Research On Scene Recognition Algorithm In Old Guardianship System

The stereoscopic scene analysis method of the old people monitoring system is a new type of information application developed by modern computer vision technology and pattern recognition method.It has the ability to real-time extract and analyze image information.Now,most of them are applied in the area of artificial intelligence,Human-computer interaction technology,robot navigation and so on.The most important technology in the old guardianship system is the recognition and segmentation of the objects in the scene.Because the foreground objects can be extracted accurately is important for further object recognition and scene analysis combined with gesture recognition.Because the environment is very complicated,the scene is not a single fixed change,the camera to the image when the environment will be most of the lights,shadows and other environmental factors and the surrounding environment noise interference.So how to accurate and complete segmentation is a certain degree of difficulty with the target audience.Based on this problem,researchers ai m to study how to enhance and optimize the dynamic robustness and efficiency of objects detection algorithm in the scene.And improve the accuracy of the semantic analysis of subsequent scene analysis.This paper first introduces the development and current situation of scene analysis application technology,and then studies and analyzes the key technologies in the elderly monitoring system: the object recognition and segmentation theory.This paper designs and implements scene object detection method which does not depend on the traditional structure light in the old guardianship system.And realizes the robustness detection of objects in the scene by using th e principle of speckle of Prime Sense.The foreground segmentation and scene reconstruction of the foreground objects in the scene are realized by using stereo vision technology.The multi-layer perceptron is used to identify foreground objects and scenes in the scene.In this paper,the scene recognition algorithm used in the old guardianship system is researched,and the simulation is discussed in depth.The results show that the recognition rate of foreground object detection in the scene is 100%,and 98% better recognition result is achieved by the combination of attitude recognition and scene analysis.And in the processing speed to achieve the performance requirements of real-time processing to meet the old guardianship system used to analyze the scene function and performance requirements.