Research On Moving Target Measurement Method Based On Image Processing

In recent years, human-computer interaction problem has been an important research direction in the field of machine vision. At present it has extensive application and potential development space in many fields, such as traffic control, intelligent monitoring, air navigation, Somatic game for body building, etc.As a hot topic in the field of human-computer interaction, This paper mainly researches the moving target detection and target tracking in the measurement of moving target.For the defects of three-frame difference algorithm in moving target detection, an improved modeling method based on Gaussian mixture model combined with three-frame difference algorithm has been proposed. The main idea is to establish multidimensional Gaussian mixture model for each of the background pixels. Then the three-frame difference method was blended in the model to determine the background area and sports area in real time. So that the holes caused by three-frame difference can be removed and External rectangle which can follow the target's movement was added in, generated Gaussian model. Thus reducing the problem of large amount of calculation as a result of the intervention of Gaussian model, Save operation time and achieve the ideal effect of noise cancellation. The experimental results show that: compared with three-frame difference method, this method has better denoising effect and reduce more computation, applicable to real-time target detection.Aiming at Camshift algorithm interferes between target color and background color around and the problem of invalidation in the complex and changeful background.this paper presents an improved Camshift algorithm based on SURF feature point matching, For example, When the hand and facial overlapping phenomenon occurs, do the SURF matching according to the current frame moving targets and the original image. Thus sustained good tracking can be achieved under the color confusion. Based on the tracking algorithm in this paper, in allusion to hand movement model in the video,measured the relevant data of moving distance, speed, direction, and researched and analyzed the date. So the effectiveness of the algorithm can be verified. According to the above identifiable hand movement, the algorithm can be applied in the design ofSomatic game.