K-均值聚类算法在海水背景石斑鱼彩色图像分割中的应用

摘要

Fish’s image segmentation is the key and prerequisite step of the behaviour observation and intelligent automatic feeding system for farmed fish.For fish segmentation of the image with seawater background,we proposed a hardware and software system in regard to acquiring colour RGB image of grouper farmed in artificial seawater environment,and applied k-means clustering algorithm in the experiment of RGB format images’segmentation.We observed and analysed the histogram characters of three planes of monochromatic pixels (R,G or B)of colour image.Using three-colour pixel points (RGB)and three planes of monochromatic pixels (R,G or B)as the clustering data sets,we employed k-means clustering algorithm for image segmentation,the clustering number value was set to 2,representing the seawater background and fish object respectively.Through observation and statistics of 100 image segmentation results,it could draw a preliminary conclusion that the results of using blue (B)plane clustering segmentation satisfied the most.%鱼类图像分割是养殖鱼类行为观测和智能化自动投饵系统的关键先决步骤。针对海水背景图像中鱼体分割，提出一种在人工海水环境中获取石斑鱼彩色 RGB 图像的软硬件系统，并将 K-均值聚类算法应用于 RGB 格式图像的分割实验。观察并分析了彩色图像的三种单色（R、G 或 B）像素的直方图特征。以 RGB 三色像素点和三种单色（R、G 或 B）像素点为聚类数据集，使用 K-均值聚类算法进行图像分割，聚类数设定为2，分别代表海水背景和鱼体目标。通过对100幅图像分割结果的观察和统计，可以得出初步结果，即使用蓝色（B）通道聚类分割的结果，最令人满意。