王 斌1，徐建瑜1，王春琳2
（1 宁波大学信息科学与工程学院，浙江 宁波 315211；
2 宁波大学海洋学院，浙江 宁波 315211）

摘要：为研究不同背景（白色无沙、蓝色无沙、蓝色有沙）对三疣梭子蟹蜕壳持续时长与蜕壳周期的影响，搭建了3组24 h连续拍摄梭子蟹蜕壳的循环水养殖实验环境，从稚蟹5期开始拍摄不同背景下养殖的梭子蟹30 d，并设计了一种基于计算机视觉技术的梭子蟹蜕壳自动检测方案。该方案运用了图像处理中数学形态学方法成功实现了无沙背景的蜕壳自动检测，检测准确率达到97.9%，同时训练Adaboost分类器自动检测有沙背景的蜕壳行为，准确率达到79.5%。结果显示，蓝色有沙背景组的梭子蟹蜕壳持续时长缩短了15.3%，蜕壳周期缩了12.0%，梭子蟹蜕壳时间主要集中在3:00—6:00及18:00—21:00两个时段。研究表明，三疣梭子蟹更适合在有沙背景中蜕壳。

关键词：三疣梭子蟹；背景选择；蜕壳；计算机视觉；自动检测

Computer-vision based molting detection of Portunus tritubercularus and effects of different backgrounds on molting
WANG Bin1，XU Jianyu1，WANG Chunlin2
(1 Faculty of Information Science and Technology, Ningbo University, Ningbo 315211, China ;
2 School of Marine Science, Ningbo University, Ningbo 315211, China)

Abstract: To observe the impacts of different backgrounds (white without sand, blue without sand, and blue with sand) on molting duration and molting cycle of Portunus tritubercularus, 3 groups of experimental aquaculture environment were built up with circulating water and 24-hour continuous monitoring. All crabs in the three different backgrounds were observed for 30 days since they grew into fifth-period juveniles, and a detection scheme for crab molting based on computer vision was designed, which, by using the mathematical morphology method of image processing, successfully realized the automatic detection of molting in the background without sand with a precision rate of 97.9%; meanwhile, molting of the with-sand group was detected using Adaboost classifier and the precision rate was 79.5%. The data showed that molting duration was reduced by 15.3% and molting cycle was shortened by 12.0% in the blue with sand group. Besides, molting activity occurred at two peak times respectively between 3-6 am and 18-21 pm. Therefore, Portunus tritubercularus prefers molting in sand background.

key words: Portunus tritubercularus; background selection; molting; computer vision; detection