基于改进YOLOv5s的水下自主机器人垃圾检测技术研究

Underwater Autonomous Robot Garbage Detection Technology Based on Improved YOLOv5s

摘要:

研究针对水下环境垃圾污染问题,提出了一种基于改进YOLOv5s的水下自主机器人垃圾检测方法；该方法采用MSCA模块结合轻量化网络MobileNet V3,对原模型的主干网络进行优化；同时引入了FCS模块,对模型的颈部进行改进；为了进一步提高模型的检测精度,研究加入了SIoU损失函数；实验结果表明,通过图像处理技术对水下图像进行预处理后,其图像质量有所提升,处理后的图像色彩度和图像清晰度评估值分别为2.65和0.59；在复杂水环境下,改进后的YOLOv5s模型平均精度达到0.948,损失函数值为0.0008；实验测试结果证明了改进的YOLOv5s水下垃圾检测技术能够有效提升模型的检测精度,且在实际应用中具有较优的性能；该技术经实际测试满足了水下生态监测和水下环境保护的应用。

Abstract:

Aiming at the problem of underwater environmental garbage pollution, an underwater autonomous robot garbage detection method based on improved YOLOv5s was proposed. The multi-spectral channel attention module combined with lightweight network MobileNet V3 is used to optimize the backbone network of the original model. At the same time, the feature pyramid network, channel attention mechanism and space pyramid pool network are introduced to improve the neck layer of the model. In order to further improve the detection accuracy of the model, SIoU loss function is added. The experimental results show that the image quality of underwater images is improved after image processing, and the evaluation values of image color and image sharpness are 2.65 and 0.59, respectively. In complex water environment, the average accuracy of the improved YOLOv5s model reached 0.948, and the loss function value was 0.0008. The experimental test results prove that the improved YOLOv5s underwater garbage detection technology can effectively improve the detection accuracy of the model, and has better performance in practical application. The technology meets the application of underwater ecological monitoring and underwater environmental protection by practical test.