浑浊水体中模拟水生动物的识别

打开文本图片集

中图分类号：O436 文献标志码：A 文章编号： 1000-5013（2025）04-0442-06

Recognition of Simulated Aquatic Animal in Turbid Water Environments

WANG Xiaoyan 1，2 ，CHAI Yuman 1，2 ，PU Jixiong 1，2

（1.College of Information Science and Engineering，Huaqiao University，Xiamen 361o21，China; Fujian Key Laboratory of Light Propagation and Transformation，Huaqiao University，Xiamen 361o21，China）

Abstract：Using turbid water as a model，this study investigates object recognition in dynamic scatering environments.. Taking simulated aquatic animal as an example，a neural network is constructed and trained to identify the species and quantity of aquatic animals through deep learning techniques. When untrained speckle images are input into a trained neural network，it outputs the categories and number of aquatic animals. The experiment results demonstrate that deep learning techniques can successully identify both the categories and quantity of aquatic animals in turbid water. The accuracy of quantity recognition reaches 100% ，while the accuracy for species recognition exceeds 99% across all tested categories.

Keywords：aquatic animals；image recognition；dynamic scatering；turbid water;deep learning；neural network

在许多生活场景和科学研究中都存在光经过散射介质传输的现象，例如，雾霾环境的光学成像、烟雾场景的目标探测、多模光纤的信息传输等。（剩余8037字）