养殖水体中校色卡识别与色彩变化分析  

Identification and analysis of color changes in aquaculture waters using color checker

作　　者：李佳康 张胜茂[1] 吴祖立[1] 石永闯 唐峰华[1] LI Jiakang;ZHANG Shengmao;WU Zuli;SHI Yongchuang;TANG Fenghua(Key Laboratory of Fisheries Remote Sensing,Ministry of Agriculture and Rural Affairs,East China Sea Fisheries Research Institute,Chinese Academy of Fishery Sciences,Shanghai 200090,China;College of Information,Shanghai Ocean University,Shanghai 201306,China)

机构地区：[1]农业农村部渔业遥感重点试验室,中国水产科学研究院东海水产研究所,上海200090 [2]上海海洋大学信息学院,上海201306

出　　处：《渔业信息与战略》2024年第1期49-60,共12页Fishery Information & Strategy

基　　金：崂山实验室专项经费(LSKJ202201804);中国水产科学研究院基本科研业务费(2020TD82);国家自然科学基金项目(61936014)。

摘　　要：水下环境中颜色变化是一种常见现象,研究其变化特征可应用于水下图像处理和水产养殖生物监测等。利用校色卡分析不同水深颜色变化的特点,实验中首先对图像进行大小归一化及颜色空间转换;其次用图像与标准色块对比计算差值图像;然后进行轮廓提取并筛选出可能的色块加入候选列表,通过候选列表中的色块计算出原图色卡的位置;再通过仿射变换将原图色卡提取出来;最后,通过比较不同水深图像中色卡色块的颜色值,分析水下环境中的颜色变化。实验结果表明,该方法能够有效地检测和分析水下环境中的色卡,在0~3 m的浅水域中检测准确率达到100%,在3~10 m的中层水域中检测准确率达到93%,在大于10 m背景较为复杂深水域中提取准确率达到73.7%,在占用极小内存空间的情况下达到了较高的识别准确率;通过水下颜色分析得出:随着水深的增加,红色光的吸收和衰减最明显,蓝色光的吸收和衰减也比较多,而绿色光的影响较小。研究内容为利用水下摄影研究水产养殖生物行为提供了参考方法。Color change is a common phenomenon in underwater environments,and studying its characteristics can be helpful in its application in underwater image processing and aquaculture biological monitoring.By analyzing the characteristics of color change at different water depths using color cards,the experiment first normalizes the size and color space of the images.Then,it calculates the difference of image by comparing the image with standard color blocks.Next,it extracts the contours and selects possible color blocks to add to the candidate list,from which it calculates the position of the original color card.The original color card is then extracted using affine transformation.Finally,the color values of color card blocks in different water depths are compared to analyze the color changes in underwater environments.The experimental results show that this method can effectively detect and analyze color cards in underwater environments.In shallow water area with water depth of 0-3 m,the detection accuracy reaches 100%.In mid-water area with water depth of 3-10 m,the detection accuracy reaches 93%.In deep water area with more complex backgrounds and water depth more than 10 m,the extraction accuracy reaches 73.7%.It achieves high recognition accuracy while occupying the minimal memory space.Through underwater color analysis,it is concluded that as water depth increases,the absorption and attenuation of red light are most pronounced,followed by blue light,while the impact of green light is relatively small.The research aims to provide a reference method for studying the behavior of aquaculture organisms using underwater photography.

关 键 词：校色卡 OPENCV 图像处理 色卡检测 颜色空间 水产养殖 

分 类 号：S951.2[农业科学—水产养殖]