Person Re-Identification in multi-camera network

  The increasing availability of visual data provided by surveillance cameras placed in several locations of large cities provides a secure environment to people circulating on those areas. In the last few years, public authorities have increased significantly the number of cameras inst

Project Title

Person Re-Identification in multi-camera network

Project Area of Specialization

Artificial Intelligence

Project Summary

The increasing availability of visual data provided by surveillance cameras placed in several locations of large cities provides a secure environment to people circulating on those areas. In the last few years, public authorities have increased significantly the number of cameras installed in large cities over the world. Due to the large amount of visual data and the difficulty to perform manual processing, the automatic understanding of human activities performed in videos is relevant so that the security personnel can be aided by automated systems to perform their duties. The main goal of such systems is to analyse the scene focusing on the detection and recognition of suspicious activities performed by humans in the scene, so that the security personnel can pay closer attention to these preselected activities

Project Objectives

The aim of the project is to develop a person re-identification system which is capable enough to enhance the surveillance capability. The objectives of the project are:

1. Person re-identification to enhance current surveillance system.
2. Suspicious activity detection.
3. Undesirable event prediction for timely alerts to security personnel.
4. Make security systems smarter and less human dependent.

Project Implementation Method

Machine Learning

Datasets

1 Sequence S2_L1 from Crowd_PETS09 dataset

2 Sequence AB_HARD from AVSS2007 dataset

3 Sequence CHAP from ICGLab6 dataset

Algorithms

1. MTIC: Multi-Target Information Consensus

        2. KCF: Kalman Consensus Filtering

Multi-target information consensus

The KCF and ICF assume that there is a single target, or for multiple targets that the measurement-to-track association is provided. For a multi-target tracking problem, the data association and the tracking steps are highly inter-dependent. The performance of tracking will affect the performance of data association and vice-versa. Thus, an integrated distributed tracking and data association solution is required where the uncertainty from the tracker can be incorporated in the data association process and vice-versa. Among many single-sensor multi-target data association frameworks, the Multiple Hypothesis Tracking (MHT) and the Joint Probabilistic Data Association Filter JPDAF are two dominant approaches. MHT usually achieves higher accuracy at the cost of higher computational load. Alternatively, JPDAF achieves reasonable results at lower computation cost. As distributed solutions are usually applied within low-power wireless sensor networks where the computational and communication power is limited, the JPDAF scheme will be utilized in the proposed distributed multi-target tracking framework herein. The proposed distributed tracking and data association framework is termed as the Multi-Target Information Consensus (MTIC)

Kalman consensus filtering

Decentralized Kalman Filters with an all-to-all communication complexity of O(n2) were applied to data fusion for sensor networks in. The author introduced the first scalable and distributed Kalman Filtering (DKF) algorithm which relied on dynamic average-consensus. Recently, it was revealed in that the performance of the DKF algorithm with embedded averaging consensus filters has a relatively weak performance that is comparable with the collective estimation error of n non-cooperative local Kalman Filters which is a trivial base performance level for distributed estimation in sensor networks. This weakness in performance of the DKF algorithm, motivated the development and introduction of the Kalman-Consensus Filter (KCF).

Benefits of the Project

1. Person re-identification to enhance current surveillance system.
2. Suspicious activity detection.
3. Undesirable event prediction for timely alerts to security personnel.
4. Make security systems smarter and less human dependent.

Technical Details of Final Deliverable

Make security systems smarter and less human dependent.

Final Deliverable of the Project

HW/SW integrated system

Core Industry

Security

Other Industries

Core Technology

Artificial Intelligence(AI)

Other Technologies

Sustainable Development Goals

Industry, Innovation and Infrastructure, Sustainable Cities and Communities

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com