基于机器视觉技术的对虾主骨架线提取
Distilling the main skeleton line of shrimp based on machine vision technology
-
摘要: 长度作为对虾外观的主要特征之一,是对其规格进行判别的主要依据,而对虾的主骨架线可以直接反映长度。骨架化在图像上又称为细化,因此本文以一种有效的二值图像细化算法为基础,结合实验对分支长度进行分析,设定初始长度阈值,并在每次去除一条分支后对阈值进行自动修正,提出了一种能够有效去除多余分支并且不影响主骨架线长度,完整地提取对虾主骨架线的算法。结果表明,根据不同状态主骨架得到的预测长度与实际长度的最大相关系数为0.946,最小相关系数为0.747。对预测样本的长度预测平均相对误差为2.06%。实验结果表明,该算法对于对虾的主骨架线提取具有实用性,提取出的主骨架线具有较好的连续性、光滑性,还在去除骨架分支的同时基本保证了单像素的宽度,更加准确和稳定地反映主骨架线的几何性状。Abstract: The length is one of main features of shrimps, which can be seen from the main skeleton line of the shrimp. So, it is very important to analyze the main skeleton line. The length of branches was studied based on a classical thinning method in this research. And the initial length was set as the threshold. The threshold should be corrected automatically after one branch was removed each time. This algorithm could remove branches of the skeleton efficiently and didn’t have an effect on the length of the main skeleton line. Actually, the length of skeleton was highly correlated with the real length, the maximum R2 was 0.946 and the minimum R2 was 0.747. The real length of another sample set was predicted with this method , and the mean relative error was 2.06%. Results demonstrated that this algorithm had prevalent applicability to distill the main skeleton line from the skeleton of the shrimp no matter what posture the shrimp’s body was. The main skeleton line produced by this algorithm not only had great smoothness and connectedness, but also kept one-pixel width. The geometric properties of the main skeleton line could be shown precisely.