Training-Free Artifact Detection Method for Radio Tomographic Imaging Based Device-Free Localization

Abstract

Radio tomographic imaging (RTI) is a promising device-free localization (DFL) method by reconstructing an attenuation image from the variations of received signal strength (RSS) measurements. Targets are considered to be at positions of some pixels with local maximums in the attenuation image. However, the number of local maximums generally exceeds the number of real targets, which will lead to large localization errors. The local maximum which does not have its corresponding real target is defined as an “artifact”. Through our analysis, two causes for artifacts are concluded: 1) multi-path interference: RSS measurements for links that are far from the targets may have variations due to multi-path interference, even though these links are not blocked by the targets. 2) intersecting affected links: the affected links blocked by different targets may intersect within the monitoring area, fabricating artifacts. At a position of artifact, there seems to be a target whose interference effect on links is similar to that of a real target. The quality of the attenuation image will deteriorate because of the above two inevitable influencing factors. To alleviate the effect of artifacts, a training-free artifact detection method is proposed. We formulate the artifact detection problem in the attenuation image as a multi-component optimization problem combing the information of feature links and the density of intersection points. The alternating direction method of multipliers (ADMM) is used to derive a closed-form solution. Experiments are conducted to validate the advantages of the proposed artifact detection method.