A research of exploring advancement in crowd sensing through infrared based pedestrian counter

A research of exploring advancement in crowd sensing through infrared based pedestrian counter

Keywords: crowd monitoring, sensors, data analysis.

Abstract: The term “crowd sensing” refers to sharing data collected by sensing devices with the aim to measure phenomena of common interest. It is increasingly finding its application in transport and traffic. The recent technological advancement in crowd-sensing has opened up new perspectives for cost-effective ways of managing the traffic congestion as well as safety in a critical situation such as evacuation. This project will explore the advancement in crowd sensing especially towards managing the passengers crowds at major transport systems. It will also provide an opportunity to work on data collection and analysis that involves crowd sensing through Smartphone.

Introduction

The management and control of crowds is a crucial problem for Human Life and Safety to carry out such task; there is an established practice of using extensive closed circuit television system. However, a significant number of video cameras often used by such systems require a massive recording and storage capacity and some people to observe the television monitors (Bertozzi, 2007, p.329). As routine monitoring is tedious, the observers are likely to lose concentration. The advantages and necessity of the automatic surveillance for routine crowd monitoring and controlling are, therefore, clear. In trying to automate the mentioned surveillance, many problems arise.

Surveillance operators are today monitoring a large number of CCTV screens, trying to iron out the cognitive chores of analyzing crowd behavior and identifying threats and other unusual behavior. Information overload is a rule rather than exclusion (Bertozzi, 2005, p.24). Moreover, CCTV footage lacks some crucial indicators revealing certain threats, and can also in other respects be complemented by data from other sensors. This article presents an approach to interpreting automatically sensor data and estimate behaviors of groups of people to provide the operator with relevant warnings. We use data from distributed heterogeneous sensors (visual cameras and a thermal infrared camera), and process the sensor data using detection algorithms. The extracted features are fed into a hidden Markov model to model normal behavior and detect deviations. This paper proposes a computer vision methodology for estimating of crowd density. We will apply an experiment; will involve the use of infrared sensors installed at strategic positions to count people. The infrared sensors will then transmit data via Wi-Fi.  I have some articles

1. Review of literature

The operation “exploring advancement in crowd sensing through infrared based pedestrians counter “, has greatly attracted the concerns of many researchers and scholars over the past years. Previous researchers have documented their findings in various articles and journals.

Visual observation and group examination in element scenes with people are exceptionally dynamic exploration points in PC vision. The conceivable applications are various, as are the quantity of productions in the zone.

This current article’s attention is on the item tracker part of the observation framework, and the references in that for a review. The study is an early distribution where a molecule channel is utilized for visual form following. In a blend molecule channel and support finder is utilized to track different articles (hockey players) in a video stream. Visual following is frequently performed in the picture plane with the advantage of keeping the state measurement low and staying away from the adjustment of outward camera parameters, i.e., the area and introduction of the camera relative a world reference outline (Goubet, 2006, p.6262c). In this study, following is performed in worldwide directions which disentangles the movement model of the objective furthermore makes it simpler to join with other following frameworks and logical information about nature. Following in worldwide directions with a dream sensor is basically equal to following with a heading just sensor which has been customarily treated in the objective following group, and the references in that.

Affiliation is a difficult issue, particularly with a solitary camera in swarmed situations with impediments. A various leveled affiliation methodology is proposed in to frame the directions of the walkers. The strategy likewise contains a programmed scene structure estimator. The study gauges the probabilities of the inhabitance algorithm in the ground plane spoke to as a lattice. The Viterbi calculation is then used to gauge target directions in an arrangement of edges (Greene, 2008, p.100). One normal methodology for taking care of impediment is to utilize different perspectives keeping in mind the end goal to have the capacity to use the profundity data. In a planar homographic requirement is utilized to find the objectives on the ground plane. Just the sorts of impediment which are articulated to stationary and known items like structures and trees are considered in this study.

In a great observation setup the vision sensors are stationary, however as of late various people on foot location and following frameworks have been proposed for moving cameras in car applications, the study utilizes structure-from-movement to appraise the ground plane that backings the objective following.

