Chromosome classification model based on grid reconstruction learning

The invention provides a chromosome classification model based on grid reconstruction learning aiming at the problems of difficulty in curved chromosome recognition, difficulty in generalization of chromosome data sets of chromosome fine granularity and different developing technologies and the like, and the accuracy of chromosome classification is effectively improved. According to the network, gridding and grid reconstruction module are specially designed, firstly, a chromosome image is gridded, and adverse effects caused by chromosome bending are weakened; and reconstruction features are screened by using the grid reconstruction module so as to improve the recognition performance of the curved chromosomes. The classification precision of the grid reconstruction learning model on three different common chromosome data sets reaches 0.973, 0.972 and 0.995. 本发明针对弯曲染色体识别困难、染色体细粒度以及不同显色技术的染色体数据集上泛化困难等问题提出基于网格重构学习的染色体分类模型，有效提高染色体分类的精度。该网络特别设计网格化及网格重构模块，首先将染色体图像网格化，弱化染色体弯曲造成的不利影响；随后利用网格重构模块，筛选重构特征，以提高弯曲染色体的识别性能。网格