طراحي و بازشناسي خودكار تارگت هاي كددار در فتوگرامتري بردكوتاه

Design and Automatic Recognition of the Coded Targets in Vision Metrology Systems

Today, vision metrology is one of the known precise and automatic 3D measurements methods in industry. In vision metrology stystems, coded targets are utilized to automatically derive the initial value of unknowns. Therefore, automatic coded target recognition is the precondition for automation of vision metrology systems. Nowadays, it is also used in other applications such as automatic registeration of multi station laser scanner data, automatic calibration of mobile mapping systems, and automatic generation of panoramic images by linear array sensors. In this research project, different methods of coded target design are reviewed first. Then a special design of coded target called, spatial orthogonal coding, is presented. The proposed code design is highly stable to code recognition, geometric (scale, rotation and projectivity) and radiometric (focus and contrast) distortions, and has small size, low complexity and high number of possible coded targets (256). After code designing, we propose a deterministic method for automatic coded target recognition in two steps: recognition of target location in the image and code extraction. Target location is recognized based on white target frame and target code is determined by sampling from coding cells after local projective distortion correction. The final section of the research is accuracy and robustness evaluation of the proposed coded target design and recognition method. The experiments on more than 180 images of 12 different retro-reflective coded targets showed that the proposed method is generally valid and has high robustness to geometric and radiometric distortions. The method was also checked by hidden areas and it was seen that it is highly sensitive to fully detectable target frame. Therefore, we suggest to study on a new coded target design and recognition method that is able to handle the free-frame coded targets.