Constantin Traistaru Autonom ous Vehicle Prof.dr.ing. Csaba Antonya Stereo Vision Introduction This research describes a stereo vision system for use… [610474]

Constantin Traistaru Autonom ous Vehicle Prof.dr.ing. Csaba Antonya

Stereo Vision

Introduction
This research describes a stereo vision system for use by a n autonomous
vehicle through a computer controlled. It should can move through a vary
environment, avoid obstacles, navigate to desired locations and build a description of
its environment. The computer vision is one of the most difficult probl ems in Artificial
Intelligence (AI) that has been challenging researche rs and engineers for a long time.
What do we know about computer stereo vision? Based on now adays the
(CSV) is transformation from a 3D information based on digital i mages, such as
those ob tained by a CCD camera. Comparing information s about a scene from two
vantage points, 3D information can be extracted by examining the relative positions
of objects in the two panels. This is the same to the biological process Stereopsis .
Stereoscopic images are often stored as MPO (multi picture object) files. Recently,
researchers pushed to develop methods to reduce the storage needed for these files
in order to maintain the high quality of the stereo im age.

Autonomous vehicle technology is a popular top ic that could increase vehicle safety
and convenience. Today, autonomous cars are tested with multiple sensors
including lidar, radar, and cameras. Vehicle perception and localization today relies
on expensive sensors like a Velodyne that costs as much as $30k; however, humans
are able to use the equivalent of one or two cameras as enough information to drive
on the road. In order for autonomous vehicles to become realistic for the general
public, there needs to be a more cost effective way to accomplish ob ject recognition
and state estimation.

Autonomous vehicle technology is a popular top ic that could increase vehicle safety and
convenience. Today, autonomous cars are tested with multiple sensors including lidar, radar, and
cameras. Vehicle perception and localization today relies on expensive sensors like a Velodyne
that costs as much as $30k; however, humans are able to use the equivalent of one or two cameras
as enough information to drive on the road. In order for autonomous vehicles to become realistic
for the general public, there needs to be a more cost effective way to accomplish ob ject
recognition and state estimation.
Computer stereo visio n is the extraction of 3D information from digital images, such as those
obtained by a CCD camera . By comparing information about a scene from two vantage points, 3D
information can be extracted by examining the relative positions of objects in the two panels. This is
similar to the biological process Stereopsis . Stereoscopic images are often stored as MPO (multi

Constantin Traistaru Autonom ous Vehicle Prof.dr.ing. Csaba Antonya
picture object) files. Recently, researchers pushed to develop methods to reduce the storage needed
for these files in order to maintain the high quality of the stereo im age.

This problem of extracting useful information
from images and videos finds application in a variety
of fields such as robotics, remote sensing, virtual reality,
industrial automation, etc. The concept of making cars drive
by themselv es has gained immense popularity today in the
AI world, mainly motivated by the number of accidents that
occur due to driver errors/ negligence. Autonomous vehicles
are designed to sense their surroundings with techniques such
as RADAR, LIDAR, GPS and comp uter vision. This array
of sensors working coherently to observe and record the
surroundings constitute the perception module of the vehicle.
The next stage in the pipeline is the localization step which
stitches together the incomplete and disconnected in formation
obtained from the sensors to identify the position, velocity
and other states of the vehicle and the obstacles (including
dynamic obstacles). The final module is the planning stage
where the vehicle has to decide what it has to do given the
situation it is in. The present day research prototypes built
by major players in the industry/ academia have LIDAR and
RADAR as their primary perception systems. They generally
provide a very accurate full 360 _ view to the vehicle making
it more informed about the environment than a normal human
driver. The downside to these systems is the cost involved in
deploying them. So an option is to use cameras and computer
vision techniques to substitute these systems.
Autonomous vehicle technology is a popular top ic that could increase vehicle safety and
convenience. Today, autonomous cars are tested with multiple sensors including lidar, radar, and
cameras. Vehicle perception and localization today relies on expensive sensors like a Velodyne
that costs as much as $30k; however, humans are able to use the equivalent of one or two cameras
as enough information to drive on the road. In order for autonomous vehicles to become realistic
for the general public, there needs to be a more cost effective way to accomplish ob ject
recognition and state estimation.
We propose a way to use “smarter” stereo cameras for both recognition and localization. One
way to demonstrate this is by recognizing stop signs and the velocity of the vehicle. An
autonomous car using the vision algo rithms of this project could know where a stop sign is and
how quickly it needs to brake. This technology could also be extended to knowing how close an
autonomous car should follow in traffic. By recognizing general features of the back of a car, an
auton omous car can judge how closely it is following the car ahead. Combined with estimation
in speed, an autonomous car can judge whether it ’s too close to the car ahead in traffic. This and
other scenarios can be eventually accomplished using stereo cameras w ith smart enough
algorithms.
Algorithm:
This paper describes a stereo vision system for use by a
computer -controlled vehicle which can move through a cluttered
environment, avoid obstacles, navigate to desired locations, and
build a description of its environment. One possible application

Constantin Traistaru Autonom ous Vehicle Prof.dr.ing. Csaba Antonya
of such a vehicle is in planetary exploration. Our experimental
vehicle is described in [41
As the vehicle moves about, it takes stereo picture pairs
from various locations This could be done with two cameras
mounted on the vehicle, but with our present vehicle with one
camera, it is done with the vehicle at two locations. Each of
these stereo pairs is processed to extract the needed
three -dimensional information, and then this information fro m
different pairs can be combined in further processing.
The processing of the stereo pairs is done as follows.
First, an interest operator finds small features with high
information content in the first picture. Then, a binary search
correlator finds the corresponding points in the other picture.
(The interest operator and the binary search correlator were
both developed by Moravec [41) Next, a high -resolution
correlator is given these matched pairs of points. It tries to
improve the accuracy of the match, and it produces an accuracy
estimate in the form of a two -by-two covariance matrix, and a
probability estimate giving the goodness of the match. The
coordinates of these matched points are corrected for camera
distortion as described by Moravec [41 A ster eo camera model
solver then uses these matched pairs of points to find the five
angles that relate the position and orientation of the two camera
locations. The accuracy estimates are used by the camera model
solver to weight the individual points in the s olution and to
compute accuracy estimates of the resulting camera model. A
dense sampling of points is now matched over the pictures. The
known camera model is used to restrict the search for these
matches to one dimension, and by first trying matches
approximately the same as neighboring points that have already
been matched, often no search is needed. In any case, the
precise matches are produced by the high -resolution correlator,
and its probability estimates are used in guiding the search.