| The vision makes human to apperceive and understand the world.The goal of the research about robot visual perception is to make the computer or machines can "see" or "know" the world outside just like human beings.From the mid-1970s to the early 1980s, professor Marr proposed a theoretical framework for the field of computer vision-calculation vision theory.Its proposing was the important progress in the field of computer vision.In the recent 30 years,computer vision develops very quickly,and its application has been throughout the industry,agriculture,the military,national defense and other fields.However, in the research and practical applications,some questions of calculation vision theory are gradually revealed,and encountered some insurmountable difficulties.Intelligent system based on calculation vision theory frequently because of the change on partial nature of objects causes the wrong recognition result.Therefore, the basic starting point to solve the bottleneck in the development of computer vision is to understand the physiology of the human visual system deeply,and study the mathematical model and simulation technology of human visual system.Human's vision cannot do without Human's intelligence, people are able to see and understand the immediate object, largely due to the brain rather than eyes.The target which be researched by robot visual perception should be human's vision and brain, and applying the neural mechanisms of human visual perception system to the computer or robot,then we can creat intelligent machines which could "think" and "live" like human. The ultimate and most fundamental task of robot visual perception is to make computers or machines to be like a human being which could access and collect information outside automatic, identify the images, deal with all topologies, think, has self-awareness. As long as we can make the handling of information in the robot visual perception system and human vision system is a fundamental counterparts relations system, we can simulate a robot visual perception system which is equivalent to the human visual system, then the robot could process visual information, cognize and think.This paper is based on the basic physiological structure of human vision system, under the premise of study about the human visual system and working mechanism, visual perception and its neural mechanisms, the theory of human's thinking in image, topology property perception theory, this paper combines the human visual system with the topology and mathematical knowledge, and analyse the mathematical structure of vision, perception and thinking. From the perspective of topology this paper discusses detailed the topological space structure of the retinal visual information, the topological properties of visual information processing, the topological properties of thinking in image. The main points of this paper are as follows:Chapter 1: Introduction. The chapter describes the purpose and direction of robot visual perception, and expatiates computer vision, physiology, artificial intelligence, psychology and so on. The present chapter introduces the research content and results.Chapter 2: Theoretical foundation of this paper. The chapter introduces the theoretical foundation from four aspects, human visual system and working mechanism, perceived law and working mechanism of visual perception system, the theory of human's thinking in image, topology property perception theory. The chapter also introduces the basic idea of topology and topological space and homeomorphism, and it is the foundation for the arguments below.Chapter 3: The topological space structure of retinal visual information. The chapter in term of structure of the retinal cells , makes the information which is carried by the retinal as the object of study, and investigates the topological structure of the retinal visual information. The chapter begins with receptor cells visual information set R constituting topological space (R, ) as a example, and proves that the elements of denotes all the receptive field in receptor cell layer; proves that the working mechanism of receptor cell layer--the visual information combinations—is consistent with the topology principle; proves that (R, ), (B, ) and (G, )were local connected spaces ;proves that (R, |R), (B, |B)and (G, |G) are are the son space of retinal visual information topological space (S, ); proves that the subset race ={R,B,G} is a radix of(S, ); proves that the (S, ) is a local connected space and the (S, ) satisfies A2 space and A1 space.Chapter 4: The topological property of visual information processing. This chapter analyzes the path by which the human visual system processes the main character of physical image. This chapter proves that physical image information topological space (W, ) changes into receptor cell layer information topological space (WR, ) ,after the processing by retina (WR, ) changes into ganglion cell layer information topological space (WG, ) ,after transmitted by LGN (WG, ) changes into visual cortical areas V1 information topological space(VA, ),(VA, ) changes into visual cortical areas V2 information topological space(VB, );proves that these spaces are all discrete topological spaces,and discusses the physiological significance in detail; elucidates that people will notice the object which has closed contour; elucidates that visual perception system to the same type of things with a feature or repeated cognition can eventually form mental image; elucidates that all the information of menal image stores in distributed in many areas of different visual cortex;for the example of physical image, elucidates the process of visual system to shape features.Chapter 5: The topological property of mental image. This chapter analyzes the selection process and structure of mental image, and discusses the topological property from three aspects of character mental imagery, mental imagery rotation, mental imagery construction. This chapter proves that the mapping from outline information topological space (LK, ) to outline mental imagery topological space (L, ) is a quotient mapping; mental imagery rotation is a common form of thinking, this chapter proves that the outline mental imagery topological space (L, ) is the same to the rotation mental imagery topological space (M, ) in homeomorphism from the aspects of identical mapping and homeomorphism; proves some topological properties in the process of mental imagery structure for the examples of "synthesis structure" , "bonding structure", "analogy structure".Chapter 6: Experiment.This chapter through three experiment "the image of retinal cells ","the selection process of mental image", "recognition for graphics using mental image " and "the transformation of mental image" achieve the mental image simulation. This is the exploration to build the robot visual perception system which can produce mental image and use mental image. The various characteristics of external things save in the form of text files in a different coding to express long memory of mental image in the brain. The identification and transformation of the mental image can no longer rely on the operation of specific image pixels , rather, it relies on the comparison of document encoding. Even if the image does not exist, people can still activate mental image, and use mental image to do imagination and various transformations. This more in line accords with the work mechanisms of the biological visual.The purpose of this paper is to build a robot visual perception system which can simulate the human vision, and the foundation of this paper is the physiological structure of the human visual system, and the mathematical tool of the paper is topology.The paper hangs human visual system with topology together, and discusses topological property and mathematical structure of vision, perception and thinking deeply , then get some good propositions and conclusions. The Experiment builds the robot visual perception system which can produce mental image and use mental image. This paper provides some ideas and methods which can be used for reference to the producing of robot which could process the visual information, cognize, think, and have self-consciousness as the human.This paper has great guiding significance to the study of robot visual perception system. |
PDF Full Text Request