基于U型神经网络两违疑似图斑识别模型研究

Research on two violation suspected spots recognition model based on U-shaped neural network

目前,违章建筑和违章占地现象时有发生,两违执法工作已是城市管理部门的常态。基于U型神经网络(U-Net)建立了两违疑似图斑识别模型,在GIS技术支持下,可对大面积正射影像进行自动搜索,标定两违疑似图斑的位置和范围,从而减轻人工作业劳动强度,提高了两违监测的自动化水平和工作效率。

At present,illegal buildings and illegal land occupation occur from time to time,and the law enforcement of two violations has become the norm of urban management departments.In this paper,a recognition model of suspected spots of two violations is established based on U-net.With the support of GIS technology,it can automatically search large-area orthophoto images and calibrate the location and range of suspected spots of two violations,so as to reduce the labor intensity of manual operation and improve the automation level and work efficiency of two violations monitoring.