自动显微注射过程中图像分割的实现

Realization of image segmentation in automatic microinjection

自动显微细胞注射能克服人工注射时间长,易疲劳的缺点,为解决卵胞浆内自动单精子注射过程中细胞和针尖定位问题,通过基于YOLOv3神经网络和水平集相结合的方法,分割卵细胞和注射针,进而定位注射针的针尖。对比实验证明了方法能够解决传统水平集难以分割细胞和注射针的问题,完整分割出显微注射过程中发生形变的细胞,以及实现细胞内针尖的定位,且误差较其他算法小。

Automatic microcell injection can overcome the shortcomings of long manual injection and easy fatigue, in order to solve the problem of cell and pinpoint location in automatic intracytoplasmic sperm injection. Based on the method of combining YOLOv3 neural network and level set, the egg cell and the injection needle are divided, and then the needle tip of the injection needle is located. The comparative experiment proved that the method can solve the problem that the traditional level set is difficult to segment cells and injection needles, and completely segment the deformed cells during microinjection as well as locate the needle tip in the cells, and the error is smaller than other algorithms.