Application Of Object Detection And Location Method In Construction Land Supervision

As one of the most important natural resources,land resource is the foundation of national economic construction.With the sustained and rapid development of China's economy,the contradiction between supply and demand of land resource has become increasingly prominent.As the core element of land resources,the development and utilization of construction land is the focus of the contradiction between supply and demand.Therefore,it is particularly important to strengthen the supervision of construction land.The traditional supervision mode of construction land is mostly post supervision,which is difficult to realize pre-warning.In view of this kind of management problem,many places carry out supervision on construction land of tower foundation monitoring video,but the existing supervision on construction land of tower foundation monitoring video is mostly manual duty mode,with low level of intelligence and unfavorable to the secondary interpretation of construction behavior.The main reasons are the imperfect system of construction behavior category and the lack of linkage between geospatial and video images.How to improve the intelligent level of construction land supervision and improve the interpretation efficiency of tower foundation monitoring video is the difficulty of the current construction land supervision of tower foundation monitoring video.Aiming at the difficult problems in the supervision of construction land in tower based surveillance video,this paper proposes a target detection and spatial positioning algorithm based on 3D Virtual Geographic scene.The research contents of this paper are as follows:(1)3D Virtual Geographic scene fusion construction algorithm.Three dimensional virtual geographic scene is the basis of target spatial location under the condition of monocular surveillance camera.Firstly,based on the inside and outside orientation elements of the tower base surveillance camera,a virtual camera is generated in the 3D Geographic scene;Secondly,the tower monitoring image is projected into the 3D Geographic scene to form a 3D Virtual Geographic scene.(2)Construction behavior target detection algorithm.Firstly,the characteristics of various construction land scenarios are analyzed,and the classification system and sample library of construction behavior target objects are formed.Secondly,fast r-cnn is used as the basic model to train the construction behavior target detection model,and the construction behavior target detection algorithm is constructed on this basis;Third,the integration of target detection model is carried out,and the 3D Virtual Geographic scene is taken as the detection object to detect the construction behavior in the screen.(3)The algorithm of geographic coordinates based on 3D Virtual Geographic scene.Based on the three-dimensional virtual geographic scene,the pixel coordinates of the corner points of the construction behavior are converted into geospatial coordinates by the coordinate transformation method,so as to realize the spatial positioning of the construction behavior,and evaluate the relationship between the positioning accuracy loss and the sight distance.(4)Architecture and function design of intelligent supervision system for construction land.Taking tower monitoring video as the data collection source,based on the target detection and positioning method studied in this paper,an intelligent supervision system of construction land is designed to realize the intelligent monitoring and positioning of construction behavior,and on this basis,the construction behavior and the spatial control data of construction land are spatially overlapped to assist manual rapid determination of the legitimacy of construction behavior,And form a comprehensive situation of spatial distribution of construction behavior.This paper takes Nanjing City as the experimental area to verify the effectiveness of the algorithm and system.Firstly,the three-dimensional geographic scene of the experimental area is constructed with the combination of DEM + DOM + DLG;Secondly,through making training samples of 37688 tower base monitoring images,the target detection model which can detect 15 kinds of construction behavior is obtained,and the model m AP accuracy is 83.44%;Thirdly,the location experiment shows that the spatial location error of the algorithm based on 3D Virtual Geographic scene increases with the increase of line of sight,which is caused by the perspective change of pixel spatial resolution;Fourth,based on Super Map GIS 10 I platform,using B/S architecture,build the construction land intelligent supervision prototype system,and access RTSP,RTMP video stream,realize the intelligent monitoring and positioning of construction behavior,which can effectively improve the level of intelligent supervision of construction land.