Research Of Camera Calibration Technologies And Its Implementation On The DSP Platform

Vision is an important means of observing and understanding the world. In recent years, visual measurement technology with its high speed, high precision, non-contact, high degree of automation has been widely applied. The optical imaging apparatus of the visual measurement are known as camera. Camera calibration is a prerequisite for vision measurement, and calibration accuracy directly affects the accuracy of measurement.This paper studies the camera imaging model, analyses advantages and disadvantages of the major existing calibration methods. Camera calibration methods can be divided into tradition calibration and self-calibration by whether need for calibration object. The ZHANG method is one of the most far-reaching method among all non-self-calibration methods. It is through establishing the homography between the model plane and its image to solve intrinsic and extrinsic parameters. The calibration methods based on active vision are an important class of self-calibration methods. These methods requires no calibration object and controls the movement of the camera to get multiple images to determine the intrinsic parameters, therefore, the calibration process is greatly simplified. The paper focuses on the ZHANG method and a method which circled around the lines paralleling to Y? axis, summarizes the implementation steps of this two methods. The later method was experimentally verified. The results show that this method just need to control the camera rotates twice, is easy to be achieved and good robustness.Unfortunately, because of harsh environment, camera calibration can not be implemented on a computer in some situations, such as outer space. Therefore, achieving camera calibration on a platform is particularly important. With dual-core, ARM and DSP, TI’s TMS320DM6446 is a processor having superior image processing performance. In this paper, we removes the correlation function of the OpenCV library, achieves camera calibration based on the ZHANG method on the hardware platform whose core is TMS320DM6446.Image acquisition is the basic of achieving camera calibration on dsp platform. In order to shorten the time of image acquisiton, a method that capture images from video stream is designed. OpenCV is a computer vision library including many image processing algorithm and the function of camera calibration in this library is based on the ZHANG method. This paper not only designs and removes the program of calibration based on OpenCV, but also improves the optimization algorithm of calibration function. Experiments show that the camera calibration procedure designed in this paper can be successfully achieved in dsp platform, and the running time is less than on the PC.