Performance Evaluation of Point Feature Detectors for Eye-in-Hand Visual Servoing

This paper presents a new approach to evaluate the performances of point feature detectors for eye-in-hand visual servoing systems. The performances are analyzed in terms of stability and robustness criteria defined for a sequence of images. The first image represents the start position and the last one contains the reaching of the desired position for grasping the target object. The performances are back-analyzed, every image being compared with the last one which contains the desired features. Real-time experiments with a set-up consisting of a six d.o.f ABB robot with an eye-in-hand configuration were used to evaluate the performance. Point features are extracted with Harris and SIFT detectors and experiments show better results for the last descriptor.