Camera Calibration

Camera calibration is an essential step to obtain three-dimensional space informationform the scope of two-dimensional image information in computer vision, and it is widelyused in the field of three-dimensional reconstruction, medical imaging, navigation, visualsurveillance, intelligent robot, traffic inspection, testing and other areas. In traffic detection,geological disasters testing and other areas, it also has important applications. Depending onthe actual situation, the camera calibration has two types which are monocular and binocular.Based on the two types, camera self-calibration will be realized by this paper.For the one CCD camera calibration, image coordinate system, imaging plane coordinatesystem, CCD camera coordinate system and geodetic coordinate system are established firstly.And then the quantitative relationship, which containing parameters, of the image coordinatesystem and the geodetic coordinate system is built by the geometrical relationship. Taking theminimize error between the theory distance and the actual distance of the key point oncalibration template as target, the optimization model is established. The optimization modelis solved by the particle swarm optimization, the internal and external parameters of cameraare obtained, the corresponding relation of spatial coordinates and the pixel coordinates isdetermined, and then finished the calibration of the coordinate system. The experimentalresults show that the model is correct, calculation accuracy are improved significantlycompared with other algorithms.For the two CCD cameras calibration, two camera coordinate system (A, B), two imageplane coordinates and two geodetic coordinate systems are set. The quantitative relationshipbetween the image coordinate system and earth coordinate system of the A and B picture wasderived, and the calibration model is set up respectively through two different angles. The leftand right pictures are obtained by the two cameras (A, B). for the corresponding points in thetwo picture, it calculated its three-dimensional coordinates, considering they are the samepoint in space, so, the optimization model Ⅰ is set up with the target of minimizing thedistance in their spatial coordinates. The corresponding three-dimensional space coordinate ofthe key points in left image is mapped to the image coordinate system from the right camera, and the optimization model Ⅱ is established by the least square principle. Finally theoptimization model is solved by particle swarm algorithm and the parameters of the cameraare obtained. The theory is verified by the experiment, the advantages and disadvantages areanalyzed, and the merits of the two models are compared. Through the example, the twomodels are all feasible, simple, fast, and can be used in the calibration of two CCD cameras.And the model Ⅱ is with smaller error, more precise and more effective.In addition, the corner detection problem is studied. There are many corner detectionmethods, the most common is the Harris corner detection, which the effect of the arrangedpoint tilt detection is not ideal, so in order to improve the effect, this paper will enhance it,ensure the integrity of the camera calibration process and improve the calibration accuracyalso.