Towards Scalable Video Analytics at the Edge

다국어 초록 (Multilingual Abstract)

Currently, Intelligent Video Analytics systems have come to play an essential role in many fields, including public safety, transportation safety, and many other industry areas as automated tools for data extracting and analyzing huge datasets such as multiple live video streams transmitted from a large number of cameras. A key characteristic of such systems is that it is critical to perform real-time analytics so as to provide timely actionable alerts on various tasks, activities and conditions. Due to the computation-intensive nature of these operations, however, video analytics servers may not fulfill the requirements when serving for a large number of cameras simultaneously. To handle these challenges, this dissertation present an edge-computing based system that minimizes transfer of video data from the surveillance camera feeds on a cloud video analytics server Based on a novel approach of utilizing the information from the encoded bitstream, the edge can achieve low processing complexity of object tracking in surveillance videos and filters non-motion frames from the list of data which will be forwarded to the cloud server. To demonstrate the effectiveness of the approach, we have implemented a video surveillance prototype consisting of edge devices with low computational capacity and a GPU-enabled server. The evaluation results show that our method can efficiently catch the characteristics of the frame and is compatible with the edge-to- cloud platform in terms of accuracy and delay sensitivity. The average processing time of this method is approximately 39 ms/frame with high definition resolution video, which outperforms most of the state of art method. In addition to scenario implementation of the proposed system, the method helps the cloud server reduce 49% load of GPU, 49% load of CPU, and 55% of network traffic while maintaining the accuracy of real-time 18 video analytics event detection.

목차 (Table of Contents)

• CHAPTER 1. Introduction 1
• CHAPTER 2. Background 5
• 2.1. Surveillance Video Analytics Architectures and Challenges 5
• 2.1.1. Edge Camera Based Implementation 5
• 2.2. Video Codec and Video Coding Motion Vectors 9
• 2.2.1. Video Codec 9
• 2.2.2. Video Coding Motion Vectors and Blocked-based Motion Estimation 11
• 2.3. Compressed-Domain Based Moving Object Detection 16
• 2.4. Pixels-Domain Based Moving Object Detection 18
• 2.4.1. Hybrid model of background subtraction and object classification Based moving object detection 19
• 2.4.2. Deep Learning based Moving Object Detection 22
• CHAPTER 3. Methodology 25
• 3.1. The Edge-to-cloud System Model for Surveillance Camera based Applications 26
• 3.2. The Light-weight Runtime Moving Object Detection In Video Compressed Domain 28
• 3.2.1. Median Filter and Moving Object Detection 29
• 3.2.2. The IoU based Moving Objects Tracking 33
• 4.3. Performance Evaluated Model 35
• CHAPTER 4 37
• Implementation and Performance Evaluation 37
• 4.1. Scenario Setup 37
• 4.2. The Light-weight Runtime Moving Object Detection in Video Compressed Domain 40
• 4.3. Performance Evaluation Results 43
• 4.3.1. Computing Resources Consumption 44
• REFERENCES 48