Monitoring a large surveillance space through distributed face matching
01 Dec 2013-pp 1-5
TL;DR: A distributed camera and processing based face detection and recognition system which can generate information for finding spatiotemporal movement pattern of individuals over a large monitored space is proposed.
Abstract: Large space with many cameras require huge storage and computational power to process these data for surveillance applications. In this paper we propose a distributed camera and processing based face detection and recognition system which can generate information for finding spatiotemporal movement pattern of individuals over a large monitored space. The system is built upon Hadoop Distributed File System using map reduce programming model. A novel key generation scheme using distance based hashing technique has been used for distribution of the face matching task. Experimental results have established effectiveness of the technique.
Citations
More filters
TL;DR: This work proposes a novel distributed protocol for a face recognition system that exploits the computational capabilities of the surveillance devices (i.e. cameras) to perform the recognition of the person.
Abstract: Video surveillance systems have become an indispensable tool for the security and organization of public and private areas. Most of the current commercial video surveillance systems rely on a classical client/server architecture to perform face and object recognition. In order to support the more complex and advanced video surveillance systems proposed in the last years, companies are required to invest resources in order to maintain the servers dedicated to the recognition tasks. In this work, we propose a novel distributed protocol for a face recognition system that exploits the computational capabilities of the surveillance devices (i.e. cameras) to perform the recognition of the person. The cameras fall back to a centralized server if their hardware capabilities are not enough to perform the recognition. In order to evaluate the proposed algorithm we simulate and test the 1NN and weighted kNN classification algorithms via extensive experiments on a freely available dataset. As a prototype of surveillance devices we have considered Raspberry PI entities. By means of simulations, we show that our algorithm is able to reduce up to 50% of the load from the server with no negative impact on the quality of the surveillance service.
27 citations
TL;DR: It is noticed that the node distribution through the proposed scheme not only provides better coverage in each layer but also minimizes both the energy-hole and the coverage-hole problems in the deployment field while maintaining longevity of the sensor network.
Abstract: Wireless sensor networks are equipped with sensor nodes having limited battery as energy source. These sensor nodes have to maintain the desirable coverage of the network to ensure the periodical communication of the sensed data to the base station. Therefore, lifetime of sensor nodes and the energy efficient network coverage are the two major issues that needs to be addressed. Effective placement of wireless sensor nodes is of paramount importance as the lifetime of the network depends upon it. In this work, a corona based energy balanced node deployment scheme for sensors with a limited sensing range is proposed in which the nodes are distributed in accordance with a probability density function (PDF). Optimal number of nodes in each corona is determined using the proposed PDF. Performance of the scheme is evaluated in terms of coverage, energy balance and network lifetime through simulation. The intrinsic characteristic of the proposed PDF has been derived. It is noticed that the node distribution through the proposed scheme not only provides better coverage in each layer but also minimizes both the energy-hole and the coverage-hole problems in the deployment field while maintaining longevity of the sensor network.
12 citations
TL;DR: The design of a Probability Density Function (PDF) targeting the desired coverage, and energy efficient node deployment scheme, and the simulation results obtained confirm the schemes superiority over the other existing schemes.
Abstract: Energy consumption is one of the important issues in wireless sensor network that rely on non chargeable batteries for power. Also, the sensor network has to maintain a desired sensing coverage area along with periodically sending of the sensed data to the base station. Therefore, coverage and the lifetime are the two important issues that need to be addressed. Effective deployment of wireless sensors is a major concern as the coverage and lifetime of any wireless sensor network depends on it. In this paper, we propose the design of a Probability Density Function (PDF) targeting the desired coverage, and energy efficient node deployment scheme. The suitability of the proposed PDF based node distribution to model the network architecture considered in this work has been analyzed. The PDF divides the deployment area into concentric coronas and provides a probability of occurrence of a node within any corona. Further, the performance of the proposed PDF is evaluated in terms of the coverage, the number of transmissions of packets and the lifetime of the network. The scheme is compared with the existing node deployment schemes based on various distributions. The percentage gain of the proposed PDF based node deployment is 32
$$\%$$
more than that when compared with the existing schemes. Thus, the simulation results obtained confirm the schemes superiority over the other existing schemes.
