FPGA Implementation of Moving Object Segmentation using ABPD and Background Model

Hardware Implementation of an efficient technique for moving object segmentation in video frame plays important role in real-time applications such as Video Surveillance system, traffic monitoring system, robotics vision, etc. This paper presents a hardware implementation of moving object segmentation using Adaptive Block Partition Decision Technique (ABPD) with a fixed threshold and pixel- based background model. The pixel-based background model is included to make background subtraction method insensitive to surrounding environment illumination variations and noise, also it makes an algorithm useful for the case where no background image exist or hard to generate. Behavior of proposed method is improved in terms of FP rate, Precision, Similarity, and F-measure. The architecture for proposed algorithm is implemented on a vertex6 xc6vlx75t device with 5 pipeline stages. This architecture achieves high throughput with less resource utilization compared to existing architectures.