Postgraduate. Ibadov S.R.
Don State Technical
University, Shakhty, Russian Federation
Postgraduate. Ibadov R.R.
Southern Federal University,
Taganrog, Russian Federation
Ph.D., Associate Professor. Dmitrienko N. A.
Don State Technical
University, Shakhty, Russian Federation
METHODS FOR DETECTION OF TRAFFIC OFFINCE NOT THE
ADMISSION OF PEDESTRIANS ON THE ZEBRA THE VEHICLE ON THE VIDEO SEQUENCE
Abstract – The article describes the method developed for
the determination of traffic violations, namely not the admission of
pedestrians the vehicle on a zebra. Work of a method is shown on the test video
sequence.
Keywords: detecting, video sequence, segmentation.
Introduction
Today means of automatic fixing of traffic
offenses work practically in all regions of Russia. In regions functioning
Centre for automated fixing administrative offenses, in which accumulated data
on violations of traffic rules, committed by drivers and recorded operating in
automatic mode complexes. Here are issued rulings on the cases of
administrative offenses, which then together with penal receipts go to owners
of vehicles.
Detecting of a zebra
At the first stage we need to detect a zebra,
the video sequence is for this purpose loaded. [1] Further, there is a
splitting the video sequence into frame and allocation of an object and
background by means of tags by the user. Stages of the offered approach are
presented in the Figure 1.
Figure 1 - Stages of segmentation
To do this, first select the object frame
is marked with white background, the background is marked in black color the
figure 2a, in the figure 2b an alpha the channel of the image received by means
of a segmentation method on the received tags at the first stage is presented. In the figure 2c
show the selected object on the original frame of the video sequence.
a)
b) c)
Figure 2 – stages of segmentation of a
zebra:
a) marking of an object and background on
the first shot, b) an alpha the channel of the image, c) the allocated object
Detecting of the vehicle
using Gaussian Mixture Model
Step 1 – Load the video sequence Figure 3a and initialize the foreground
figure 3b. Instead of at once processing all video sequence, processing begins
with receiving an initial shot of video, in which moving objects are segmented
from a background. It helps to enter gradually the steps used for processing of
all video sequence.
The detector of the foreground demands a certain number of the video
footage for initialization of model of Gauss. [2] In our example the first 50
frames for initialization of three Gaussian fashion in this model are used.
After the training, the detector will begin to remove more reliable
results of segmentation.
a)
b)
Figure
3 – detecting of the foreground of the video sequence:
a)
first shot of the video sequence, b) foreground
Step 2 - Detection
of vehicles on the first shot of the video sequence.
Into the forefront process of segmentation isn't perfect and often
includes undesirable noise. In this example morphological approach is used to
remove noise and to fill in blanks in the found objects the figure 4a.
Further, we find the limiting squares of each connected component
corresponding to vehicles by means of an object vision.BlobAnalysis
figure 4b. Object in addition filters the found foreground, rejecting clots
which contain less than 150 pixels.
a) b)
Figure
4 – detecting of the vehicle:
a)
the pure foreground, b) the detected vehicles
Step 3 – Processing of the remained video footage
At the final stage, we process the remained video footage. Then we need
to detect pedestrians on all frames of the video sequence.
Detecting of pedestrians
We detect pedestrians by means of Motion-Based Multiple Object Tracking
of a method.
Detecting of moving objects and tracking of their movement is an
important element of computer sight. [3]
Detection of moving objects uses the algorithm of subtraction of a background
based on Gaussian mixture of model. Morphological operations are applied to a
resultant mask of the foreground the figure 5a to elimination of noise. Finally, the analysis reveals a group of related
pixels that are likely to correspond to a moving object, the figure 5b.
The association of detection besides to an object is based only on the
movement. The movement of each traced object is estimated by means of Kallman's
filter. The filter is used for forecasting of tracking of provision of objects
in each shot.
a) b)
Figure
5 – detecting of the pedestrian:
a) the foreground, b) the detected
pedestrian
Kalman filter is used to predict the midpoint of each of the following
structural frame and updates the bounding rectangle respectively.
Comparison of shots
The second stage of an estimated method of definition
not of the admission by the vehicle of the pedestrian on a zebra consists in
comparison of shots with the allocated objects. To reveal violation not of the
admission of the pedestrian on a zebra it is necessary to impose video sequence
shots with the allocated objects at each other. If on the same shot of the
video sequence the pedestrian and the vehicle cross a zebra, the vehicle means
violated traffic regulations, namely not the admission of the pedestrian on a
zebra.
For example, when comparing the 130th shot of the
video sequence with the allocated vehicle the figure 6a the 1st shot of the
video sequence with the allocated zebra the figure 6b we see that the vehicle
crossed a zebra the figure 6c.
Figure 6 – Crossing by the vehicle of a zebra.
a) the
allocated vehicle, b) the allocated zebra, c) crossing of a zebra by the
vehicle
Further we check for crossing the 130th frame of the
video sequence with the allocated pedestrian the figure 7a, with the 1st frame
of the video sequence with the allocated zebra the figure 7b we see that the
pedestrian crossed a zebra the figure 7c.
a) b) c)
Figure 7 – Crossing by the zebra pedestrian:
a) the
allocated pedestrian, b) the allocated zebra, c) crossing of a zebra by the
pedestrian
Therefore, the vehicle violated traffic regulations,
without having missed the pedestrian on the 130th frame of the video sequence.
Bibliography
1.
Voronin V.V., Gapon N.V., Sizyakin R.A.,
Ibadov R.R., Ibadov S.R., Semenishchev E.A. Research video recovery method
based on the spatio-temporal processing. Innovation, ecology and
resource-saving technologies (InERT 2014) [electronic resource]: Proceedings of
the XI International scientific-technical forum / DSTU; ed. A.D. Lukyanov -
Rostov n / D: DSTU, 2014. pp 1360-1366.
2.
Detecting
Cars Using Gaussian Mixture Models
https://www.mathworks.com/help/vision/examples/detecting-cars-using-gaussian-mixture-models.html?requestedDomain=www.mathworks.com
3.
Motion-Based
Multiple Object Tracking
https://www.mathworks.com/help/vision/examples/motion-based-multiple-object-tracking.html