The Principles of Organizing the Search for an Object in an Image, Tracking an Object and the Selection of Informative F
A significant expansion of the scope of computer vision, in particular in real-time systems, places very high demands on them in terms of productivity and efficiency of information processing, and in feedback systems, it also requires information lag in i
- PDF / 1,379,197 Bytes
- 23 Pages / 439.37 x 666.142 pts Page_size
- 67 Downloads / 190 Views
Abstract. A significant expansion of the scope of computer vision, in particular in real-time systems, places very high demands on them in terms of productivity and efficiency of information processing, and in feedback systems, it also requires information lag in it. Such requirements are not ensured by traditional approaches. The way out of the situation may be to use as a prototype the principles of organization of the human visual system, which has a very high selectivity of perception of video information. The paper presents a generalized dynamic model of the organization of these principles. It is proposed to use them to organize the search for an object in a coarse image of a scene, to track an object and, if necessary, to carry out its classification or recognition at a more detailed level. Keywords: Retinal neural network · Receptive fields · Image preprocessing · Object search · Informative signs · Local (ring) organization of neurons · Adaptation mechanisms · Intelligent video systems
1
Introduction
Intelligent video cameras and real-time video systems play a large role in automation systems for production processes, visual quality control of products, robotics, security and military systems, automation systems for scientific and biomedical research, etc. Moreover, the range of their application, the requirements for them are constantly expanding. This is especially true for video systems with feedback, where the results of real-time information processing are used to control the process or for other actions. Such systems put forward increased demands not only on the performance of computing facilities, but also on the lag of information in the feedback loop, which are not provided within the framework of traditional approaches [7,9]. On the other hand, the human visual system has improved over millions of years and has reached an extremely c Springer Nature Switzerland AG 2020 S. Babichev et al. (Eds.): DSMP 2020, CCIS 1158, pp. 22–44, 2020. https://doi.org/10.1007/978-3-030-61656-4_2
The Principles of Organizing the Search for an Object
23
high level of organization. Therefore, the phenomenon of vision provides an extremely many diverse solutions for computer vision systems. Despite the enormous amount of information in the image, and especially in the video sequence, the human visual-analyzing system very effectively and efficiently copes with these problems due to its extremely high selectivity [4,8,16,18,25]. There is a significant semantic gap in how a person perceives and describes an image and how the image is perceived by the video system. A person identifies the semantics of the image, and the video system represents the visual content of the image in the form of its low-level characteristics such as color, texture, orientation, shape, the presence of movement, etc. [22]. The second section briefly discusses the principles of organization of the human visual system, the third - the state of the problem, in the fourth - a generalized model of human perception in dynamics. The fifth section is devoted
Data Loading...
