|
In order to generate synthetic digital close range images for testing the accuracy and reliability of the advanced matching operators especially dedicated application Image Generator was developed. Image Generator can generate circular dark targets on bright background and the other way bright targets on dark background. Generating test digital images (*.bmp) with artificial targets in Image Generator requires defining the following parameters: image resolution, pixel size, target radius, distance between targets, brightness of the target and its background, background noise, target edge blur, sharpness level, calibration parameters of digital camera and interest area. Radiometric parameters of the image are being defined by R, G, B color components and additionally by Gaussian noise with any size and noise strength and two boxfilters (9 and 25 elements mask). Brightness change of targets requires gradient introduction. Geometric image distortions are being modeled by systematic errors by Brown’s and El-Hakim’s and Beyer’s model which includes symmetric radial distortion, asymmetric and tangential radial distortion affinity and a shear of the sensor matrix. The free oriented images are processed with 2D projective transformation. The application additionally allows for creation of interest area for Matching software, image fragment (crop) generation, listing with pixel coordinates and brightness level. The Image Generator application was developed in Delphi 7 programming language. Tests of digital images with artificial targets of various sizes were conducted by automatic coordinates measurement in authors application Matching (CWM – center weighted method and LSM – least square method were applied), as well as in commercial software Pictran DE (LSM method). Measured pixel coordinates of targets centers were compared with theoretical coordinates, which were calculated using Newton method. Pixel coordinates discrepancies on radiometric and geometric distorted images amounted to RMS .x'y' . 0.3.
|