Foundation Research Of Image Understanding Progress For Biology Characteristic-based Robot

Image understanding plays an important role in many fields such as defence, aerospace, industry, medicine and so on. The technology of image understanding for biology characteristic- based robot is one of the original technology that should be researched and developed prior in the program of the state's intermediate duration science and technology development. In this paper, based on the all-around using of certain knowledge of the subjects such as physiological psychology, brain science, perceive science, complex system methodology, modern mathematics, robotics and etc, the author explored and studied the image understanding process for biology characteristic-based robot primarily. The abstract of mathematic structures can reach is the deepen for people to perceive object from phenomena till essence. Hence, this paper is aimed to try to explore and find some mathematic structures about the work mechanism of creature visual perception system and reproduction the work mechanism in robot's visual perception system according as these mathematic structures.In this paper, based on the detailed awareness of physiology knowledge about visual perception data-driven course, the author seeked after and demonstrated some mathematic structures existing in the physiology mechanism. Some specialties are as follows:1. The mathematic structure of every neurone's receptive field is a category, and each object in this category is a zero object, and they are equal to each other. The receptive field categorys of neurons which have symmetrical receptive field structures are athwart categories to each other. This conclusion reflects the symmetrical relationship of the structure between the On and Off types of receptive field which have the same region of photo-receptor cells.2. The mathematic structure of the relationship among each horizontal cell and its conjoint photo-receptor cells is a non- bijection transformation group, and the relationship among each amacrine cell and its conjoint bipolar cells has the same mathematic structure. This conclusion opens out the mechanism of the retina lateral communication.3. The mathematic structure of the color code set in creature's color vision mechanism is a infinity Abelian group; The mathematic structure of the spectra corresponding color set is also a infinity Abelian group. They are isomorphism groups.Concept-driven is a process in visual perception course about the direction from the advanced conceptual work to the vision perceiver. That there is already a certain concept system in the brain of the robot is one of its basic qualification. In this paper, a simple structure model frame about concept system, which includes notion expression, notion relationship, matter notion, affair notion and etc, is present. The author also seeked after and demonstrated some mathematic structures existing in this system. Mostly are as follows:1. The mathematic structure of the concepts on one branch of the conceptional tree is a ace hemigroup pseudo-in order category; The mathematic structure of the whole conceptional tree is a ace hemigroup category; And the previous one is the subcategory of the post one. This conclusion reflects the arrangement relationship of the conceptional tree.2. All of the mathematic structures about"object image"and its corresponding"visual image"and"imago"are categories; The"visual image"category is the child of the"object image"category, and the"imago"category is the child of the"visual image"category, and what's more, there is a relation named covariant functor between every two categories. This consecution opens out the logic relation among different phases about visual information process result. It also opens out the essence mechanism of vision recognition. Hence, it can supervise the robot to set up the mental ability with"vision lasting".Based on correlation theories on human retina and the production of this paper, the author designed a primary robot retina. At the same time, according as the production and correlation theories about"spot cell"in perceive neurosciences, the author simulates the"spot"perceiver for the robot visual perception system, and presents the perceiver results. Thereby, this ascertained the visual acuity of the robot retina. This experiment incarnates the data-driven progress of the whole image understanding progress. What's more, it validates that the robot's visual perception system can also have the mental ability of perceiving a spot"object image"as a"spot"automatically.