A METHOD FOR AUTOMATIC RELOCATION OF SKIN MARKERS IN REARFOOT MOTION ANALYSIS

Abstract

INTRODUCTION: The authors are collaborating in a project aiming at developing computer based strategies to improve the analysis of the visual and ground reaction forces signal during passive or active foot load. The main objectives of the project are the automatic detection of the skin markers in image sequences for kinematics analysis, and the synchronisation the visual and ground reaction force data acquisition systems to achieve dynamic joint measurements. In this paper, a method for automatic relocation of skin markers in rearfoot motion analysis is proposed. The results of preliminary tests are critically described on full paper. METHODS: The visual signal is acquired using a high-speed digital camera (DALSA CA-D1), at 225 frames per second. A prototype system is being built, based on an IBM-PC running the Windows NT operating system. A windows interface has been developed, in order to facilitate data analysis. Data can be exported to a spreadsheet or to a statistical package for further processing. The experiments conduced focused on the observation of the ankle joint complex kinematics behaviour on the posterior aspect of the frontal plane during stance phase of walking barefoot. The skin markers have been clearly marked on the rear part of the leg and on the heel (black ink on a white stripe of tape). These markers are localised in the computer screen by hand, using the mouse, in the first frame of the sequence. The markers are automatically located in the other image frames of the sequence, using the proposed method, based on optimised block-matching. The space calibration procedure uses a scale marked by two targets located perpendicularly on the camera axis. These two points mark the horizontal. Comparison between the real measures of one well know rigid body and measures calculated trough the computer helps preventing some misbehaviour of the automatic analysis. RESULTS: The prototype system was applied to recorded data of angular motion of the ankle joint complex during two walking tasks. Our detailed motion analysis scheme during these motor tasks was successful in using the proposed automatic relocation method, with great economy to the operator. CONCLUSION: This automatic localisation solution represents a step towards the analysis of sequences acquired with high frame rates, justified by the fact that image acquisition with standard equipment (25 or 30 frames per second) presents strong aliasing problems. The method presented here is expected to be a useful tool for other kinematics behaviour studies.