鱼体背部轮廓BPR算法的淡水鱼种类识别方法研究

RESEARCH ON IDENTIFICATION METHOD OF FRESHWATER FISH SPECIES USING BPR ALGORITHM BASED ON FISH BACK CONTOUR

针对淡水鱼种类的在线自动识别问题,利用机器视觉技术,提出一种鱼体背部轮廓弯曲潜能比率BPR(Bending Potential Ratio)的无损鱼体识别方法。首先通过分析采集的鲫、草鱼、鳊、鲤四种常见鱼类样本集,采用最小二乘算法建立相应的背部轮廓数学模型,并对鱼的BPR值分布区间作统计分析;接着对需要在线识别的鱼体通过摄像头获取相应彩色图像,并对获取的彩色图像进行相关预处理得到鱼体轮廓;最后计算鱼体图像的BPR值并根据建立的鱼体BPR分布模型识别出鱼的种类。实验结果表明,所提方法具有算法简单、识别准确率较高的特点,在采集的数据库上识别精度达到95%以上。

In view of the online automatic identification of freshwater fish species, we use machine vision technology and put forward a non-destructive fish body identification method using BPR （bending potential ratio）, which is based on the fish back contour. Firstly, through analyzing the collected sample sets of crucian, grass carp, bream and cyprinoid these four common fish species, we adopt least-square algorithm to establish the corresponding mathematical model of fish back contour, and give the statistical analysis of BPR value distribution interval. Then we use camera to obtain the corresponding colour image of fish body, which needs online automatic identification, and we can get the fish back contour through obtained colour image after pretreatment. Finally, we calculate the BPR values of fish body and use the established BPR value distribution model to identify fish species. Experimental results indicate that the method is simple and has high identification accuracy rate, which is above 95% on the collected database.