| 摘要: |
| 海洋牧场是实现水产资源可持续利用的重要手段，精确识别其中的经济生物对于生态监测与资源管理具有重要意义。本文构建了一套基于推扫式成像原理的水下高光谱成像系统，旨在满足对海洋牧场经济生物探测识别需求。在实验室条件下，使用该系统对五种海洋牧场经济生物（虾夷扇贝、栉孔扇贝、仿刺参、脉红螺和皱纹盘鲍）进行了静态高光谱成像，基于光谱角制图（SAM）、随机森林（RF）和支持向量机（SVM）三种算法进行了分类实验，分类精度分别达到71.2894%，81.0478%和85.0482%，验证了水下高光谱在海洋牧场典型物种识别中的适用性与构建系统的稳定性。虽然本研究以实验室测试为主，但所采集的高分辨率光谱数据在海洋牧场生物分类中展现出良好的技术适用性与支撑潜力。尽管仍面临光谱重叠、图像畸变与数据处理压力等挑战，但其非接触、高分辨率等技术优势使其在海洋牧场资源监测、生态评估与智能化管理等领域展现出广阔的应用前景。 |
| 关键词: 水下高光谱 推扫式成像 光谱特征分析 水下生物分类 |
| DOI： |
| 投稿时间：2025-02-24修订日期：2025-04-16 |
| 基金项目:国家重点研发计划“南海海洋牧场构建与生态农牧化开发新模式(2022YFD2401304)“项目 |
|
| Application of Pushbroom Hyperspectral Imaging Technology for the Classification and Identification of Economically Important Marine Organisms in Ocean Ranching |
|
tangtianyi, duzengfeng, zhangjianxing, luanzhendong
|
| (IOCAS) |
| Abstract: |
| Marine ranching is a vital approach to achieving the sustainable utilization of aquatic resources, and the accurate identification of economic species is essential for ecological monitoring and resource management. In this study, a push-broom-based underwater hyperspectral imaging system was developed to meet the detection and classification needs of economically important species in marine ranch environments. Under laboratory conditions, hyperspectral imaging was conducted on five representative marine ranch species—Mizuhopecten yessoensis, Chlamys farreri, Apostichopus japonicus, Rapana venosa, and Haliotis discus hannai. Supervised classification experiments were performed using three commonly used algorithms: Spectral Angle Mapper (SAM), Random Forest (RF), and Support Vector Machine (SVM), achieving classification accuracies of 71.2894%, 81.0478%, and 85.0482%, respectively. These results verified both the applicability of underwater hyperspectral imaging for marine species identification and the operational stability of the developed system. Although this research was conducted primarily under laboratory conditions, the acquired high-resolution spectral data have demonstrated good technical applicability and potential to support marine ranch resource evaluation and ecological monitoring. Despite ongoing challenges such as spectral similarity, image distortion, and data processing complexity, the non-invasive nature and high spectral resolution of underwater hyperspectral imaging offer promising prospects for applications in marine ranch monitoring, ecological assessment, and intelligent management. |
| Key words: Underwater hyperspectral, Pushbroom imaging, Spectral feature analysis, Underwater organism classification |
