Detection and Tracking Software Framework for Automotive Perception System Development

Prism™ AI

Go to Product Support
product image

Prism AI is a software framework that provides classification, object detection, and object tracking enabling perception engineers to quickly start integrating thermal cameras for advanced driver assistance systems (ADAS) and autonomous vehicle (AV) systems. Developers can use Prism AI as the primary perception software or as reference while taking advantage of camera calibration tools, along with visible-and-thermal fusion and advanced image processing capabilities that offer superior perception capability, especially for pedestrian and animal detection.

Compatible with Teledyne FLIR market-leading thermal cameras including Boson®, Tau® 2, and the FLIR ADK, Prism AI features camera-to-ECU authentication, a key requirement for automotive production deployment. Prism AI tools provide simplified data integration with Teledyne FLIR Conservator™ dataset development software as well as the industry’s largest thermal-and-visible training dataset. The combination of validated models with successful performance in NCAP-based automatic emergency braking (AEB) tests, AI tools, ONNX model plugins, and annotation services decreases development cost and shortens time to market for a thermal-enabled ADAS or AV systems.

  • Immediately start evaluating thermal sensor detection capabilities

    Reference software delivers baseline performance and lowers investment for thermal-camera-enabled perception testing

  • Leverage Teledyne FLIR perception ecosystem to shorten time to market

    Suite of development tools provide an open, “white-box” foundation for ADAS perception software with thermal infrared capability

  • Built for developers with the Teledyne FLIR brand promise

    Trusted global leader in thermal infrared sensing and solutions in automotive

Specifications
Software Features
Authentication
Camera to ECU key handshake with 256-bit challenge/response
Camera Interface
USB, GMSL1, GMSL2, and FPD-Link III
Dewarping
Radial and tangential lens dewarping
Imaging Blending
Multi-spectrum blended image for display that automatically adjusts blending factor for day and night
Multi-Object Tracker
Lightweight, signal-processing based tracker reduces false positives and maintains target lock between frames in the absence of detections. Image-plane and ground-plane configurations available.
Object Detector (CNN)
7 visible and thermal object detection categories: Person, car, large vehicle (bus, semi-truck, train, etc.), bicycle, motorcycle, traffic light, and large animal (deer, cow, horse)
Rectification
Visible & thermal camera calibration process with pixel-accurate registration of visible & thermal
Sensor Fusion
Lightweight late-fusion implementation for object perception from thermal and color information from visible (brake lights, traffic lights, street signs)
Supported Hardware
X86 Desktop (CUDA 6.1 (Pascal), 7.0 (Volta), 7.5 (Turing), 8.6 (Ampere)), Nvidia Jetson Xavier AGX, Nvidia Jetson Xavier NX, Ambarella CV series, and Qualcomm RB5
Thermal Image Processing
Automatic Gain Control (AGC), super-resolution, denoising, and sharpening
Media Gallery
Related Documents
Export Restrictions

Export Restrictions

The information contained in this page pertains to products that may be subject to the International Traffic in Arms Regulations (ITAR) (22 C.F.R. Sections 120-130) or the Export Administration Regulations (EAR) (15 C.F.R. Sections 730-774) depending upon specifications for the final product; jurisdiction and classification will be provided upon request.

Prism™ AI