Direction Relation Matrix Model Considering the Location Characteristics

Abstract

Abstract To resolve the problem that the conventional direction relation matrix (DRM) models are not sensitive to the location characteristics of the target object, a DRM model that takes into account the location characteristics is proposed. Based on spatial cognition, the concept of taking the direction field of the target into consideration is provided, and a probability density function of the direction field is established. The cumulative probability of the region where the target object is located is used to replace the area in the conventional DRM model, and two methods are given for cumulative probability calculation. The cumulative probability is used to assign values to the direction relationship matrix. Experimental results show that, compared with the conventional DRM model, when the target is present in multiple regions, the method proposed in this paper leads to results that are more consistent with human intuitive experience and describes the direction accurately. When the target is only present in a certain fixed region, the proposed method has the same performance as the conventional DRM model. Therefore, the method proposed in this paper alleviates to a certain extent the problem that the conventional DRM models are not sensitive to location characteristics.