A Comprehensive Survey on Computer-Vision Object Detection, Segmentation, Tracking, and Feature Extraction

Abstract

Computer-Vision [CV] applications like surveillance, autonomous vehicle navigation involves object detection and tracking using multiple features[1]–[4]. Video surveillance technology can be effectively implemented in a real-time application like public security, logistics management, and others, which involves continuous monitoring of the environment. In a video sequence, moving object detection from a video is the primary step which helps in object tracking and behavior understanding. Video surveillance in a dynamic scene with objects like vehicles and humans is a challenging task in the CV. Moving object detection involves designing efficient video surveillance algorithms for complex environments like the variation in illumination, brightness and environmental effects. Capturing and studying the moving object segmentation is an important task in vision-based object movements. Monitoring human behavior is an important task in a security surveillance system, hence it has become an active research area due to the need for such systems in civic areas. Object tracking based algorithms such as point tracking, Kernel tracking, and Silhouette tracking detect immobile background objects when the camera is fixed and the change in illumination condition is sluggish, and they separate foreground objects from the present frame. Static background surveillance systems are monitored by individuals, viewing several screen display through various camera feeds. Dynamic background detection is very difficult. Manual detection of the dynamic scene is very less effective. Hence leads to automating the entire process. Processing of a large amount of data in a video sequence which makes manual initialization cost-prohibitive. Research in CV has addressed semi-automatic and automatic methods to automate the contents of video data, to aid human operators. Advances in computation and communication techniques have increased the research interest in automating video content analysis.