IMAGE PROCESSING
APPLIED TO TRAFFIC QUEUE
DETECTION ALGORITHM
INTRODUCTION
Digital processing is done with a digital computer
or some special purpose digital hardware.
The word Digital implies that the information in
the computer is represented and sent by
variables that taken limited number of discrete
values.
Increasing demand for road traffic data of all
sorts.
Variation of parameters in real-world traffic.
Aimed to measure queue parameters accurately.
Queue detection algorithm has two operations :
• vehicle detection
• motion detection.
Motion detection is based on applying a
differencing technique on the profiles of the
images along the road.
The vehicle detection is based on applying edge
detection on these profiles.
DIGITAL IMAGE PROCESSING:
IMAGE PROCESSING APPLIED TO TRAFFIC:
Need for processing of traffic data:
Traffic surveillance and control, traffic management, road safety
and development of transport policy.
Traffic parameters measurable:
Traffic volumes, Speed, Headways, Inter-vehicle gaps, Vehicle
classification, Origin and destination of traffic, Junction turning.
Image analysis system structure
Stages of image analysis:
Image sensors used:
Improved video cameras: automatic gain control, low SNR
ADC Conversion:
Analog video signal received from video camera is
converted to digital/binary form for processing
Pre-processing:
High SNR of the camera output reduces the quantity of
processing enormous data flow.
Two jobs to be done:
Green light on:
Determine no. of vehicles moving along particular
lanes and their classification by shape and size.
Red light on:
Determine the backup length along with the
possibility to track its dynamics and classify
vehicles in backup.
Queue Detection Algorithm
Approach described here is a spatial-domain technique
to detect queue - implemented in real-time using low-
cost system.
For this purpose two different algorithms have been
used,
• Motion detection operation
• Vehicle detection operation
Motion detection is first – as in this case vehicle
detection mostly gives positive result, while in reality,
there may not be any queue at all
•George Bush Drive and Wellborn Road in College Station, Texas
Motion detection operation:
Differencing two consecutive frames.
Histogram of the key region parts of the frames is
analyzed by comparing with the threshold value.
Key region should be at least 3-pixel-wide profile
of the image along the road.
.
A median filtering operation is firstly applied to
the key region (profile) of each frame and one-
pixel-wide profile is extracted.
Difference of two profiles is compared to detect
for motion.
When there is motion, the differences of the
profiles are larger than the case when there is no
motion. The motion can be detected by selecting a
threshold value.
Vehicle detection algorithm:
Following the application of the motion detection
operation, a vehicle detection operation is applied
on the profile of the unprocessed image.
To implement the algorithm in real time, two
strategies are often applied:
• Key region processing
• Simple algorithms.
Most of the vehicle detection algorithms
developed so far are based on a background
differencing technique, which is sensitive to
variations of ambient lighting.
The method used here is based on applying edge
detector operators to a profile of the image .
Edges are less sensitive to the variation of
ambient lighting and are used in full frame
applications (detection).
Methods of vehicle detection:
Background frame differencing:
Grey-value intensity reference image
Inter-frame differencing:
Incoming frame itself becomes the
background for the following frame
Segmentation and classification:
Sub division of an image into its
constituent parts depending on the context
Edge detectors consisting of separable medium filtering and
morphological operators, SMED (Separable Morphological
Edge Detector) are applied to the key regions of the image.
(The SMED approach is applied (f) to each sub-profile
of the image and the histogram of each sub-profile is
processed by selecting Dynamic left-limit value and a
threshold value to detect vehicles.
SMED has lower computational requirement while having
comparable performance to other morphological operators
SMED can detect edges at different angles, while other
morphological operators are unable to detect all kinds of
edges.
Canny edge detection operator:
Canny is the most famous edge detection operator.
It acts as either 1st order or 2nd order gradient operator.
Canny operator mainly consists of two operations.
• Non-maximal Suppression.
• Hysteresis Thresholding
In the Non-maximal Suppression, the non-maximal pixel values
in the image will suppress to zero and maximal pixel values in the
image will tend to a maximum value.
In the Hystersis thresholding, the detection of the threshold
will take place.
Image captured by camera Canny operator applied to the
image
This program selects a grey value from the
histogram of the window, where there are approx.
zero edge points above this grey value.
When the window contains an object, the left-limit
of the histogram shifts towards the maximum grey
value, otherwise it shifts towards the origin.
This process is repeated for a large no. of
frames(100),and the minimum of the left-limit of
these frames is selected as the left-limit for the
next frame.
Fig: histogram of the image
Fig: Image captured by camera
Threshold selection program:-
The no. of edge points greater than the left limit grey value of each window is extracted
for a large no. of frames (200) to get enough parameters below and above a proper
threshold value.
These nos. are used to create a histogram where its horizontal and vertical axes
correspond to the no. of edge points greater than left limit and the frequency of
repetition of these numbers for a period of operation of the algorithm (200 frames).
This histogram is smoothed using a median filter and we expect to get two peaks in the
resulted diagram, one peak related to the frames passing a vehicle and the other related
to the frames without vehicles for that window.
However, as it can be seen in the figure given below, there are other number of edge
points (32-40) between peaks 20 & 60, which are related to those vehicles in each
profile. We use statistical approach based on selecting a point on the horizontal axis,
where the sum of the entropy of the points above and below this point is maximum.
This point is selected as the threshold value for the next period.
Results and Discussions:
The main queue parameters we were
interested in identifying were the length of the
queue, the period of occurrence and the slope of
the occurrence of the queue behind the traffic
lights.
To implement the algorithm in real-time, it was
decided that the vehicle detection operation
should only be used in a sub-profile where we
expect the queue will be extended.
Any queries ?