Spatial Occlusion Model Based On Fixed Viewpoint

Spatiotemporal reasoning is the reasoning of the object that take the space and change with time .In artificial intelligence(AI)field, The enthusiasm of time and space concepts and reasoning methods from the perspective of epistemology led to the birth and development of the space-time reasoning. The method of spatiotemporal reasoning improved with the development of AI. Particularly in Computer Vision (CV) areas, the overlapping of subject grows more obvious for description of entity. In recent years, the field of artificial intelligence is a very active branch.The occlusion of space is an important research topic in spatial representation and reasoning. In video and image processing, visual space occlusion is the most common phenomenon. Most studies of the phenomenon of space occlusion must consider time factors. Therefore, the research of space occlusion put forward new demands of the integration of time and space information. The representation and reasoning of space occlusion has very broad application prospects in Computer Vision, Image Processing, Geographic Information Systems, Robot Navigation and many other fieldsThe model of space occlusion representation and reasoning are mostly based on Regional Connection Calculus (RCC). Given a new definition of the original language in a new model, one can use the original language to expand the relationship of the RCC to enrich the ability of the new model. Although the new model increasing the ability of representation, one can not according to actual needs for expansion. At the same time, increasing a new original language means that the model will bring an additional burden. This paper proposed a new model to address the problem. The new model composed of two parts, Images occlusion model and entity occlusion model. Images occlusion model give 14 image relations which are jointly exhaustive and pair-wise disjoint, Then we give a composite table of the 14 image relations and the concept neighborhood graph. Table inference algorithms can be used to inference complex image relationship automatically. Entity occlusion model gives a uniform definition of the entities occlusion and reasoning method of entities. In this way, one will be able to make use of the theoretical result of the RCC which is a mature theory, And the new model would not makes complexity of reasoning grow more difficulty.RCC is a famous model in Qualitative Spatial Reasoning(QSR). It use region as the basic spatial representation. It define the basic topologic relation C(x, y), which read as"x connects with y". It meansthat the topological closures of regions x and y share a common point. The relation C(x, y) is reflexive and symmetric. Using C(x, y), a basic set of dyadic relations are defined: DC(x, y) indicate x is disconnected from y; EC(x, y) indicate x is externally connected with y; EQ(x, y) indicate x is identical with y; PO(x, y)indicate x is externally connected with y; TPP(x, y) indicate x is a tangential proper part of y; NTPP(x, y) indicate x is a non-tangential proper part of y; TPPI(x, y) indicate y is a tangential proper part of x; NTPPI(x, y) indicate y is a non-tangential proper part of x.RCC model which using region as space element accord with the feeling of space. For decades, scholars have very in-depth study of the theory of RCC, made many important theoretical results. However, the RCC model has its inherent flaws in the area of occlusion. How to expand RCC, so as to be able to carry out the representation and reasoning of space occlusion becomes a hot issue. Galton studied the nature of the cause and aimed at the convex hull entities to distinguish between the 14 types of occlusion. Note nature of the occlusion associated not only with the location of the space objects, and the Viewpoint.To achieve the representation and reasoning of real relationship between the entities overhead and reasoning, which come from video or image data, The new model will be divided into two parts, the first part is image occlusion model, in order to RCC theoretical conclusions in the new model would continue to apply, we have to re-interpret the region, image, entity. Given function and predicate, linked. The image occlusion model reuse the representation method of image which given by Galton. We introduced the affect relations to expand the relation of RCC, it made up of 14 relations: D(A, B) indicate A is disconnect of B;JC(A, B) indicate A is just disconnect of B;PH (A, B) indicate A partially hides B; PHI(A, B) indicate A is partially hidden by B;JH(A, B) indicate A just hides B; JHI(A, B) indicate A is just hidden by B; H(A, B) indicate A hides B; HI(A, B) indicate A is hidden by B; EH(A, B) indicate A exactly hides B; EHI(A, B) indicate A is exactly hidden by B; F(A, B) indicate A is in front of B; FI(A, B) indicate A has B in front of it; JF(A, B) indicate A is just in front of B; JFI(A, B) indicate A has B just in front of it.The images occlusion model gives the definition of the 14 image relations. And the axiom system is given to reason the relationship of image. One can infer the relationship of image using the relationship of RCC and a relationship between the image and entity. The composite table of image is given, Composite table reasoning method can be used to reason complex relationship of image.The second part of the new model is entity occlusion model. Because the image occlusion model only study the relationship ofimage, but not the real occlusion relationship of entity, particularly for the non-convex hull entity, its image is always composed of a number of images, in which the occlusion are already exist between each others, This has brought about many new problems in reasoning with the entity. Based on the image occlusion model, we defined the non-convex hull of the entities mapping methods. Using mapping method, correspondence between entities, we define the occlusion relationship of entities. In this way, the entity occlusion model not only used to describe the relationship between the convex hull entities and convex hull entities, but also used to describe the relationship between the convex hull entities and the non-convex hull entities or the relationship between the non-convex hull entities and the non-convex hull entities. Finally, we give the reasoning method of the occlusion relationship of entities and some examples for verification.We consider the viewpoint exists in the new model, but does not take into account the changes of viewpoint. If the change of viewpoint has taken into account, we need to know the quantitative information of viewpoint. But only received an absolute coordinates of the location of viewpoint, without the target location coordinates of the object, the change of location has no contrast. If the location is relative coordinates, the movement can be converted into perspective all the observed changes in the location of entities. Therefore, in order to simplify the model, we do not consider the changes of viewpoint.This paper presents the model to some extent shielded room to improve and add space overhead theory. In the fields of computer vision and image processing have potential applications.