Object detection is important and challenging task in computer vision applications such as surveillance, vehicle navigation, and human tracking. Video surveillance is a key technology to fight against terrorism and public safety management. In video surveillance, detection of moving objects from a video is important for object detection and behaviour understanding. Detection of moving objects in video streams is important process of revelation and background subtraction is popular approach for foreground segmentation. In this paper high speed background subtraction algorithm for moving object detection is proposed. The video is first converted to streams and then applied to convolution filter which removes high frequency noise components to obtain smoothened images. The smoothened images are then applied to background subtraction algorithm with adaptive threshold which gives detected object present in background image. The detected object is then applied to convolution filter to remove the spurious distorted pixels which improves the quality of image. The proposed architecture was designed using VHDL language and implemented using Spartan-6 (XC6SLX45-2csg324) FPGA kit. It is observed that the proposed technique is better compared to existing method in terms of image quality and speed of operations.
展开▼
