China
Africa
Explore chapters and articles related to this topic
Hardware
Published in Hanky Sjafrie, Introduction to Self-Driving Vehicle Technology, 2019
One of the key properties of stereo cameras is the ability to carry out correspondence search, which involves searching for similarities in different sensors to create a continuous image of the surroundings. Various algorithms can be used to do this. Area-based algorithms consider a small area of one image and look for a similar area in another image. In contrast, feature-based algorithms identify unique features in each image in order to match up common points. Rather than computing an area, the calculations can be drawn from much smaller identifiable components of the image, including edges, corners, and lines. Epipolar geometry can be used as a basis for reducing the complexity of correspondence searches, as long as there are several pixels in one image that correspond to pixels in the other image. Correspondence can be generated using epipolar lines, a process in which an object in one camera view is projected onto the image plane of the second camera. This feature-based technique allows the system to identify similarities, and combine the images. A similar principle is used to generate a surround view (360 degrees) from four or more fish-eye cameras with overlapped horizontal fields of view (FOV).
Sensors for Autonomous Vehicles in Infrastructure Inspection Applications
Published in Diego Galar, Uday Kumar, Dammika Seneviratne, Robots, Drones, UAVs and UGVs for Operation and Maintenance, 2020
Diego Galar, Uday Kumar, Dammika Seneviratne
Common types of cameras used on AVRPs include regular color cameras, monochrome cameras, and stereo cameras. Monochrome cameras produce images in a single hue. This allows them to capture the actual light intensity values for each pixel, rather than using a color filter array (CFA) like a color camera does. In turn, this provides a better gradient, which can allow monochrome images to be better at tasks such as edge detection. Stereo cameras emulate human vision by using multiple cameras. This gives the platform the ability to create 3D images that can be processed to determine depth.
Automated driving technologies
Published in Tom Denton, Automated Driving and Driver Assistance Systems, 2019
Front mounted cameras are relatively cheap, but like all cameras, they are limited by environment conditions. Front cameras are typically mounted between the rear mirror and the windscreen; they are often part of the mirror assembly. The screen protects the camera and is cleaned by the washers and wipers. Front cameras can also be installed inside the vehicle between the dashboard and the screen, outside on the bumper or at the centre of the front roof edge. Stereo cameras are usually used because they provide distance estimation.
Virtual training and commissioning of industrial bin picking systems using synthetic sensor data and simulation
Published in International Journal of Computer Integrated Manufacturing, 2022
Maximilian Metzner, Felix Albrecht, Michael Fiegert, Bastian Bauer, Susanne Martin, Engin Karlidag, Andreas Blank, Jörg Franke
Metzner et al. (2019) propose a concept for the simulation of stereo vision-based bin picking systems by implementing a virtual stereo camera and using stereo matching algorithms as in real stereo cameras. Complex shader-based rendering allows for the mapping of effects caused by materials and surfaces or lighting. When applied to this synthetic data, post-processing and feature-based object detection and pose estimation algorithms produce similar results compared to real sensor data. Simulation is also carried out in a physics-simulation environment. No synthetic training data generation for learning-based methods is conducted. (Metzner et al. 2019)