Dynamic Focus Capture Method and Its Application
Shoujin Wang, Yang Hua Cao, Jingang Shi · 2016
With the development of computer science and technology, motion capture is increasingly applied in many fields such as film production, research analysis, human-computer interaction, intelligent monitoring and so on.The main task of the motion capture is to analyze and process the data from the sensor, and abstract to recognize the action.How to achieve fast and accurate implementation of the action capture has been a hot research in recent years. Background Sampling MethodThe background in this paper should be defined as camera environment that a pending motion captured object locates.For the proper operation of the entire system, it is important to select the background image.Ideally, the background should not be changed with the moving of pending motion-captured object, dividing image with background difference under ideal background is the most accurate.However, in practice, background captured by the camera is sure to impact by the motion of pending motion-captured object [1,2].For example: set the room as the background, a person to be the pending captured object.When the person moves within the camera shooting range, it will cause changes in the overall image brightness, the focal length of the shooting and shadows.At present, for better shooting results, most cameras will be automatically adjusted according to shooting images, such as adjusting the sensitivity and lens focal length [3,4].This is going against with the immobilization of background shooting, shooting parameters should be fixed as much as possible.The changing of shadow moving object is inevitable, light direction adjustment can alleviate this problem.In some cases, the background image itself needs to be updated appropriately.For example, during prolonged outdoor shooting, outdoor light will change gradually over time; in this case, the background image should be updated at regular intervals.Because all detected objects will be withdrawn every time when background being updated.And updating background frequently is relatively troublesome.By using background adaptation algorithm can be programmed to automatically update the background.Specific ideas are as follows: Storage record for a period of time, analyze each frame of the video frame-by-pixel and generate a new background image pixel by pixel [5,6].New background pixel should meet the following criteria: 1.The RGB color value of the pixel of recurring or similar appears in the same position for recording, and more than a certain percentage (e.g.related pixels within the ceiling of the room should be kept in the same color).2. The color of the pixel should be close to the corresponding pixel of last background image. Image Segmentation and PreprocessingImage Preprocessing.In order to improve the efficiency and reliability of the system, we must process the preprocessed image.The method used in this paper is gray image processing.In the RGB model, if R = G = B, the color shows a gray color, wherein the value of R = G = B is called the gray values, so each pixel of the grayscale image only requires one byte to store ash value (also known as the intensity value, brightness value), the range of grayscale is 0-255.There are four general methods of color images gray scaling [7,8]. Average Method.Take the average of the three-component color image as the gray value.