Obstacle avoidance with stereo vision in Self-driving cars

-
✔️Obstacle avoidance with stereo vision in Self-driving cars




✔️Motivation

Autonomous vehicles are the future of mobility and to realize this future, the costs associated with the current technology have to be reduced significantly. Most autonomous vehicles use lidar systems, with prices in the range of $60,000-$80,000, for perception. An alternative to lidar systems for perception are cameras and vision systems. Vision systems are comparatively inexpensive but are not nearly as accurate as lidar systems. As part of this project, we plan to explore the potential of vision systems to detect obstacles using stereo vision.

✔️Description

The plan is to implement a simple obstacle detection program and test it on an existing self driving car. The current self driving car we plan to do our tests on is X1, Stanford’s student built drive by wire, steer by wire test vehicle. The two major stages that constitute the technology of an autonomous vehicle are 1)Perception and 2) Planning. Perception is the stage where the car observes its surroundings and makes a map of all the neighbouring obstacles relative to its location. Planning is the stage where the car decides on what to do once it knows everything about its environment. Currently a sampling based motion planning algorithm FMT* has been developed and tested on X1 and thus the planning part of the car is already implemented. We intend to add a perception framework on X1 that utilizes stereo-vision to complete the loop. The project will involve installing cameras on the X1 and using stereo vision techniques as done in [1] and [2] that employ the cameras to create depth maps and extract 3D information about the surroundings. We plan to use two 360 degree cameras as done in [3] to replace the usual lidar. In this way we are not restricted to mapping only the direction in which the cameras are pointed . This also enables the car to do more complex manoeuvres such as parallel parking or backing up once it has a more detailed knowledge about all its surroundings.

Planned demonstrations:
• Obstacle avoidance by changing lanes
• Path following
• Parallel parking and other complex low speed manoeuvres.

✔️Implementation

Hardware:
Ricoh theta S spherical digital cameras may be used to obtain 360 degree images. An alternative would be to use multiple cameras which cover the entire 360 degree field of vision.

Software:
We plan to develop our algorithms in python or matlab and may use openCV. We will not be developing applications on android.

✔️References

[1] Massimo Bertozzi and Alberto Broggi. Gold: A parallel real-time stereo vision system for generic obstacle
and lane detection. Image Processing, IEEE Transactions on, 7(1):62–81, 1998.
[2] Reinhard Klette. Stereo-vision-support for intelligent vehicles-the need for quantified evidence. Springer,2008.
[3] Hansung Kim and Adrian Hilton. 3d scene reconstruction from multiple spherical stereo pairs. International journal of computer vision, 104(1):94–116, 2013.

✔️PROJECT SOURCE CODE CLICK HERE

Comments

Post a Comment

Popular posts from this blog

Light Field Images for Background Removal

-
Image
Light Field Images for Background Removal OUTPUT OVERVIEW Standard edge detection or foreground/background separation techniques, such as Otzu’smethod, require color or intensity differences between the background and regions that need tobe separated. For example, green screens are routinely set up as the background in a scene so thatthere is a clear difference in color between the background and foreground.Light filed images, captured in 4D and passively containing depth information for thescene can be used to approximate this effect. Depth estimation in the scene alone could provide a metric to separate the foreground and background, but more sophisticated methods are available.Considering the edge detection from a single image from a single viewpoint and analyzing the depths from the light field around the edges in that image, occluded edges and intensity or color edges can be distinguished. Along with the rest of the depth information this can allow the foreground and b
Read more

Using Image Processing to Identify and Score Darts thrown into a Dartboard

-
Image
✔️Using Image Processing to Identify and Score Darts thrown into a Dartboard ✔️Introduction The game of darts is a throwing sport in which participants toss projectiles into a circular target  attached to a vertical surface. The target is divided into many regions which correspond to different point and multiplier values. Darts is commonly played across North America and Europe and is a popular past time for many so an application that allows players to more conveniently keep score would be widely  beneficial. “Darts” is a general term for a targeting game following this basic premise and many game variations exist within this archetype; all utilizing the standard dart projectiles with a regulation dart board. Each game variant may have different objectives for which players to aim but identifying the region in  which the dart has hit in the dartboard is necessary for proper scoring. This project proposes the use of image processing to identify thrown darts and to det
Read more

Digital Make up Face Generation

-
Image
Digital Make up Face Generation Goal Current make-up applications rely on using photoshop tools to apply makeup on the target's digital  faces and generate results. While these applications allow customization, a customer who wants to  quickly decide on the type of makeup kit to buy at a store will not find it useful. The customer might just  want to find out how the make-up look on the cover or billboard will look like on her face. The goal of  this project is to use an existing reference image of another subject with a make-up applied, and  transfer the reference's make up on the target's face. The application can be further extended to photo  retouching and illumination transfer from the reference image to the target. Methodology In order to transfer the make-up from the reference onto target in a pixel by pixel basis, the areas of  interest must align. Face features such as eyes, nose, mouth and contours of the face, will be recognized  using Active Shape
Read more