A Survey on Object Tracking Based System for Safer

Transportation in Smart Cities

Helena Thomas#1, Josna Jose#2
#1
PG Scholar, Computer Science and Engineering,
College of Engineering Kidangoor, Kottayam, Kerala
#2
Asst.Professor, Computer Science and Engineering,
College of Engineering Kidangoor, Kottayam, Kerala

Abstract—Now a day’s safety and security are the most discussed are extracted from the current frame, the objective is to track
topics in the road and railway transportation field. While them. To achieve that, an object’s tracking method based on
trains are more convenient for travel and for transporting optical flow is applied. The tracking process starts by
goods, they have become a great danger over the years as their computing optical flow of corner points, extracted by Harris
speed has increased. In the current transportation system it is
operator, using the Lucas–Kanade algorithm.
becoming ever necessary to install safety elements to avoid
accidents. One of the causes that can provoke serious accidents
is the existence of obstacles, either fixed or mobile. Latest
security initiatives in the field of railway transportation propose
to implement video surveillance at level crossing (LC)
environments. This paper aims to compare the existing systems
and find an efficient method towards safer transportation. At
present video based analysis method is efficient. Four hazard
scenarios are tested and evaluated with different real video image
sequences: presence of the obstacle in the rail-road crossing,
presence of the stopped vehicles line, vehicle zigzagging between
two closed half barriers, and pedestrian crossing the area.

Keywords—Video based surveillance system, Hidden Markov

Model, Dempster-Shafer data fusion technique

I. INTRODUCTION
Improving rail-road crossings safety has become an important Fig 1: General representation of the architecture of the
field of academic research. Improving the safety of people and PANsafer project [1]
rail-road facilities is an essential element to ensure good
operation of the road and railway transport. Statistically,
nearly 44% of rail-road users have a negative perception of the
environment, which consequently increases the risk of II. LITERATURE REVIEW
accidents. The proportion of accidents at railway-road level
crossings is not very high for the society, but these accidents In the following section we will be discussing about various
are very dangerous. The risk of fatal outcome in an accident at methods promoting safer transportation:
a level crossing is twenty to forty times greater than in an
average road accident. Intensive users (mostly car drivers) are
highly exposed to the risk, and the population living in the
vicinity of level crossings is the most endangered one. Special A. Intra-Vehicle Networks: A Review
cases are multi-fatality accidents and especially accidents Automotive electronics is a rapidly expanding area with an
involving school buses. The method starts by detecting pixels increasing number of safety, driver assistance, and
affected by motion as a pre-treatment phase. To detect and infotainment devices becoming standard in new vehicles.
separate objects, this method consists of clustering, moving Current vehicles generally employ a number of different
pixels by comparing specific energy vectors associated with networking protocols to integrate these systems into the
each target and each pixel affected by motion. Once the targets vehicle. The introduction of large numbers of sensors to
 provide driver assistance applications and the associated
high bandwidth requirements of these sensors has
accelerated the demand for faster and more flexible network
communication technologies within the vehicle. This paper
presents a comprehensive overview of current research on
advanced intra-vehicle networks and identifies outstanding
research questions for the future.[11]

