Accuracy evaluation method of multispectral data fusion for robotic systems

Abstract

The aim of the paper is to describe method for accuracy evaluation of optical multispectral data-fusion systems. The method was primarily developed for evaluation of robotics systems formed in our laboratory, but its usage is without doubt much wider. Described method of evaluation consists of multispectral target design, pattern extraction and computation of criteria for data fusion evaluation. Two completely different robotic systems have been evaluated in this paper. The first one is mobile robot Orpheus-X4 with its advanced visual system that uses data fusion for automatic environment mapping and visual telepresence. The fusion is realized by means of spatial data from Swissranger TOF camera; the thermal and CCD camera data are comprised in one multispectral 3D model for mapping purposes or stereo image presented in real-time to a monitor or a binocular head-mounted display. The second system is scanning robotic system RoScan with industrial manipulator that can be used for many different applications in medical domain, such as “diabetic foot” necrotic tissue detection, inflammation diagnosis, or monitoring of tissue recovery process. Fusion is realized by means of spatial data from triangulation line scanner; the thermal and CCD camera data are comprised in one multispectral 3D model for medical purposes.