9 citations
01 May 2017
TL;DR: The protocol organizes the distribution of a classification library among the cameras involved, which also participate actively to the target recognition phase, and minimizes the network overhead towards the centralized server while keeping high the speed of recognition.
Abstract: Video surveillance is an important security enforcement operation in many contexts, from large public areas to private smart homes and smart buildings. Today's video surveillance systems are much more than mere recording storages, as the advancement in classification and recognition allow for an immediate target recognition without the intervention of human operators. These smart video surveillance systems usually rely on a central server as the main coordination of recognition and tracking, which can represent a performance or economical bottleneck. In this paper, our contribution focuses on a decentralized protocol with the aim of eliminating such bottleneck. Our protocol organizes the distribution of a classification library among the cameras involved, which also participate actively to the target recognition phase. The protocol minimizes the network overhead towards the centralized server while keeping high the speed of recognition making use of a system to predict the movements of the targets. We tested the protocol by means of simulations, exploiting a realistic indoor human mobility model.
8 citations
Cites methods from "Monitoring a large surveillance spa..."
...The work of Mishra et al.[7] proposes a novel video surveillance application using HDFS and the MapReduce framework....
[...]
28 Dec 2015
TL;DR: A novel distributed protocol is proposed that exploits the computational capabilities of the surveillance devices (i.e. cameras) to perform the recognition of the person to reduce up to 50% the load of the server with no negative impact on the quality of the Surveillance service.
Abstract: Video surveillance systems have become an indispensable tool for the security and organization of public and private areas. Most of the current commercial video surveillance systems rely on a classical client/server architecture to perform person and object recognition. In order to support the more complex and advanced video surveillance systems proposed in the last years, companies are required to invest resources in order to maintain the servers dedicated to the recognition tasks. In this work we propose a novel distributed protocol that exploits the computational capabilities of the surveillance devices (i.e. cameras) to perform the recognition of the person. The cameras fall back to a centralized server if their hardware capabilities are not enough to perform the recognition. By means of simulations, we show that our algorithm is able to reduce up to 50% the load of the server with no negative impact on the quality of the surveillance service.
7 citations
Cites background from "Monitoring a large surveillance spa..."
...In this work we do not investigate and propose new image recognition algorithms....
[...]
References
More filters
TL;DR: In this paper, a face detection framework that is capable of processing images extremely rapidly while achieving high detection rates is described. But the detection performance is limited to 15 frames per second.
Abstract: This paper describes a face detection framework that is capable of processing images extremely rapidly while achieving high detection rates. There are three key contributions. The first is the introduction of a new image representation called the “Integral Image” which allows the features used by our detector to be computed very quickly. The second is a simple and efficient classifier which is built using the AdaBoost learning algorithm (Freund and Schapire, 1995) to select a small number of critical visual features from a very large set of potential features. The third contribution is a method for combining classifiers in a “cascade” which allows background regions of the image to be quickly discarded while spending more computation on promising face-like regions. A set of experiments in the domain of face detection is presented. The system yields face detection performance comparable to the best previous systems (Sung and Poggio, 1998; Rowley et al., 1998; Schneiderman and Kanade, 2000; Roth et al., 2000). Implemented on a conventional desktop, face detection proceeds at 15 frames per second.
13,037 citations
07 Jul 2001
TL;DR: A new image representation called the “Integral Image” is introduced which allows the features used by the detector to be computed very quickly and a method for combining classifiers in a “cascade” which allows background regions of the image to be quickly discarded while spending more computation on promising face-like regions.
Abstract: This paper describes a face detection framework that is capable of processing images extremely rapidly while achieving high detection rates. There are three key contributions. The first is the introduction of a new image representation called the "Integral Image" which allows the features used by our detector to be computed very quickly. The second is a simple and efficient classifier which is built using the AdaBoost learning algo- rithm (Freund and Schapire, 1995) to select a small number of critical visual features from a very large set of potential features. The third contribution is a method for combining classifiers in a "cascade" which allows back- ground regions of the image to be quickly discarded while spending more computation on promising face-like regions. A set of experiments in the domain of face detection is presented. The system yields face detection perfor- mance comparable to the best previous systems (Sung and Poggio, 1998; Rowley et al., 1998; Schneiderman and Kanade, 2000; Roth et al., 2000). Implemented on a conventional desktop, face detection proceeds at 15 frames per second.