Target following with street system data requires philosophies which can keep the characteristic multi-methodology of the fundamental likelihood densities. The main endeavors utilized the hop Markov (non) linear frameworks in mix with the interfacing numerous model (IMM) calculation with augmented Kalman channels (EKFs) as sub-squares (Li, 2010, p.268). Since the diverse street sections compare to various modes in these IMM calculations, there are an excessive number of them to be considered at a solitary stride of the different model channel. Thus, these calculations connected the purported variable structure cooperating different model (VS-IMM) calculation which includes/expels modes into/from the channel when fundamental.

Essential distinct options for IMM based strategies propose variable structure various mode

Molecule channels (VS-MMPF) as an expansion of the VS-IMM approaches. Since the molecule channels can deal with nonlinear and non-Gaussian models, the client has substantially more flexibility than in VS-IMM demonstrating (Aburafah, 2012, p.135). The street requirements are taken care of utilizing the idea of directional procedure commotion. In the streets are 3D bends spoken to by straight sections and the street system is spoken to as a diagram with streets and crossing points as the edges and hubs, separately. The position and speed along a solitary street are displayed by a standard straight Gauss-Markov model. The objective can be veiled both by the disorder indent of the sensor and by landscape snags. The outcomes for a Gaussian entirety channel and a standard bootstrap molecule channel methodology are displayed.

2. Gap in research

Some essential things have been left out in some researches that have been in conducted on this topic. Analysis of crowd control and crowd management should also incorporate formation of expectations regarding visitor behaviors and practices based on typologies. Typologies found in scientific literature distinguish between crowds on the basis of several traits for instance the public’s goals and plans, the duration of an event, the size of the crowd, event locations and visitor movements between these locations, the general atmosphere of the event and the composition and size of various subgroups. There is also inconsistency in time plan implementation, which involves control of groups of intersections among crowds using online generated timing plans and also traffic adjusted control among the traffics. There is also need for modification of presence and passage sensors for better analysis and excellent results.

The documentations entail the following.

• Why choose infrared?
• What is the appropriate sensor and data acquisition system, which can be used in real-time monitoring and control?
• What is the appropriate S/W algorithm to estimate crowd control?
• What is the dynamic model of the flow of large crowd of pedestrians?

2.1 Why use infrared?

In the light of scholarly articles and journals on real-time monitoring using infrared thermal video sequences,

Moving a giant number of people in an unordered and uncontrolled manner led to many accidents in the past twenty years. Avoiding critical crowd densities and triggering rapid group movement in a relatively short period of time have become critical tasks required to avoid future accidents and keeping the sanctity of emotions its best. The proposed technique uses a far infrared camera (sometimes also called FLIR camera) for monitoring and estimating the density of crowd in real time. Although far infrared cameras were originally developed for military use, lowering their prices make them now accessible for civilian use. There are some applications of it in industry especially inequality control. However, its technology is not yet widely available, and many researchers lack the experience in Acquiring far infrared.

Use of infrared in monitoring the crowd involves use of infrared thermographs (Bu, 2005, p.104). The existence and size of a crowd can be estimated by detecting persons and counting the number of detections. This can be done with both visual and TIR data. However, counting the number of detections may be difficult if the crowd is dense and specific individuals cannot be separated.       Infrared sensor networks used in mobile crowd sensing has unique properties that bring both new problems and opportunities. Mobile devices have significally more computing, communication and storage resources than mote class sensors. They are equipped with multimodality sensing capacities. Instead of installing road-side cameras and loop detectors, we can collect traffic data and detect congestion levels using smart phones carried by drivers. Such solutions reduce the cost of deployment of specialized sensing infrastructure. The dynamic conditions of the set of mobile devices and the need for data reuse across different applications in MCS are also quite different from those of traditional sensor networks. In MCS, the population of mobile devices, the type of sensor data each can produce and the quality in terms of accuracy, latency, confidence can change all the time due to device mobility, variations in their energy levels and communication channels, and device owners’ preferences. Identifying the right setoff devices that can produce the desired data and instructing them to sense with proper parameters to ensure desired quality is a complex problem.

