High-Precision Photogrammetry Using RGB Color Information

Abstract

This paper discusses the possibilities of using all three channels of a true color image for high-precision photogrammetric 3-D point measurement. After a survey of common types of digital color sensors, the wavelengthdependent imaging errors of commonly used lenses are discussed, as well as how the different options for correcting these errors are derived. The resulting effect on image and object accuracy is verified by bundle adjustments with selfcalibration. For standard imaging configurations in closerange photogrammetry, a significantly enhanced precision (a factor of 1.7) can be proven for different camera and lens combinations. It is shown that the mathematical model matches the physical imaging properties of the lens down to a circular aperture (ca) of 1/50th of a pixel. However, a significant improvement of standardized length measuring errors could not yet be achieved. In industrial photogrammetric 3-D metrology, high-resolution SLR cameras are usually used for offline applications. Most photogrammetric image analysis programs are unable to process true color pictures, rather, they are restricted to working with only one color channel (usually green) or with converted gray level (intensity) images. Hence, loss of geometric resolution occurs as a function of the color sensor and the post-processing of the image data, either within the camera firmware or offline, leading to reduced image and object point accuracy. In addition, wavelength-dependent imaging errors within