Auto Calibration System Of Camera With Radial Distortion

We can use rectangular coordinate (x , y) or Polar Coordinate(r , ?)to indicate a point in a image. The Distortion of a point in a image is dx = x '- x, dy = y' - y or dr = r' - r, d? = ?'- ?. When dr changes , camera radial distortion will happen. The reasons for camera radial distortion include Flawed Imaging Sensor, uncorrected assembly of important parts and Flawed lens,etc. The distortion of camera makes images look like fish-eye distortion and unbeautiful.Camera distortion is unpermitted in Vision Measurements, 3D reconstruction, image recognition and etc. Such as Robot Visual Orientation and 3D Scanner, all of Which restore the 3D coordinate from the 2D image pixel. So, we must assure the accuracy of the sampled 2D image. In two-dimensional code identification field, when the image distortion exists, the results may be wrong.So, it is necessary to calibrate the Camera to make sure that the 2D coordinate mach the 3D world coordinate accurately. Currently there are several ideal means both at home and abroad, such as DLT, RAC, Plane calibration from Zhang Zhengyou, Plane Circle calibration from Meng Hu, parallel Circle calibration from Wu YiHong and etc.Base on the above calibration theory, this paper presents a mean of auto calibration of camera with radial distortion and validates the real-timing of the mean through ARM Embedded System. The first step is to pre-process the BMP image, which includes Converting the BMP image to gray scale image with median filtering, equalizing the image by adjusting the expected means and variance, splitting the image into foreground and background, binarizing the image bases on the splitting image, and thinning the binarize image by using image morphology. Base on the thinning image, this paper creatively present tracking curve with template, finding distortion center with variance principle and finding crossing point with Harris detector. Finally use RAC theory to calibrate the camera.