Object Detection and Spatial Location Method for Monocular Camera Based on 3D Virtual Geographical Scene

With the rapid development of video surveillance and 3D geographic information system (GIS), object detection and spatial location have gradually become one of the supporting capabilities of various applications. The primary application mode of traditional surveillance video and 3D geographic information fusion is video projection to the 3D geographic scene to form the 3D virtual geographic scene. Research on object detection and spatial location method based on the 3D virtual geographic scene can significantly improve intelligence and spatialization application ability. This paper proposes a method of object detection and spatial positioning of monocular camera based on 3D virtual geographical scene and realizes monocular camera object detection and geographic coordinate calculation. The method is composed of three algorithms, (1) fusion constructing algorithm of the 3D virtual geographic scene, (2) object detection algorithm based on convolutional neural network, (3) coordinate calculation algorithm based on the 3D virtual geographic scene. After the location error and application experiment verification, the positioning accuracy error of this method is positively correlated with the viewing distance, and the object detection and spatial location are realized. The method is propitious to various 3D geographic scenes, which can effectively improve the spatial understanding of the object in the surveillance. It has application value to all kinds of 3D GIS.