Another way of approximating the size of the crowd is to estimate the part of the image where foreground objects are present, and divide this number with the total amount of pixels. With calibrated cameras, and the head detection as described above, the distance to the crowd can be known. This improves the estimation of the crowd size. A threshold value is used to indicate when a crowd is assumed to be present. Sometimes it is of interest to estimate the existence and size of a crowd within a specific area in the image(Hancke,2012 p.399). This can be done by limiting the search for foreground objects to this specific area. What is considered as a large crowd will differ from case to case.

To obtain a realistic estimation of the crowd size, measurements should be performed over a certain time period. Detecting and tracking people in dense crowds is a challenging problem. The idea in this paper is instead to regard the crowd as one unit that emits observations which can be used to estimate the behavior of the crowd as being normal or abnormal. Another way of approximating the size of the crowd is to estimate the part of the image where foreground objects are present, and divide this number with the total amount of pixels. With calibrated cameras, and the head detection as described above, the distance to the crowd can be known (Hancke, 2012, p.398). This improves the estimation of the crowd size. A threshold value is used to indicate when a crowd is assumed to be present. Sometimes it is of interest to estimate the existence and size of a crowd within a specific area in the image. This can be done by limiting the search for foreground objects to this specific area. What is considered as a large crowd will differ from case to case. To obtain a realistic estimation of the crowd size, analysis and measurements should be performed over a certain time period.

Infrared Thermography. Thermo graphic images represent the electromagnetic radiation of an object in the far infrared range, which isthe principle of Thermography is based on the physical principle that any body of a temperature above absolute zero degrees emits electromagnetic radiation (Suard, 2006, p.207). The electromagnetic spectrum is the range frequencies of electromagnetic radiation and is divided, low wavelength to highest wavelength, for instance the infrared rays.

The thermographic image depicts emitted energy from an object, and the relation between the temperature of an object and the emitted energy is given by the Stefan Boltzmann formula, thermography can be used to emit energy that is described by wavelength according to the following equation: be simply interpreted as a measurement technique, which, in most cases, is able to quantitatively measure surface temperatures of objects with different pixel intensities representing different temperatures. Thermography makes it possible to see one’s environment with or without visible illumination (Li, 2010, p.268). The amount of radiation emitted by an object increases with temperature; therefore, thermography allows one to see variations in temperature. When viewed through a thermal imaging camera, warm objects stand out well against cooler backgrounds; humans and other warm-blooded animals become easily visible against the environment, day or night.

• What is the appropriate sensor and data acquisition system, which can be used in real-time monitoring and control?

The experimental sensor system consists of gyro-stabilized gimbals with IR and CCD video sensors, and an integrated high-performance navigation system. The navigation system combines GPS with data from an inertial measurement unit (IMU) mounted with reference to the optical sensors. However, in the experiments presented in this article external landmarks with known location have also been used to estimate the orientation of the camera relative the world frame by using standard camera calibration techniques.

The IR sensor in the gimbals is a FLIR system ThermaCAM SC3000, which is a long-wave infrared (LWIR) sensor with a quantum well infrared photo detectors (QWIP) focal plane array. It has a low noise equivalent temperature difference (NETD) of (Goubet, 2006, p.100-105). The detector array is composed of  pixels with a comparatively narrow spectral sensitivity of which corresponds to the wavelength peak of an equivalent black body radiator atThe digital output has a resolution of  and a frame rate of  The mounted optics has a field-of-view of which gives a spatial angular resolution of per pixel.

2.3 What is the dynamic model of the flow of large crowd of pedestrians?

The detection problem is to find targets in cluttered backgrounds and the output from the detector is a set of image coordinates for all detections in each video frame. In this study a sliding window approach is used to detect pedestrians in cluttered backgrounds (Bertozzi , 2007, p.327-332). At each image position, the content of a local image region is fed into a classifier that decides whether or not the region contains a target.