10,592 citations
"Monitoring a large surveillance spa..." refers methods in this paper
...We are using Viola-Jones Face Detection algorithm [3] for the purpose of detection of the frontal faces at the server end....
[...]
TL;DR: This paper focuses on motion tracking and shows how one can use observed motion to learn patterns of activity in a site and create a hierarchical binary-tree classification of the representations within a sequence.
Abstract: Our goal is to develop a visual monitoring system that passively observes moving objects in a site and learns patterns of activity from those observations. For extended sites, the system will require multiple cameras. Thus, key elements of the system are motion tracking, camera coordination, activity classification, and event detection. In this paper, we focus on motion tracking and show how one can use observed motion to learn patterns of activity in a site. Motion segmentation is based on an adaptive background subtraction method that models each pixel as a mixture of Gaussians and uses an online approximation to update the model. The Gaussian distributions are then evaluated to determine which are most likely to result from a background process. This yields a stable, real-time outdoor tracker that reliably deals with lighting changes, repetitive motions from clutter, and long-term scene changes. While a tracking system is unaware of the identity of any object it tracks, the identity remains the same for the entire tracking sequence. Our system leverages this information by accumulating joint co-occurrences of the representations within a sequence. These joint co-occurrence statistics are then used to create a hierarchical binary-tree classification of the representations. This method is useful for classifying sequences, as well as individual instances of activities in a site.
3,631 citations
TL;DR: An EM-based algorithm to compute dense depth and occlusion maps from wide-baseline image pairs using a local image descriptor, DAISY, which is very efficient to compute densely and robust against many photometric and geometric transformations.
Abstract: In this paper, we introduce a local image descriptor, DAISY, which is very efficient to compute densely. We also present an EM-based algorithm to compute dense depth and occlusion maps from wide-baseline image pairs using this descriptor. This yields much better results in wide-baseline situations than the pixel and correlation-based algorithms that are commonly used in narrow-baseline stereo. Also, using a descriptor makes our algorithm robust against many photometric and geometric transformations. Our descriptor is inspired from earlier ones such as SIFT and GLOH but can be computed much faster for our purposes. Unlike SURF, which can also be computed efficiently at every pixel, it does not introduce artifacts that degrade the matching performance when used densely. It is important to note that our approach is the first algorithm that attempts to estimate dense depth maps from wide-baseline image pairs, and we show that it is a good one at that with many experiments for depth estimation accuracy, occlusion detection, and comparing it against other descriptors on laser-scanned ground truth scenes. We also tested our approach on a variety of indoor and outdoor scenes with different photometric and geometric transformations and our experiments support our claim to being robust against these.
1,484 citations
08 Apr 2005
TL;DR: This survey describes the current state-of-the-art in the development of automated visual surveillance systems to provide researchers in the field with a summary of progress achieved to date and to identify areas where further research is needed.
Abstract: This survey describes the current state-of-the-art in the development of automated visual surveillance systems so as to provide researchers in the field with a summary of progress achieved to date and to identify areas where further research is needed. The ability to recognise objects and humans, to describe their actions and interactions from information acquired by sensors is essential for automated visual surveillance. The increasing need for intelligent visual surveillance in commercial, law enforcement and military applications makes automated visual surveillance systems one of the main current application domains in computer vision. The emphasis of this review is on discussion of the creation of intelligent distributed automated surveillance systems. The survey concludes with a discussion of possible future directions.
712 citations
"Monitoring a large surveillance spa..." refers background in this paper
...Common application domains are airports [6], railways, underground, marine environments [7], traffic monitoring [9], public places like banks, super markets, homes, departmental stores [10], parking lots [12] and remote surveillance to find presence in football matches [11] or other activities....
[...]