基于改进YOLOv9c的海洋垃圾检测研究

Marine Litter Detection Based on Improved YOLOv9c

为了提高解决海洋垃圾问题的效率,保护海洋环境。文章提出了一种基于改进YOLOv9c的的海洋垃圾检测算法的研究。由于海洋垃圾浮游于海中,受较暗光线以及海水颜色的影响较难检测识别,在预处理时对数据进行色彩增强与图像增亮的处理,提高了图像的辨识度。并采用最新的YOLOv9c作为目标检测的骨干网络,引入Squeeze and Excitation注意力机制,提高了特征的敏感度,增强了网络的泛化能力和效率。并且将下采样替换为基于Haar小波下采样,在降低特征图的同时尽可能保留更多的信息,提高处理的效率。经训练后,在J-EDI海洋垃圾数据集上进行验证,其mAP达到了70.5%,模型的参数也只有12.5M,FPS为75。表明改进后的算法有较好的效果。

In order to improve the efficiency of solving the marine litter problem and protect the marine environment.The article proposes the study of a marine litter detection algorithm based on improved YOLOv9c.Since marine litter floats in the sea and is more difficult to detect and identify due to the influence of darker light and the color of seawater,color cnhancement and image brightening are applied to the data during preprocessing to improve the recognition of the image.The latest YOLOv9c is used as the backbone network for target detection,and the Squeeze and Excitation attention mechanism is introduced to improve the sensitivity of features and enhance the generalization ability and efficiency of the network.And the downsampling is replaced with Haar wavelet-based downsampling to retain as much information as possible while reducing the featurc map and improve the efficicency of procssing.After training,it is validated on the J-EDI marine litter dataset,and its mAP reaches 70.5%,and the parameters of the model are only 12.5M,and the FPS is 75.It shows that the improved algorithm has better results.