Abstract For offshore aquaculture cages, structural damage can cause serious economic property damage to the aquaculture industry, thus it is necessary to carry out regular inspections of the cage structure. At present, the main method of detecting netting damage is that the staff sneak into the water for manual investigation. This method not only does not guarantee personal safety, but is also labor-intensive and inefficient. This paper proposes a new feature curve method for non-contact underwater netting damage detection, which can effectively recognize damaged netting and analyze the degree of damage for the underwater net and return to its damage position. The new method presented in this paper was designed taking into consideration of effect of seaweed growth: Firstly, an image block within the ROI (region of interest) region was processed by bilateral filter. Secondly, the binary image was obtained via the OSTU (the maximum inter-class variance method) method and connected domain detection was performed. Thirdly, the feature gradient histogram was calculated according to the area of the mesh hole and the local peaks of the curve (named as feature curve) was searched to determine the position of damage in the netting. The proposed method combined the image processing technology with aquaculture engineering seamlessly, and reduced the complexity of the detection system greatly, and significantly improved the efficiency of the netting detection. Finally, the MATLAB program was developed to realize the netting detection process and the proposed method had been verified by the actual underwater netting experiment. The experimental results showed that the netting damage detection method proposed in this paper can successfully detect crack of the netting despite image degradation in water.

中文翻译：

---

基于图像处理技术的渔网损伤检测自适应方法

摘要 对于近海养殖网箱，结构损坏会对养殖业造成严重的经济财产损失，因此有必要对网箱结构进行定期检查。目前，检测渔网破损的主要方法是工作人员潜入水中进行人工调查。这种方法不仅不能保证人身安全，而且劳动强度大，效率低下。本文提出了一种新的非接触式水下渔网损伤检测特征曲线方法，能够有效识别受损渔网，分析水下渔网的损伤程度并返回其损伤位置。本文提出的新方法是考虑到海藻生长的影响而设计的：首先，ROI（感兴趣区域）区域内的图像块由双边滤波器处理。其次，通过OSTU（最大类间方差法）方法获得二值图像并进行连通域检测。第三，根据网孔面积计算特征梯度直方图，搜索曲线（称为特征曲线）的局部峰值，确定网中损伤的位置。该方法将图像处理技术与水产养殖工程无缝结合，大大降低了检测系统的复杂度，显着提高了网状检测的效率。最后，开发了MATLAB程序来实现结网检测过程，并通过实际的水下结网实验验证了所提出的方法。