Abstract: In this paper, aiming at the problem of the automatic identification of freshwater fish species, using the machine vision technology to put forward a kind of method for the fish species identification, which based on the correlation coefficient algorithm of the fish back contour. Firstly, according to the four kinds of freshwater fish pictures of crucian, grass carp, bream and cyprinoid, the image processing method is applied to the extraction of the fish back contour. And the least square algorithm is used for fitting a curve of the fish body back contour to establish the mathematical model of the back contour of four kinds of freshwater fish. Then, in order to achieve the automatic identification of fish species, for the fish body to be recognized, using the machine vision technology to obtain the fish body contour, and calculating the correlation coefficient value of the fish contour and the mathematical model of the back contour of four kinds of freshwater fish. Finally, the 60 fishes alive are randomly selected from the market have been tested by the proposed method, which shows that the algorithm is simple and has higher recognition accuracy. It can provide new ideas for the identification of the freshwater fish species and improve the automation level of aquaculture.

Key words: machine vision, the contour of fish body back, least square algorithm, correlation coefficient

摘要： 针对淡水鱼种类的自动识别问题，利用机器视觉技术，提出一种基于鱼体背部轮廓相关系数算法的鱼体种类识别方法。首先根据采集的鲫、草鱼、鳊、鲤四种淡水鱼图片，将图像处理方法应用到鱼体背部轮廓的提取上，并采用最小二乘算法对鱼体背部轮廓进行曲线拟合，建立这四种淡水鱼的背部轮廓数学模型；接着对要识别的鱼体，通过机器视觉技术获得鱼体轮廓，并计算提取的鱼体轮廓与建立的四种鱼背部轮廓数学模型的相关系数值，达到对鱼体种类自动识别目地；最后采用提出的方法对市场随机选取的各60条活鱼进行了测试，测试结果表明，该算法简单，识别准确率较高，能够为淡水鱼种类识别方法提供新的思路，提高水产养殖的自动化水平。

关键词: 机器视觉, 背部轮廓, 最小二乘算法, 相关系数