Detection of Fish Cage Net Damage Using Image Processing with Mesh-Hole Grouping

Abstract

In this paper, we propose a Mesh-hole grouping algorithm that detects damaged areas by comparing the area between neighboring net holes in order to detect damaged parts of net that occurs in a wagging fish cage net. An image pre-processing is performed to extract the net shape from the underwater net image and convert it into a binary image. Each net hole in the binarized net image is assigned a number, and the net holes adjacent to the reference net hole are grouped together into one group. These grouped net holes are then arranged in ascending order based on their area size. Then, if the difference between the area of the first widest hole and the area of the second widest hole in the group is greater than the average hole area of the corresponding group, it is detected as damaged. The net damage detection algorithm was evaluated on a dataset of 600 images and achieved the following performance metrics: accuracy 0.86, precision 0.86, and recall 0.88.