Written & Interviewed by : LITEON Editorial Team (Corporate Brand Value Development Center)
Technical review :LITEON Center of Core Competence
A vehicle can detect lanes, nearby traffic, and obstacles but miss a risk within arm’s reach. That is, a distracted driver, an awkwardly seated passenger, or a child left behind. Road-facing advanced driver assistance systems (ADAS) need complementary info about the people inside.
In-Cabin Sensing uses cameras, radar, and AI to assess presence, position, behavior, posture, and selected physiological signals. Thanks to the usage of camera radar sensor fusion, complementary measurements support a more comprehensive understanding of cabin conditions. The direction is human-centered, including moving from evaluating a single driver to safeguarding everyone on board. Responses are according to the situation.
Human attention is increasingly being taken into consideration alongside road hazards in safety regulations. Crashes involving distracted drivers killed 3,208 people in the United States in 2024.
From 7 July 2024, EU rules extended driver drowsiness warnings and reversing detection through cameras or sensors to all new motor vehicles. Moreover, advanced driver distraction warnings followed for new passenger cars and vans from 7 July 2026.
Euro NCAP provides a separate incentive through consumer safety ratings. Its assessments include occupant protection functions beyond the legal baseline.
Meanwhile, higher levels of automated driving could give occupants more freedom to work or relax during permitted operation. A cabin becomes a living space, as well as the demands on an occupant monitoring system are changed. Whenever there is a need for a handover, it is very important to determine whether or not the driver is able to regain control.
A Driver Monitoring System (DMS) analyzes eye gaze, head direction, eyelid closure, and blinking in order to assess if the driver is fatigued, distracted, or ready to retake control of the vehicle. It’s best understood as a driver analysis and assessment system.
Context is important. For example, a brief glance toward a mirror and prolonged attention away from driving shouldn’t receive identical judgments. Along these lines, temporal patterns allow the system to assess behavior across successive observations.
Just open eyes don’t establish takeover readiness. During a requested handover, the information on the driver's status must be evaluated within the context of the control-transition strategy of the vehicle. This includes the amount of time that is available for the individual to resume driving.
An occupant monitoring system (OMS) extends coverage across seating rows. It determines which seats are occupied and assesses occupant location, posture, seat-belt status, occupant classification, and child presence. These inputs can support occupant protection functions and more adaptive restraint strategies. When an occupant monitoring system uses camera radar sensor fusion, it must associate observations with the correct person. Otherwise, movement from one passenger could be assigned to a neighboring seat. Whole-cabin understanding needs that spatial context.
These terms describe different scopes within a shared safety architecture.
|
What does an occupant monitoring system enable beyond identifying an occupied seat?
Further, Euro NCAP’s 2026 CPD criteria demand direct sensing and coverage of children through six years old. When it comes to a detected unattended child who has been locked inside, the protocol permits no more than 15 seconds from locking to the initial warning.
Camera sensing extracts facial landmarks, eye activity, head orientation, and body geometry from images. Such features provide the visual evidence available to camera radar sensor fusion.
LITEON also analyzes subtle optical changes that are associated with facial microvascular blood flow in order to estimate heart-rate-related signals. Basically, the principle is remote photoplethysmography. Here, pulsatile blood-volume changes affect reflected light. Such estimates provide additional state information, but accuracy depends on operating conditions. Lighting, viewing angle, and occlusion might be a few challenges. A 2023 university study evaluated camera-based pulse estimation across five datasets, including night driving, rain, and longer journeys. This shows the breadth of validation required. The study cut nighttime heart-rate estimation error (root mean square error) by up to 42.6% vs. its benchmark.
For an occupant monitoring system, frequent false alerts create a human problem too. People may stop trusting the warnings or cover the camera. That’s a poor outcome for any safety feature.
60 GHz millimeter-wave radar makes use of radio reflections for the purpose of estimating position and detecting small body movements without depending on visible light. Respiration produces chest motion that can indicate a living occupant. Academic work has demonstrated seat-occupancy and breathing-rate sensing using this frequency band.
At 60 GHz, the wavelength is approximately 5 millimeters. Changes in the returning signal allow suitable radar systems to resolve movements smaller than that wavelength. Consider a rear-facing child seat. The camera's direct view may be obscured by its shell, but the radar will continue to have a route that it may use to identify breathing-related motion. Selected implementations may further explore heartbeat-related micro-motion sensing.
There are a number of factors that influence performance, including antenna design, mounting position, cabin structure, seating arrangement, obstructions, as well as validated algorithms. Vehicle vibration and ordinary body movement can interfere with physiological estimates.
