摘要： 【目的】为解决传统生物污损评价难以量化、效率低、主观性强等问题，提出基于图像识别的自动化定量评价方案。【方法】在舟山六横岛海域开展 7个月实海浸泡实验，完整记录了六横岛海域 3类样板的污损情况，覆盖了污损生物生长淡季到旺季。采用百分格度板法、标注掩码法和图像识别法对比分析污损图像，并建立改进的污损值评价模型。【结果】 U-Net模型对污损图像的识别精度较高，识别速度较人工方法提升 2个数量级，仅为 71. 6 ms/张。通过引入物种权重系数提出污损值评价模型， 6—9月期间空白板污损程度从 15. 5上升到 157. 4。水解型涂层 A较自抛光型涂层 B具有更低的污损值和更好的防污性能，表明不同实海条件下的涂层防污性能有差异。【结论】图像识别技术为海洋生物污损定量评价和防污涂层性能检测提供了高效可靠的技术手段，显著提升了污损评价的准确性与合理性。

关键词: 生物污损, 图像识别, 量化评价, 涂层, 防污性能

Abstract: ［Objective］ To address the issues of traditional biofouling evaluation，such as difficulty in quantification，low efficiency，and strong subjectivity，an automated quantitative evaluation scheme based on image recognition technology was proposed.［Methods］ A seven-month marine immersion test was conducted in the waters of Liuheng Island，Zhoushan. We fully documented the biofouling progression on three types of test panels，covering the period from the off-season to the peak season of fouling organism growth. The biofouling images were analyzed and compared using the quadrat method，the annotation masking method and the image recognition method，and an enhanced fouling rating evaluation model was established.［Results］The U-Net model achieved high recognition accuracy for biofouling images，and its recognition speed was two orders of magnitude faster than manual methods，requiring only 71. 6 ms per image. By introducing species-specific weighting coefficients，a fouling rating evaluation model was proposed. From June to September，the fouling ratingof blank panel increased from 15. 5 to 157. 4. Hydrolytic coating A exhibited a lower fouling rating and superior anti-fouling performance than self-polishing coating B，indicating that the anti-foulingefficacy of coatings differed under different actual marine conditions.［Conclusion］Image recognitiontechnology provided an efficient and reliable technical means for the quantitative assessment of marinebiofouling and the evaluation of anti-fouling coating performance，and significantly improved the accuracy and rationality of fouling evaluation.

Key words: biofouling；image recognition；quantitative evaluation；coating；anti-fouling performance 

中图分类号: 

• TQ637. 3