图像噪声对无人机水面目标识别率的影响研究  

Study on the Impact of Image Noise on Water Surface Target Recognition Rates

作　　者：涂强 王留杰 TU Qiang;WANG Liujie(Guangdong Research Institute of Water Resources and Hydropower,Guangzhou 510635,China)

机构地区：[1]广东省水利水电科学研究院,广州510635

出　　处：《广东水利水电》2024年第12期88-93,共6页Guangdong Water Resources and Hydropower

摘　　要：随着无人机和高分辨率卫星技术在水利行业的广泛应用,水面目标图像识别的需求增多,精度要求提高。该文研究图像噪声对水面目标识别影响,达到优选降噪方法的目的。通过对含有高斯噪声,脉冲噪声及其他复合噪声的真实图像数据集,基于深度学习图像识别模型,分析不同噪声水平对图像识别率的影响;对比传统与深度学习降噪技术对提高目标识别准确性的作用,结果表明:适当的降噪处理显著增强了模型在噪声影响下的鲁棒性和准确性,凸显了在水面目标识别中应用高效图像降噪技术的重要性。With the development of storage and transmission technologies and the advancement of observational methods such as drones and high-resolution satellites,image data sources in hydrological systems have grown significantly,leading to a corresponding increase in the demand for object recognition within image data.This paper focuses on studying the impact of image noise on the recognition rate of hydrological targets.By introducing Gaussian noise,impulse noise,and other composite noises into a self-constructed real-world dataset and utilizing deep learning image recognition models,we analyze the effects of different noise levels on recognition rates.Additionally,we explore the role of traditional and deep learning denoising techniques in improving target recognition accuracy.The results demonstrate that appropriate denoising significantly enhances the robustness and accuracy of models under noisy conditions,underscoring the importance of applying efficient image denoising techniques in water surface target recognition.

关 键 词：水面目标识别 图像噪声 深度学习 识别率 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] P231[自动化与计算机技术—计算机科学与技术]