渔业现代化

渔业现代化杂志

• 论文 • 上一篇    下一篇

一种边缘检测方法在头足类角质颚识别中的应用

  1. （1上海海洋大学信息学院，上海 201306；
    2上海海洋大学海洋科学学院，上海 201306；
    3大洋渔业资源可持续开发教育部重点实验室， 上海 201306；
    4 国家远洋渔业工程技术研究中心，上海 201306；
    5农业农村部大洋渔业开发重点实验室，上海201306；
    6农业农村部大洋渔业资源环境科学观测实验站，上海 201306）
  • 出版日期:2022-08-20 发布日期:2022-11-04
  • 通讯作者: 刘必林（1980—），男，博士，教授，研究方向：渔业资源生物学。E-mail: bl-liu@shou.edu.cn
  • 作者简介:王冰妍（2001—），女， 研究方向：计算机视觉与数字图像处理。E-mail:13965618443@163.com
  • 基金资助:
    国家重点研发计划（2019YFD0901404）；国家自然科学基金面上项目（NSFC41876141）；上海市高校特聘教授“东方学者”岗位计划项目（0810000243）；上海市科委地方高校能力建设项目（20050501800）；上海市科技创新行动计划（10DZ120750019DZ1207502）

The application of an edge detection algorithm in cephalopod beak recognition

  1. (1 College of Information Technology  Shanghai Ocean University, Shanghai 201306, China;
    2 College of Marine Sciences  Shanghai Ocean University, Shanghai 201306, China;
     3 The Key Laboratory of Sustainable Exploitation of Oceanic Fisheries Resources, Ministry of Education, Shanghai 201306, China;
    4 National Distant-water Fisheries Engineering Research Center，Shanghai 201306, China;
    5 Key Laboratory of Oceanic Fisheries Exploration, Ministry of Agriculture，Shanghai 201306, China;
    6 Scientific Observing and Experimental Station of Oceanic Fishery Resources, Ministry of Agriculture and Rural Affair, Shanghai 201306, China)
  • Online:2022-08-20 Published:2022-11-04

PDF

摘要： 针对采用传统的边缘检测算子Canny算子提取得到的边缘轮廓易产生冗余轮廓信息、边缘轮廓缺损的问题，而无法满足利用计算机视觉进行进一步研究的要求，提出了一种边缘检测方法，并应用于头足类角质颚的形态研究。该方法先后通过图像灰度化、滤波去噪、图像二值化、构造目标连通域、提取边缘轮廓5个步骤，实现边缘检测。结果显示：利用该方法进行边缘检测，能够有效区分信噪，提高了目标选择的准确性，同时又能够在误差允许的范围内保证轮廓的完整性，应用在头足类角质颚的识别中效果良好，像素准确率达到97.79%。研究表明，与 Canny算子的检测结果相比，本研究中的检测结果直观、完整、精确度更高，为头足类角质颚等生物形态特征的研究提供了新的思路。

关键词: 头足类, 角质颚, 边缘检测, 数字图像处理, 计算机视觉, 二值化, Canny算子

Abstract: Aiming at the problem that the edge contour extracted by the traditional edge detection operator,the Canny algorithm, is prone to produce such as redundant contour or edge contour defect, which can not meet the requirements of further research by computer vision, an edge detection algorithm is put forward and applied to the morphological study of cephalopod beak in this text. The algorithm realizes edge detection through five steps, including gray processing, filtering, binarization, constructing target connected domain and extracting edge contour. The results show that using this algorithm for edge detection can effectively distinguish signal and noise, improve the accuracy of target selection, and ensure the integrity of contour within the allowable range of error. It has a good effect in the recognition of cephalopod beak, and the pixel accuracy can reach 97.79%. Compared with the results of Canny algorithm, the results of the improved algorithm are intuitive, complete and more accurate, which provides a new idea for the study of biological morphological characteristics such as beaks of cephalopods.

Key words: Cephalopod, Beak, Edge detection, Digital image processing, Computer vision, Binarization, Canny algorithm