<p style="text-align: justify">Autonomous vehicles promise a safer future with a cleaner, more cost efficient, and more reliable transportation system. However, the current approach to autonomy for vehicles has focused on building small, expensive, disparate intelligences that are closed off to the rest of the world, where each vehicle uses its own suite of perception and navigation sensors and actuators&nbsp;to understand the surrounding environment, predict the behavior of other road users, and follow a safe trajectory.&nbsp;In this approach, even if several autonomous vehicles are traveling in the same environment at the same time, they each have to carry expensive sensing, navigation, and processing hardware and still, lacking coordination with other road users, they may get into accidents.&nbsp;Vehicle to Vehicle (V2V) and Vehicle to Infrastructure (V2I) connectivity can bridge some of that gap through enabling communication between the vehicle and its surrounding environment, enabling a a paradigm shift in communications and coordination, cooperative sensing, and real time dynamic planning and controls that can&nbsp;be effective in improving traffic congestion, road user safety, and overall efficiency. This has to be paired with a low-cost approach to perception to enable broader research into connected and autonomous vehicles (CAVs). This will eventually lead to the&nbsp;adoption of connected and autonomous vehicles in the broader soceity. Cameras are one of the most important elements of vehicle perception, because they&nbsp;enable road surface detection, lane line detection,&nbsp;object detection and tracking, and scene segmentation. However, current machine vision approaches used in the research community for autonomous driving are prohibitively expensive, preventing many scholars from conducting valuable research.&nbsp;To that end, this paper presents the development and calibration&nbsp;process of a low cost, easy to deploy machine vision pipeline for multiple cameras&nbsp;that does not sacrifice image quality or frame rate. Instead of using expensive machine vision cameras with on-board image signal processors (ISPs), this approach&nbsp;utilizes low-cost autonomotive cameras that send image data over the GMSL2 protocol to a single Jetson device for processing and compression using a GStreamer pipeline. The paper also discusses methods for calibrating the camera distortion, white balance, saturation, noise, and color accuracy.&nbsp;The paper also addresses some of the challenges encountered during the development of the&nbsp;pipeline.&nbsp;These challenges include&nbsp;synchronization of image frames to arrive at the main processing hardware at the same time,&nbsp;and integration of the pipeline with the CARMA Platform, developed by FHWA for connected vehicle research that enables cooperative autonomy.</p>

Development and Calibration of a Low-Cost Machine Vision Pipeline for Connected and Autonomous Vehicle (CAV) Research

Paper Proceedings: https://asmedigitalcollection.asme.org/IMECE/IMECE2021/volume/85543

ASME 2021 International Mechanical Engineering Congress and Exposition (IMECE2021)

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.