5
IoT-Enabled Vision System for Detection of Tint Level
as image color, image position, image size, image type, and origin. The images are initially
preprocessed. The preprocessing includes filters, brightness and contrast equivalization,
and image conversion. Using adaptive background subtraction method, the foreground
object (moving vehicle) is only extracted. From this processed image, the region of interest
(ROI), namely windshield/window region, is extracted, and then this ROI is cropped from
the original color image and is passed on to the tint level calculation module. Three different techniques of tint detection were employed in the present work. If the tint level exceeds
the prescribed limits, then the process switches over to license plate detection to uniquely
identify the vehicle, and a ticket is issued, if necessary.
Figure 1.4 shows the operational environment of the proposed system in which the camera is placed at a certain height h from the ground level. The height is generally about 8–10
feet above the ground level so that the windshield/window can be easily visible with
desirable coverage area that can also easily locate the number plate of the vehicle, if necessary. While capturing the image, the vehicle is in a position such that “r” is the shortest
distance from the base of the pole on which the camera is mounted on the roadside and
the front end of the vehicle. Here a and b are the horizontal and vertical components of this
distance r, respectively. The camera inclination position should generally be not more than
30° with respect to the windshield/window level so that the internal view of the captured
image (vehicle internal view) will be good.
Figure 1.5 shows the methodology used for finding the distance of the vehicle from the
camera which in turn can be used for classifying the vehicle. For this purpose, the ground
region is segmented into various imaginary grid lines both along x- and y-axes (x 1 , x 2 ,…
and y 1 , y 2 , y 3 ,…) and is overlaid on the captured image. Assume that the vehicle is present
Input
Background
image
Background
subtraction
Object
detection
Obtain region
of interest
Identifying
tinting level
of glass
License
plate
detection
Foreground
image
Process
O utput
FIGURE 1.3
Overview of the proposed windshield tint level detection system based on real-time vision.
a
r
h
b
FIGURE 1.4
The operational environment of the proposed system.
IoT-Enabled Vision System for Detection of Tint Level
as image color, image position, image size, image type, and origin. The images are initially
preprocessed. The preprocessing includes filters, brightness and contrast equivalization,
and image conversion. Using adaptive background subtraction method, the foreground
object (moving vehicle) is only extracted. From this processed image, the region of interest
(ROI), namely windshield/window region, is extracted, and then this ROI is cropped from
the original color image and is passed on to the tint level calculation module. Three different techniques of tint detection were employed in the present work. If the tint level exceeds
the prescribed limits, then the process switches over to license plate detection to uniquely
identify the vehicle, and a ticket is issued, if necessary.
Figure 1.4 shows the operational environment of the proposed system in which the camera is placed at a certain height h from the ground level. The height is generally about 8–10
feet above the ground level so that the windshield/window can be easily visible with
desirable coverage area that can also easily locate the number plate of the vehicle, if necessary. While capturing the image, the vehicle is in a position such that “r” is the shortest
distance from the base of the pole on which the camera is mounted on the roadside and
the front end of the vehicle. Here a and b are the horizontal and vertical components of this
distance r, respectively. The camera inclination position should generally be not more than
30° with respect to the windshield/window level so that the internal view of the captured
image (vehicle internal view) will be good.
Figure 1.5 shows the methodology used for finding the distance of the vehicle from the
camera which in turn can be used for classifying the vehicle. For this purpose, the ground
region is segmented into various imaginary grid lines both along x- and y-axes (x 1 , x 2 ,…
and y 1 , y 2 , y 3 ,…) and is overlaid on the captured image. Assume that the vehicle is present
Input
Background
image
Background
subtraction
Object
detection
Obtain region
of interest
Identifying
tinting level
of glass
License
plate
detection
Foreground
image
Process
O utput
FIGURE 1.3
Overview of the proposed windshield tint level detection system based on real-time vision.
a
r
h
b
FIGURE 1.4
The operational environment of the proposed system.