Project number
27062
Organization
ACABI
Offering
ENGR498-F2026-S2027
Project Goal/Summary: Advance last year’s CAMDAR prototype into an integrated, working system that uses two video cameras to identify and localize an at-risk individual in three-dimensional space, directs an existing alt-azimuth radar mount toward that person, measures non-contact vital signs with updated millimeter-wave radar, and displays the person, posture, location, vital signs and alert status for a responder.
Project Background: Last year’s team established the core CAMDAR architecture, built the camera/computer-vision pipeline and a functional aimable altitude–azimuth radar platform. The major unfinished tasks were reliable two-camera stereovision/3D localization, robust radar vital-sign acquisition, and complete subsystem integration. Their final report specifically recommends improved stereovision calibration/depth estimation, radar processing, mount stability/pointing accuracy, and earlier incremental integration. This year is an ADVANCE + INTEGRATE project—not a restart.
Requirements:
• Step 1 – Inherit + baseline. Begin with last year’s hardware, code, camera subsystem, Jetson/main console and alt-azimuth mount. Reproduce known functions and define quantitative baseline tests before modifying the system.
• Step 2 – Build true 3D vision. Integrate two synchronized/calibrated cameras with overlapping fields of view. Detect/track multiple people, assess posture (lean/slump/fall/collapse), and generate stable real-time X-Y-Z coordinates for the selected person of interest.
• Step 3 – Coordinate transform + aiming. Convert camera-derived 3D coordinates into radar azimuth/elevation commands. Reuse the existing turret design; refine rigidity, calibration, repeatability and pointing accuracy rather than redesigning from scratch.
• Step 4 – Upgrade mmWave radar. Integrate the identified newer millimeter-wave radar board(s) and establish reliable non-contact respiratory-rate and heart-rate measurement. Temperature may be added if supported by a separate validated sensor; do not make it a core radar requirement.
• Step 5 – Close the autonomous loop. VISION → PERSON-OF-INTEREST → 3D LOCATION → AIM → RADAR ACQUISITION → VITAL-SIGN ANALYSIS. Demonstrate repeatable handoff from vision to radar without manual aiming.
• Step 6 – Build the CAMDAR display. Create a clear real-time GUI/heads-up display showing the monitored scene, highlighted person of interest, posture/risk state, 3D position, radar targeting status, respiratory rate, heart rate, confidence/quality indicators and system state.
• Step 7 – Alert + validate. Generate a responder alert when predefined visual + physiologic criteria are met. Validate with staged non-clinical scenarios involving multiple people; quantify localization error, aiming error, vital-sign accuracy versus reference devices, detection/alert latency, false alerts and end-to-end reliability.
Project Background: Last year’s team established the core CAMDAR architecture, built the camera/computer-vision pipeline and a functional aimable altitude–azimuth radar platform. The major unfinished tasks were reliable two-camera stereovision/3D localization, robust radar vital-sign acquisition, and complete subsystem integration. Their final report specifically recommends improved stereovision calibration/depth estimation, radar processing, mount stability/pointing accuracy, and earlier incremental integration. This year is an ADVANCE + INTEGRATE project—not a restart.
Requirements:
• Step 1 – Inherit + baseline. Begin with last year’s hardware, code, camera subsystem, Jetson/main console and alt-azimuth mount. Reproduce known functions and define quantitative baseline tests before modifying the system.
• Step 2 – Build true 3D vision. Integrate two synchronized/calibrated cameras with overlapping fields of view. Detect/track multiple people, assess posture (lean/slump/fall/collapse), and generate stable real-time X-Y-Z coordinates for the selected person of interest.
• Step 3 – Coordinate transform + aiming. Convert camera-derived 3D coordinates into radar azimuth/elevation commands. Reuse the existing turret design; refine rigidity, calibration, repeatability and pointing accuracy rather than redesigning from scratch.
• Step 4 – Upgrade mmWave radar. Integrate the identified newer millimeter-wave radar board(s) and establish reliable non-contact respiratory-rate and heart-rate measurement. Temperature may be added if supported by a separate validated sensor; do not make it a core radar requirement.
• Step 5 – Close the autonomous loop. VISION → PERSON-OF-INTEREST → 3D LOCATION → AIM → RADAR ACQUISITION → VITAL-SIGN ANALYSIS. Demonstrate repeatable handoff from vision to radar without manual aiming.
• Step 6 – Build the CAMDAR display. Create a clear real-time GUI/heads-up display showing the monitored scene, highlighted person of interest, posture/risk state, 3D position, radar targeting status, respiratory rate, heart rate, confidence/quality indicators and system state.
• Step 7 – Alert + validate. Generate a responder alert when predefined visual + physiologic criteria are met. Validate with staged non-clinical scenarios involving multiple people; quantify localization error, aiming error, vital-sign accuracy versus reference devices, detection/alert latency, false alerts and end-to-end reliability.