Abstract
Moving object detection and tracking is the process of identifying and locating the class objects such as people, vehicle, toy, and human faces in the video sequences more precisely without background disturbances. It is the first and foremost step in any kind of video analytics applications, and it is greatly influencing the high-level abstractions such as classification and tracking. Traditional methods are easily affected by the background disturbances and achieve poor results. With the advent of deep learning, it is possible to improve the results with high level features. The deep learning model helps to get more useful insights about the events in the real world. This chapter introduces the deep convolutional neural network and reviews the deep learning models used for moving object detection. This chapter also discusses the parameters involved and metrics used to assess the performance of moving object detection in deep learning model. Finally, the chapter is concluded with possible recommendations for the benefit of research community.
TopIntroduction
Video surveillance is an emerging field with more advent for anonymous activity monitoring in the restricted areas and it greatly becomes a part of the life today. Usually, the video surveillance systems observe and analyze the huge amount of visual information to find out the suspicious activities in the given image frame. However, it is difficult to store and analyze the substantial surveillance data manually due to boredom and exhaustion. Alternatively, an intelligent video surveillance system can support the manual operations in case of event detection and other activity based analysis. A typical video frame comprises both foregrounds as well as backgrounds information. The pixel points which describe the target features in the region of interest are considered as foreground information and the rest of the feature points are treated as background information. In most of the cases, existing moving object detection approaches concentrate only on the foreground information and frequently ignored the background information. As a result, trackers will be deviated away from the target and detect the non-foreground objects. Figure 1 shows a typical smart video surveillance system which includes major steps such as object detection, object classification, object tracking, and event analysis.
Figure 1. Typical video surveillance systems
Key Terms in this Chapter
Deep Neural Network (DNN): It is a network with more than two layers and the word “deep” refers to the number of layers through which the data is transformed.
Neural Network: It is a computing system with interconnected nodes that can recognize hidden patterns and their correlations in input data.
Object Classification: It is the second step in smart video surveillance system. It will classify the objects into people, vehicle, animal, and other targets such as toys, buildings, etc.
Moving Object Detection: It is the process of identifying the class objects such as people, vehicle, toy, and human faces in the video sequences.
Mini-Batch: It is the number of samples used to train a model in each iteration.
Epochs: The number of data passes in DNN model is called as epochs.