Arka plan çıkarma başarımı iyileştirmek için yeni yaklaşımlar

Separation of the foreground from background on a processed image, namely background modelling, positively affects performance of certain computer vision applications. It has considered as preprocess for many tasks including moving object recognition, person tracking, traffic monitoring, motion capturing, teleconference and security surveillance systems. Video backgrounds can be considered in two categories as static and dynamic backgrounds. To improve the performance of background subtraction, we have developed four different methods by using different tools in case of distance computation between test and background frame and integrating a feedback mechanism that works beyond dynamic controller parameters. These methods are called as Background Modelling Using Common Vector Approach (BMCVA), Background Modelling Using Common Matrix Approach (BMCMA), Sliding Window-Based Change Detection (SWCD) and Common Vector Approach Based Background Subtraction (CVABS). Various experiments have conducted on different problem types related to dynamic backgrounds over CDnet2014 and Wallflower datasets. Several types of metrics calculated over the results of True-Positive (TP), True-Negative (TN), False-Positive (FP) and False-Negative (FN) counts, have utilized as objective measures and the obtained visual results are judged subjectively. Once the obtained results inspected, it has observed that the proposed methods generate successful results for different challenges.