Research On Camera Self-calibration Method For Building Reconstruction

Building reconstruction is the key technology of digital city and even the wisdomurban construction, while camera calibration facing buildings scene plays a necessarybase role in building’s3D reconstruction. Further more, camera calibration can alsobe widely used in robot navigation, video monitoring and scene roaming etc.Therefore, camera calibration is a very important research subject in the field ofcomputer vision research.Camera calibration is the process to calculate the camera’s parameters. Camera’sparameters mainly include the intrinsic parameters (focal length, principal point’sposition and pixel aspect ratios, etc) and the extrinsic parameters (camera’s rotationmatrix and translation vector in different views). The methods of camera calibrationhave roughly three categories: the traditional calibration approach, calibration methodbased on active visual and self-calibration method. The traditional calibrationapproach need a calibration pattern, it’s not very suitable for complex calibrationoccasions of outdoor building scene; Calibration method based on active vision needto strictly controlled movement of the camera platform, so it is not flexible enough inthe actual operations; While calibration method is more convenient and flexible. Inaddition, since the universality and portability of digital cameras, digital cameracalibration increasingly become the research focus, so this paper mainly studies thedigital camera calibration method in building scene.This paper firstly studied the vanishing points detection related to cameracalibration. In view of the shortcomings of the existing detection algorithms, thispaper first make full use of line information, demonstrates the linear distributionrelationship in parameters space of the segments forming a vanishing point, and thenputs forward a vanishing point detection with robust regression method based on thislinear distribution relationship. All kinds of experiments show that the vanishing pointdetection algorithm can get a higher coordinates accuracy. And, the vanishing pointobtained can lay foundation for the camera calibration.On this basic, this paper puts forward a kind of linear calibration approach withvanishing points; solve both the intrinsic parameters and the extrinsic parameters. This method needs not a calibration pattern or a specific camera platform, while theexperimental operation is simple and flexible. In the calculations, this method canavoid the basis matrix’s complex calculation process, and the complexity of thecalculation is lower; In the scope of application, this approach ensure the cameracalibration’s precision, and meanwhile, solve the calibration problem with fewernumber of vanishing points obtained, so the applied range is broadened.To prove the effective and feasible of the camera linear calibration algorithm,this paper carries out a series of experiments in the end. On the one hand, take thesimulation image from our camera simulation platform to test the calibration resultscalculated by related algorithms. On the other hand, this paper also use real image toanalysis the error between original image and the afresh projection image. All theexperimental analysis and comparison proved that the proposed camera calibrationalgorithm is effective and feasible, and has good application value.