Abstract:

Koi is graceful and colorful.In the process of breeding, monitoring the growth state and predicting the yield of koi are both important.Almost every koi will surface on the water when feeding.At this time,the digital camera captures the feeding state image rapidly.The head and mouth of every koi show the semi-spindle shape and bright color significantly contrast with the water background in the image.Based on the method of OpenCV image vision,feature triangle was defined based on the extracted colorful semi-spindle shape automatically.The number and projection area of these triangles in the image were computed through the vector graphics calculation by computer.A great deal of measurement data was taken for analysis on the correlation between the shape parameters and the body mass of koi by MATLAB.A correlation analysis model was established between the projection area of the feature triangle and the body mass.Through this method, the koi yield including the number and average body mass in the aquaculture pond was estimated automatically.The result showed that,the estimating error of the number was lower than 14%,and the estimating error of the average body mass was lower than 12%.The method

could satisfy the needs for the daily koi culture.

Key words: Koi, Image recognition, Feeding state, Extracting automatically, Semi-spindle shape;Feature triangle, Correlation analysis model, Yield estimation

摘要： 锦鲤体态优美，色彩斑斓，在其苗种池塘养殖过程中需要长期监测其生长状态，并分阶段估算其产量。针对锦鲤在饲料投喂时间集中浮出水面摄食的行为特点，采用基于OpenCV的图像视觉方法，自动识别锦鲤摄食状态图像中锦鲤头口部位半纺锤形特征及眼球特征，定义特征三角形，统计特征三角形数量，并计算其平均投影面积数据。同时，利用MATLAB对采样数据进行分析，建立锦鲤头口部特征三角形投影面积与体质量的相关性分析模型，实现池塘养殖锦鲤产量（出塘尾数及平均体质量）的估算。结果表明，该方法对锦鲤出塘尾数的估算误差小于14%，对出塘平均体质量规格的估算误差小于12%，可以满足日常锦鲤养殖的应用需求。

关键词: 锦鲤, 图像识别, 摄食状态, 自动提取, 半纺锤形, 特征三角形, 相关性分析模型, 产量估算