How to get a relevant camera quality evaluation is becoming critical in the Automotive industry with the advent of Autonomous driving. This tutorial will give the keys to getting a comprehensive evaluation of the hardware information capacity and also ISP & tuning performances according to the final usage of the camera whether it is design for computer vision or human vision.

The tutorial will be split into two major parts.

Since a good camera image requires a good camera module, in the first part, the principle of evaluation and validation of the camera sensor and lens characteristics will be explained. The camera model from light to RAW pixel values will be used to demonstrate the reliability of the measurement of the lens and sensor system/hardware design.

In the second part of the tutorial, we will see how important it is to relate the image quality evaluation to the final use case. With this in mind, we will introduce our methodology to establish a comprehensive camera testing protocol based on laboratory tests and perceptual analysis in driving conditions.

What will participants learn?

• Overall camera evaluation principle,
• Difference between RAW and RGB image analysis,
• How to validate the specifications of lens and sensor from RAW images,
• Introduction of a sample automotive camera module evaluation report,
• Study of the correlation between camera module evaluation and detection probability
• How to define the right evaluation protocol to test the performances of an ISP and its optimization
• Results of the benchmark of 2 ISP for Automotive industry on a front viewing ADAS camera

The semiconductor foundry plays an important role in achieving the OEM’s stringent ADAS performance requirements, as it manufactures and supplies the ICs used in such systems. This tutorial will review the radar system requirements and introduce the audience to what it takes to build a differentiated automotive radar chip from a foundry perspective. We will discuss requirements for the radar chip and its building blocks: transmitter, receiver, and frequency synthesizer. Furthermore, we will use three foundational circuit blocks as a case study: power amplifier, low noise amplifier, and voltage/digital-controlled oscillator. After the tutorial, the audience will have a better understanding of how foundry can help improve circuit level performance for auto radar application.

Dr McManamon will cover an introduction to lidar, its history, and the types of lidar.  He will cover atmospheric effects on lidar, lidar range equation, signal-to-noise ratio, and basic detection theory.  He will then cover major components in a lidar, the laser source, the receiver possibilities, and the methods of steering the transmitted and receive optical axis.  There will be a brief discussion of lidar processing, the testing of lidars, and lidar performance metrics, followed by lidar application design examples, including autonomous vehicle applications.

Enabled by machine learning and artificial intelligence, sensor fusion is poised to play an increasingly important role in ADAS and autonomous vehicle perception and safety systems. This tutorial will provide an overview the basics of sensor fusion and the different levels at which it can be performed. It will also explore benefits and opportunities for sensor fusion as well as risks involved. A variety of multi-modal sensor fusion examples will be reviewed that use combinations of color video, lidar and Doppler radar.

HD Maps are a critical part of the AV sensor stack. This presentation explores how HD maps are being used to power a variety of use cases, and across various SAE levels of autonomy in R&D and production, to help autonomous vehicles make safer and more pro-active driving decisions.