People gathering recognition based on dynamic texture detection

Gatherings of people always involve a common goal, and when the majority of people in a crowd are in an unstable mood, they will react with each other, and this could easily lead to unfortunate accidents or confrontations. As a result, violent attacks may take place and facilities may be destroyed. Such gatherings are often the beginning of security problems, about which the police are eager to obtain important relevant information. Many recent studies have focused on crowd scene analysis, while texture analysis has also been employed as a feasible method. Dynamic texturing involves patterns in a sequence of visual images with time-variation and repetition within a spatial domain, such as smoke, foliage, crowds or traffic conditions. Basically, dynamic texture lacks a uniform definition, yet exhibits a statistical regularity, with repetition of seemingly random movement patterns within a spatial domain, and exhibiting stability within a time domain. In a word, texture patterns always maintain similar characteristics in a sequence of visual images. This study uses a fixed camera to monitor a measured area, and proposes a grid model to describe crowd distribution within that area; the status value in the grid model shows various texture patterns, reflecting the crowd´s motion. While the activities of a crowd are often limited to a specific space, they are not stagnant at all, as some people may occasionally join or leave; the pattern in a sequence of images of a crowd shows an obvious dynamic texture distribution. Since the dynamic textures in temporal and spatial domains have some degree of similarity, the grey relational method in gray system theory is adopted to detect the similarity among a series of test images, and to identify the dynamic texture. Under these circumstances, when the number of congested cells in the grid model increases and a high degree of similarity exists in a series of images, it is possible to identify the occurrence of dynamic texture, and this serves as a basis for recognizing the `people gathering´ phenomenon. This method can be used in public places to provide useful data when crowds gather in order to maintain public safety.