全绳由于细长、形状多变以及环境变化等因素较为难以准确识别；为解决以上问题, 并确保能够在不同环境下能够准确识别安全绳, 现提出一种使用YOLOv5目标检测算法, 首先通过改进的FasterNet模块进行上下文信息提取, 在Neck网络中使用改进的多维动态卷积保留更多特征信息, 使用WIoU_Loss损失函数来提高定位精度, 在训练过程中使用动态调整学习率的策略；实验结果表明, 改进后的算法在降低计算复杂度的情况下提高了3.0%的检测精度, mAP@0.5提高了4.3%, 经过在实际场景应用, 满足项目对实时检测精度及速度的要求。

Abstract:

In the process of industrial construction, worker safety has become an increasingly important issue. Wearing safety equipment such as safety rope is an important measure to protect workers" life safety when working at height. In the process of modern production and construction, it is more and more common to use surveillance camera equipment combined with artificial intelligence algorithm to detect workers wearing safety ropes and other equipment, but safety ropes are difficult to accurately identify due to factors such as slender, changeable shape and environmental changes. In order to solve the above problems and ensure that the safety rope can be accurately identified in different environments, an object detection algorithm using YOLOv5 is proposed. Firstly, context information is extracted by the improved FasterNet module, and more feature information is preserved by the improved multidimensional dynamic convolution in the Neck network. The WIoU_Loss loss function is used to improve the positioning accuracy, and the strategy of dynamically adjusting the learning rate is used in the training process. Experimental results show that the improved algorithm improves the detection accuracy by 3.0% and mAP@0.5 by 4.3% under the condition of reducing the computational complexity. After application in actual scenarios, the proposed algorithm can meet the requirements of the project on real-time detection accuracy and speed.