Designing a gaussian mixture model algorithm using background identification for high definition video segmentation
Loading...
Date
item.page.authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Nowadays the real time object detection and identification which are in motion is more important in real time applications such as traffic monitoring, vehicle detection etc. Several methods or techniques are developed to detect the objects in a variable background region. By comparing two consecutive frames of the real time input video sequence, dynamic parts of foreground image is identified with the help of temporal difference technique. Objects which are in motion are detected by analyzing the pixel value difference between the foreground and reference background image. A multimodal
newlinebackground is the one where scene objects showing repetitive motions. Foreground objects are not stationary always. For example waves in the ocean, leaves in the tree, flickering light etc. The pixels in the multimodal background will take one or more values. This fluctuation of the values will lead to false detection in the foreground image analysis. Sudden changes in the intensity level or movement of leafs in the tree will leads to trigger the CCTV to record the changes in the foreground image with respect to reference background image. This leads to waste of storage memory by capturing the unnecessary data. Also if the motion of the object is faster, then the quality of the image captured is degraded.
newline
newlineIn order to overcome the above said problems we are designed and developed new improved version of Gaussian Mixture Model (GMM) and Background Identification Module. This will reduces the false alerts and increases the quality of image for motion detection application. In this research work we can avoid the false alert which is triggered to capture the scene. This new design will increase the memory utilization and also it reduces the power consumption. The quality of the image captured is also increased irrespective of the object motion with the help of Feature Extraction and ANN Model.
newline
newline