| 摘要: |
| 海上紫菜养殖筏架分布无规律、大小不规则且数量较多，现场测量难度大、卫星影像空间分辨率低，不能精确测量出紫菜筏架面积。无人机机动性强、影像空间分辨率高，可在紫菜养殖调查中发挥重要作用。本研究以连云港海州湾紫菜养殖筏架为研究对象，开展可见光波段在养殖筏架与水体的光谱可区分度研究，基于6种植被指数，进行自动提取实验，以目视解译结果作为真值，进行精度分析，同时利用不同时相、不同区域的无人机可见光影像，开展方法的普适性研究。结果表明：绿色度坐标（green， G）和植被（vegetativen，VEG）指数方法在浅水区和深水区均表现较好，养殖筏架个数识别精度、面积识别精度均超过91.00%。基于此，利用上述两种方法，开展其他区域自动提取实验，养殖筏架个数识别精度、面积识别精度分别超过93.02%和89.37%。结果验证了无人机可见光影像可以实现紫菜养殖筏架的自动提取，精度基本满足紫菜养殖调查需求。 |
| 关键词: 海洋物理学 无人机 可见光 紫菜养殖筏架 自动提取 连云港 |
| DOI：10.3969/J.ISSN.2095-4972.20220902001 |
|
| 基金项目:自然资源部国土卫星遥感应用重点实验室2023年度开放基金（KLSMNR-G202306）；国家自然科学基金（41706105，41506106）；国家重点研发计划（2018YFB2100705）；连云港市测绘地理信息科研项目（LYGCHKY201902） |
|
| Information extraction method applied for laver farming rafts based on UAV visible light image |
| XU Haipeng,GAO Guangjun,YANG Hui,LU Xia,SUN Yuxi,CHU Jialan |
| (Oceanic Administration and Protection Center of Lianyungang , Lianyungang 222001, China;Satellite Application Technology Center of Lianyungang, MNR, Lianyungang 222001, China;School of Marine Technology and Geomatics, Jiangsu Ocean University, Lianyungang 222005, China;National Marine Environmental Monitoring Center, Dalian 116023, China) |
| Abstract: |
| The field monitoring of laver farming rafts is difficult as they have irregular sizes of distribution and large numbers. The satellite images cannot provide accurate area of laver farming rafts due to low resolution. In contrast, UAV plays an important role in raft aquaculture surveys for its strong maneuverability and high image resolution. This paper takes the floating raft aquaculture in Haizhou Bay of Lianyungang as the research object, and studies the distinguishable degree of the visible light spectra of the breeding raft and the water body. The automatic extraction experiments for floating raft aquaculture are performed based on 6 vegetation indexes, and the accuracies of these experiments are analyzed by comparing the automatically extracted estimates with the visual interpretation values. Moreover, the applicability of the methods is analyzed based on UAV images obtained at different phases and regions. Results show that the green light band (Green, G) and vegetation index(Vegetativen，VEG) methods perform well whether in shallow waters or in deep waters with accuracies exceeding 91.00% for both number and area of laver farming rafts. These two methods also perform well in other seas with accuracy for number exceeding 93.02% and for area exceeding 89.37%. The paper reveals that UAV visible image can realize the automatic extraction of laver farming rafts and the accuracy meets the needs for laver farming surveys. |
| Key words: marine physics UAV visible bands laver farming rafts automatic extraction Lianyungang |
