基于计算机视觉的水下海参识别方法研究  

Research on underwater sea cucumber identification based on computer vision

作　　者：李娟[1] 朱学岩 葛凤丽[3] 白皓然[1] Li Juan;Zhu Xueyan;Ge Fengli;Bai Haoran(School of Mechanical and Electrical Engineering,Qingdao Agricultural University,Qingdao,266109,China;School of Technology,Beijing Forestry University,Beijing,100083,China;School of Management,Qingdao Agricultural University,Qingdao,266109,China)

机构地区：[1]青岛农业大学机电工程学院,山东青岛266109 [2]北京林业大学工学院,北京市100083 [3]青岛农业大学管理学院,山东青岛266109

出　　处：《中国农机化学报》2020年第7期171-177,共7页Journal of Chinese Agricultural Mechanization

基　　金：国家自然科学基金项目(41674037、61374126);山东省研究生教育质量提升计划项目(SDYY18168)。

摘　　要：海参识别是海参捕捞自动化和智能化的前提和关键。海参生活环境复杂且海参体色随环境变化,为解决复杂多变环境下海参识别的难题,提出基于视觉图像和海参特性的水下海参识别算法。利用海参体色随环境变化而海参刺颜色稳定呈黄绿色的特点,在图像预处理的基础上设计图像融合算法突出海参刺,然后通过边缘检测获取海参刺轮廓并计算轮廓质心坐标,进而通过椭圆拟合实现海参的识别。不同于已有的利用海参主干进行海参识别的方法,提出的基于直接最小二乘的海参刺质心椭圆拟合识别算法以海参刺的分割和椭圆拟合为途径,解决海参体色随环境变化而引起的识别率降低的问题。试验结果表明,该算法对自然环境下的海参识别平均准确率为93.33%,具有一定的实用价值。Sea cucumber identification is the premise and crux to the automation and intelligentization of sea cucumber fishing.The living environment of sea cucumber is complex,and the body color of sea cucumber changes with the environment.In order to solve the problem of sea cucumber identification in complex and changeable environment,it was proposed an underwater sea cucumber identification algorithm based on visual image and the characteristics of sea cucumber.The color of sea cucumber’s thorn is stable yellow-green color with the change of environment while the body color changes with the environment.On the basis of the image preprocessing,the image fusion algorithm was designed to highlight the sea cucumber thorn,then the sea cucumber thorn contour was obtained by the edge detection and the centroid coordinates of the contour were calculated,and then the recognition of sea cucumbers through ellipse fitting.Different from the existing methods of identifying sea cucumber by using the sea cucumber’s truck,it was proposed an algorithm of identifying sea cucumbers by the ellipse fitting of sea cucumber thorn centroid taking sea cucumber thorn segmentation and ellipse fitting as the approach,which solved the decrease problem of the recognition rate of sea cucumber due to the body color changing with the change of environment.The experimental results showed that the algorithm had an average accuracy rate of 93.33%for sea cucumber recognition in natural environment,which has certain practical value.

关 键 词：海参捕捞 图像识别 图像融合 边缘检测 椭圆拟合 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]