B. Hierarchical and Networked Vehicle Surveillance in ITS:

Traffic surveillance has become an important topic in
intelligent transportation systems (ITSs), which aims at
monitoring and managing traffic flow. The authors present
a review of the literature on the video-based vehicle Fig 2: Synopsis of the video surveillance system [1]
surveillance systems in ITS. They analyze the existing
challenges in video-based surveillance systems for the
vehicle and present a general architecture for video D. Efficient Pedestrian Detection via Rectangular Features
surveillance systems, i.e., the hierarchical and networked Based on a Statistical Shape Model:
vehicle surveillance, to survey the different existing and The authors propose a simple yet effective detector for
potential techniques. Then, different methods are reviewed robust pedestrian detection. Observing that pedestrians
and discussed with respect to each module. Applications usually appear upright in video data, the authors employ a
and future developments are discussed to provide future statistical model of the upright human body where the head,
needs of ITS services. [12] the upper body, and the lower body are treated as three
distinct components. The main contribution is to
systematically design a pool of rectangular features that are
C. Video-Analysis-Based Railway-Road System for tailored to this shape model. This approach does not avoid
Detecting Hazard Situations at Level Crossings: exhaustive searches over all possible configurations of
In this paper, the authors explore the possibility of rectangular features nor does it rely on random sampling. It
implementing a smart video surveillance security system that thus marks a middle ground among recently published
is tuned towards detecting and evaluating abnormal situations techniques and yields efficient low-dimensional yet highly
induced by users (pedestrians, vehicle drivers, and unattended discriminative features. Experimental results on the well-
objects) in level crossings. This intelligent security system established INRIA, Caltech, and KITTI pedestrian data sets
starts by detecting, separating, and tracking moving objects show that our detector reaches state-of-the-art performance
shot in the level crossing. Then, a hidden Markov model is at low computational costs and that our features are robust
developed to estimate ideal trajectories, allowing the detected against occlusions. [7]
targets to discard dangerous situations. After this, the level of
risk of each target is instantly estimated by using the E. A Pedestrian-Detection Method Based on Heterogeneous
Dempster–Shafer data fusion technique. The proposed Features and Ensemble of Multi-View-Pose Parts:
analysis also allows recognizing hazard scenarios. Four hazard Vision-based pedestrian detection remains a challenging
scenarios are tested and evaluated with different real video task so far. The detection performance often suffers from
image sequences: presence of obstacle in the level crossing, the various appearances of pedestrians, the illumination
presence of stopped vehicles line, vehicle zigzagging between changes, and possible partial occlusions. This paper
two closed half barriers, and pedestrian crossing LC area. [1] aims at resolving these challenges. Firstly, a new linear
 kernel function is proposed to effectively combine two H. Nonparametric Technique Based High-Speed Road
heterogeneous features, HOG and LBP, which enhances the Surface Detection:
pedestrian description ability to illumination conditions and Existing road surface detection methods attempt to fit the
cluttered background. Then, a novel multiple view–pose road surface into rigid models (e.g. planar, clothoid, or B-
part ensemble detector is proposed in order to better handle spline), thereby restricting to road surfaces that match
pedestrian variability, views, and partial occlusions. specific models. In addition, the curve fitting strategies
Experimental results on public data sets demonstrate that employed in such techniques incur high-computational
the proposed feature combination method significantly complexity, making them unsuitable for in-vehicle
improves the description capabilities of pedestrian features. deployments. In this paper, the authors propose an efficient
Compared with the existing multipart ensemble approaches, nonparametric road surface detection algorithm that
the proposed multiple view–pose ensemble detector boosts exploits the depth cue. The proposed method relies on four
higher detection accuracy. [6] intrinsic road scene attributes observed under stereo
geometry and has been shown to reliably detect both planar
and nonplanar road surfaces efficiently. Extensive
F. Longitudinal Model Identification and Velocity Control of evaluations are performed on three widely used benchmarks
an Autonomous Car: (i.e., enpeda, KITTI, and Daimler), encompassing many
This paper presents the model identification and the complex road scenarios. The experimental results show that
velocity control of an autonomous car. Since the the proposed algorithm significantly outperforms the well-
determination of the vehicle’s model is a difficult step in known techniques both in terms of detection accuracy and
the design of model-based controllers, the main runtime performance. [15]
contribution of this paper is the use of an empirically
determined model. In this work, the structure of the model
was conceived from the car’s physics equations, but its I. Analysis and Modelling of Pedestrian Crossing
parameters were estimated using data-based identification Behaviour During the Pedestrian Flashing Green
techniques. An important contribution of this paper is the Interval:
fact that, although the model is strictly linear, the authors The objective of this study is to analyze and model the
can change its parameters in function of the operation point probabilistic behaviour of pedestrians after the onset of
of the vehicle to represent engine’s and transmission’s PFG, which contains pedestrians’ decision of whether to
nonlinear behaviours. In addition, in this paper, the authors give up crossing, as well as pedestrian speed distribution.
propose a way to include changes in the longitudinal Empirical data analysis showed that longer crosswalks
dynamics caused by automatic gear shifting.[4] lead to significantly higher pedestrian stop probabilities.
Furthermore, there is a significant difference between
pedestrians’ speeds in the first and second halves of
G. Traffic Flow Prediction with Big Data: A Deep crosswalks. The results of a Monte Carlo simulation
Learning Approach: showed that the estimated models closely represent the
Over the last few years, traffic data have been exploding overall stochastic behaviour of pedestrians from the onset
and we have truly entered the era of big data for of PFG until the completion of crossing. [9]
transportation. Existing traffic flow prediction methods
mainly use shallow traffic prediction models and are still
insufficient for many real-world applications. This situation III CONCLUSION
inspires us to rethink the traffic flow prediction problem
based on deep architecture models with big traffic data. In The literature review presents the fact that there are large
this paper, a novel deep learning based traffic flow number of object detection techniques. We have discussed so
prediction method is proposed, which considers the spatial far many techniques, their advantages and disadvantages.
and temporal correlations inherently. The stacked auto Further the review reveals the fact that even though abundant
encoder model is used to learn generic traffic flow features, information on object detection is present, the integration of
and it is trained in a greedy layer wise fashion. To the best some of the techniques provide much more efficiency. At
of our knowledge, this is the first time to apply the deep present video based analysis is efficient. In this paper, the
architecture model using auto encoders as building blocks method starts by detecting and tracking objects seen in the
to represent traffic flow features for prediction. In addition, monitored zone by a video camera. The second stage of the
experiments demonstrate that the proposed method for method consists in predicting for each tracked object the ideal
traffic flow prediction has a superior performance. [14] trajectory avoiding potential dangerous situations.
 REFERENCES