Camera radar sensor fusion improves robustness by leveraging the complementary strengths of cameras and radar. While cameras provide rich visual and behavioral information, radar can remain effective in low-light or partially occluded conditions. Together, they help maintain reliable occupant awareness across a wide range of cabin scenarios.
Effective camera radar sensor fusion needs synchronized observations and confidence estimates. If an image is unreliable, its influence should decrease. Contradictory measurements deserve further assessment. This method may help limit the number of false positives and unnecessary responses without assuming that either sensor is infallible. Apart from that, missing data also needs an explicit status. For instance, losing sight of a seat must never silently be considered as a conclusion that the seat is empty.
The trend in vehicle electronics is to move away from the pattern of each sensor having its own dedicated processor. While the computing capacity and vehicle programs may differ, the central computers and domain controllers may share resources across functions.
Accordingly, an occupant monitoring system should be adaptable to different Electrical/Electronic (E/E) architectures, including:
Provided that bandwidth, latency, functional safety, and fault-handling requirements are satisfied, camera-radar sensor fusion solutions can be deployed across these E/E architectures.
One representative processing flow is:
Camera and Radar → Hardware Noise Filtering → AI Computing → Sensor Fusion → State Assessment → Warning or Active Protection
Hardware filters remove high-frequency interference and unwanted electrical noise. AI Computing extracts the relevant features. At the same time, camera radar sensor fusion fuses the evidence for the subsequent state assessment. Implementations may order these stages differently.
Safety-critical In-Cabin Sensing decisions must be made internally. Edge Computing minimizes communication delays and dependency on cloud connectivity, and reduces the need to transmit sensitive images and physiological data. Controls of access and retention policies are still valid.
E.g., a controller may share an occupancy state and confidence value with another vehicle function and not retain the image. Data handling deserves attention during system design.
Responses may escalate from a fatigue warning, stimulating audio, or seat stimulation to an emergency call or a controlled emergency stop. Options for an unsupervised child include notification to the parents, ventilation, or air conditioning. Euro NCAP’s 2026 CPD criteria specify escalation warnings repeated at least every minute for a minimum of 20 minutes. Regarding remote notifications, detection may run locally but needs a communication path.
Actions depend on vehicle architecture, automation level, safety validation, and regulations. Sensing supplies evidence to authorized vehicle controllers that execute the response.
LITEON's Advanced Overhead Console (AOC) integrates camera sensing, 60 GHz radar, AI processing, and vehicle interfaces into a unified roof-mounted cabin sensing platform.
This architecture enables camera-radar sensor fusion at the system level, helping maintain reliable occupant awareness across diverse cabin conditions while simplifying vehicle integration. Its higher, elevated cabin-wide sensing perspective can extend coverage to front and rear occupants.
The usable coverage of an occupant monitoring system is dependent on camera optics, placement of the radar antenna, cabin dimensions, seat arrangement, and sensing angles. Installation on the roof offers an opportunity for greater visibility and each vehicle configuration must be validated.
LITEON's decades of experience in optics, imaging, lighting, electronics, mechanical design, and manufacturing allow sensing, illumination, thermal management, and vehicle interfaces to be engineered as a coordinated system.
That combination creates engineering challenges within a compact roof enclosure. Optical reflections from the trim, heat generated by the electronics and lighting, electromagnetic interference, and tolerances in assembly. Even the cover material can affect radar transmission or the optical path of the camera. These interactions must be considered in unison.
This packaging can result in fewer separate modules, integrated wiring, easier space planning and calibration, and a more streamlined approach to supplier coordination. Potential savings on a total system basis depend on the vehicle program and integration requirements.
Modular interfaces support adaptation across vehicle models, seating layouts, and centralized or distributed computing arrangements. LITEON thus becomes a technology and a system-integration partner of human-centered cabin safety.
Replacement and recalibration should also be part of service planning for an occupant monitoring system. A repaired module needs to be restored to its intended alignment and configuration before its safety functions can be relied upon.
“Sensing First” establishes the engineering priority. Inputs that may be relied upon must come before the judgment and protective action of AI. When determining acceptance, take into account missing events, the incidence of false alarms, and response timing across actual settings.
The In-Cabin Sensing will be developed further to include more physiological sensing, better AI integration, and connectivity with Mini LED, Micro LED, and interior lighting. LITEON's vision extends beyond individual sensing functions. Through its Advanced Overhead Console platform, LITEON aims to create a human-centered cabin architecture that brings sensing, intelligence, safety, and vehicle interaction together within a unified system.
So, contact LITEON to learn how Camera, 60 GHz Radar, and AI integration can help you reach your vehicle’s in-cabin sensing and occupant protection goals.