An investigation of color based marker recognition for 3D instrument localization in surgical applications

Color based object recognition offers an attractive option in the field of machine vision in which using color, objects which have the same morphological features will become distinctive. Inspired by this idea, a research in color markers for instrument localization is proposed. This intention originates after the development of modiCAS^® (Modular Interactive Computer-Assisted Surgery) which provides a cooperative solution for surgical operation by researchers in Center for Sensor Systems, University of Siegen. Besides being expensive, the current solution for instrument localization relies on separate computation for recognition of reference bodies´ geometry since their morphological features are the same. Hence this paper describes an investigation for color based marker recognition under extreme illumination of surgical environment. The challenges include extreme illumination intensity which is more than 40 times daylight intensity, small marker sizes, speed of the image processing and color stability of the markers under strong shadows. Previous works in color constancy field of research is analyzed to find suitable algorithm for real-time application. A concise overview of color based recognition strategy is presented, results of experiments under mimicry of surgical environment are analyzed and this paper will finally conclude the suitability of this method for instrument localization in surgical environment.