[1] Houssam Salmane, Louahdi Khoudour, and Yassine Ruichek, “A [8] Government’s IT Strategy Headquaters: New IT Reform Strategy, The
Video-Analysis-Based Railway–Road Safety System for Detecting Hazard Realization of Society, inWhich Everyone Can Benefit
Situations at Level Crossings,”IEEE Transactions on intelligent From IT, Anytime and Anywhere, Jan. 19 2006, 19.
transportation systems. [9] L. Khoudouret al., “PANsafer project: Towards a safer level crossing,”
[2] Nelson, “The UK approach to managing risk at passive level crossings,” in presented at the 11th Level Crossing Symposium, Tokyo, Japan, 2010.
Proc. 7th Int. Symp. RailRoad-Highway Grade [10] N. Fakhfakhet al., “Background subtraction and 3D localization of
Crossing Res.Safety, 2002. moving and stationary obstacles at level crossings,” in Proc.
[3] L. Khoudour, M. Ghazel, F. Boukour, M. Heddebaut, and M. El-Koursi, [11] S Tuohy, M Glavin, C Hughes, E Jones, M Trivedi, L Kilmartin.
“Towards safer level crossings: Existing recommendations, new applicable “Intra-vehicle networks:A review”IEEE Transactions on Intelligent
technologies and a proposed simulation model,” Eur. Transp. Res.Rev., vol. 1, Transportation Systems.
no. 1, pp. 35– 45, Mar. 2009. [12] B Tian, BT Morris, M Tang, Y Liu, Y Yao, C Gou, D Shen, S Tang.
[4]. J. E. A. Dias, G. A. S. Pereira, and R. M. Palhares, “Longitudinal model “Hierarchical and networked vehicle surveillance in ITS: a survey”
identification and velocity control of an autonomous car,” IEEE Transactions IEEE Transactions on Intelligent Transportation Systems.
on Intelligent Transportation Systems, vol. 16, April 2015. [13] J. J. Garcia et al., “Efficient multisensory barrier for obstacle detection
[5]. N. Fakhfakhet al., “Background subtraction and 3D localization of on railways,” IEEE Trans. Intell. Transp. Syst., vol. 11,
moving and stationary obstacles at level crossings,” in Proc. no. 3, pp. 702–713, Sep. 2010.
Int. Conf. ImageProcess. Theory, Tools Appl., Paris, France, 2010. [14] Yanjie Duan, Yisheng Lv, Fei-Yue Wang.,”Performance evaluation of
[6] Priolettiet al., “Part-based pedestrian detection and feature-based tracking the deep learning approach for traffic flow prediction at different times” IEEE
for driver assistance: Real-time, robust algorithms, and evaluation,” IEEE International Conference on 10-12 July 2016.
Trans. Intell. Transp. Syst., vol. 14, no. 3, pp. 72–74, Sep. 2013. [15] Meiqing Wu, Siew Kei Lam,T. Srikanthan.,“Nonparametric Technique
[7] Eshed Ohn-Bar and Mohan M. Trivedi.,”Can appearance patterns improve Based High-Speed Road Surface Detection”. IEEE Trans. Intelligent
pedestrian detection?”.IEEE Intelligent Vehicles Symposium, June 2015. Transportation Systems 16(2): 874-884 (2015).