The classifier is trained using a variant of boosting. Boosting iteratively builds a highly discriminative classifier by combining the outputs of many component functions often referred to as “weak learners”. Applying the resulting classifier to an image window x, the output can be written as and the window is classified as containing a target if the confidence sum(x) is greater than a threshold that is set to achieve an acceptable false alarm rate(Hancle,2012).Viola and Jones .Proposed a highly efficient cascade-structured detector architecture where each stage is a boosting classifier that is trained to reject a moderate fraction of the remaining background examples, while retaining a large fraction of the target examples. This leads to an exponential decay in the probability that a retained window belongs to the background class. Another important contribution by is the design of weak learners that can be computed very efficiently.

3. Conclusion

In this paper, I have highlighted different technique that can be of great use in remote sensing, as discussed use of I advocate for use of body sensors in improving crowd security. According to the research body sensors can be used in crowd density and behavior estimation in real time as it has been proposed. This technique is applied when infrared FLIR camera is concerned. The technique used is able to process the thermal camera output such as video sequences in real time and produce an estimation of the crowd density. According to the research thermography can be applied as the best technique in crowd sensing due to the following reasons. This new technique   is non-contact. It involves use of remote sensing and therefore keeping the user safe. It keeps people privacy intact, no intruding. The produced images give excellent overview of the target without body or facial recognition of respective person. The paper also shows how the contributions from different infrared sensors affect the overall opinion of crowd behavior. The different sensors also show abnormality at different times during different times, different scenes and different time periods. Advancement in crowd sensing through infrared will improve crowd security.

4. Reference

Bertozzi, M., Broggi, A., Gomez, C.H., Fedriga, R.I., Vezzoni, G. and Del Rose, M., 2007, June. Pedestrian detection in far infrared images based on the use of probabilistic templates. In Intelligent Vehicles Symposium, 2007 IEEE (pp. 327-332). IEEE.

Bertozzi, M., Broggi, A.Lasagni, A. and Rose, M.D., 2005, June. Infrared stereo vision-based pedestrian detection. In Intelligent Vehicles Symposium, 2005.Proceedings. IEEE (pp. 24-29). IEEE.

Al-Salhie, L., Al-Zuhair, M. and Al-Wabil, A., 2014. Multimedia surveillance in event detection: crowd analytics in Hajj. In Design, User Experience, and Usability. User Experience Design for Diverse Interaction Platforms and Environments (pp. 383-392).Springer International Publishing.

Abuarafah, A.G., Khozium, M.O. and AbdRabou, E., 2012.Real-time crowd monitoring using infrared thermal video sequences. Journal of American Science, 8(3), pp.133-140.

Bu, F. and Chan, C.Y., 2005, June. Pedestrian detection in transit bus application: sensing technologies and safety solutions.InIntelligent Vehicles Symposium, 2005.Proceedings. IEEE (pp. 100-105). IEEE.

Goubet, E., Katz, J. and Porikli, F., 2006, May. Pedestrian tracking using thermal infrared imaging. In Defense and Security Symposium (pp.62062C-62062C).International Society for Optics and Photonics.

Hancke, G.P. and Hancke Jr, G.P., 2012.The role of advanced sensing in smart cities. Sensors, 13(1), pp.393-425.

Suard, F, Rakotomamonjy, A., Bensrhair, A. and Broggi, A., 2006, June. Pedestrian detection using infrared images and histograms of oriented gradients. In Intelligent Vehicles Symposium, 2006 IEEE (pp.206-212).IEEE.

Greene-Roesel, R., Diogenes, M.C., Ragland, D.R. and Lindau, L.A., 2008. Effectiveness of a commercially available automated pedestrian counting device in urban environments: comparison with manual counts. Safe Transportation Research & Education Center.

Li, J., Gong, W., Li, W. and Liu, X., 2010.Robust pedestrian detection in thermal infrared imagery using the wavelet transform. Infrared Physics & Technology53 (4), pp.267-273.