Thermal imaging camera vehicle temperature monitoring and accident prediction

Abstract

A system consisting of at least one vehicle-mounted thermal imaging camera, a processing unit for analyzing thermal images and data from auxiliary sensors, and a communication interface, wherein the processing unit detects temperature anomalies on the vehicle and predicts the risk of accidents in order to generate real-time warnings to drivers and authorities.

Description

AREA OF INVENTION

[0001] The invention relates to vehicle safety devices that combine thermal imaging cameras, on-board sensors and machine learning to monitor vehicle temperatures, predict accident risks and issue real-time warnings to drivers and authorities. BACKGROUND OF THE INVENTION

[0002] Conventional driver assistance systems rely on cameras, radar, and lidar, but their performance deteriorates in low light conditions, smoke, fog, snow, and glare. This leads to detection gaps in safety-critical situations. Long-wave infrared thermal imaging cameras, on the other hand, maintain their performance even in adverse visibility conditions and can detect thermal anomalies that are not visible in RGB images. Thermal monitoring of vehicle subsystems can identify dangerous heat patterns, such as overheated brakes, wheel hubs, or batteries, thus enabling preventative intervention to avoid fires or failures. Deep learning methods are used for the automatic temperature measurement of components such as brake drums. Recent studies show that machine learning with thermal imaging supports early collision avoidance and reliable vehicle detection and tracking in adverse weather conditions.AI-powered driver condition monitoring and near-miss analysis improve the prediction of safety events when combined with multi-sensor data. A device-centric architecture that integrates thermal imaging cameras, temperature analysis, and predictive models to forecast accident probability and issue timely warnings addresses unmet needs in proactive traffic safety management for vehicle fleets and public roads. SUMMARY OF THE INVENTION

[0003] The invention relates to a system comprising one or more vehicle-mounted thermal imaging cameras, additional sensors, a processing unit for executing machine learning models, and a communication interface for providing early warnings to drivers and authorities. The cameras capture thermal images of the vehicle and surrounding traffic to monitor component temperatures and detect hazardous situations in poor visibility conditions. Models estimate the risk of an accident based on temperature deviations, motion data, and environmental information. The processing unit analyzes the data in real time to detect overheating of brakes, tires, or the powertrain and to predict the risk of collision by detecting thermal images and movements under adverse weather conditions.Warnings with location and severity information are issued via the user interface in the vehicle interior and secure telemetry data to enable proactive intervention and emergency measures, thus increasing safety. DETAILED DESCRIPTION

[0004] The device includes at least one long-wave infrared thermal imaging camera for monitoring the vehicle's wheel suspension and powertrain, as well as optional outward-facing cameras for capturing the road environment. It provides temperature maps that are robust against the light, smoke, fog, and snow conditions commonly encountered in accident scenarios. A processing unit acquires thermal images and additional data (speed, braking behavior, steering, ambient temperature), performs preprocessing steps such as inhomogeneity correction and emissivity compensation, and detects spatial hotspots and temperature gradients indicative of abnormal heating. Component-specific analyses include temperature estimation of brake drums and discs using deep learning.This method combines dual-band inputs or learned features to classify the severity of overheating and the remaining safe operating margins before performance degradation or failure. Scene recognition models trained on thermal imaging data detect vehicles and pedestrians and estimate their distances in adverse weather conditions. They support early collision avoidance by detecting heat silhouettes where visible sensors may not function optimally. Long-range detection and distance estimation using thermal images have already been demonstrated. A predictive module combines temperature anomalies with kinematic signals and contextual information to calculate accident risk. It learns from historical near misses and incident data to predict failures or accidents.The module prioritizes immediate risks such as brake overheating during downhill driving or tire thermal runaway. The human-machine interface (HMI) displays intuitive warnings with component temperatures, risk levels, and recommended actions. The telematics module transmits georeferenced alerts and diagnostic data to fleet operators and safety authorities to accelerate response times and prevent secondary accidents. The system includes self-tests, calibration routines, and drift compensation for temperature sensors. It logs events for subsequent analysis and model improvement, enabling continuous performance enhancements through over-the-air software updates. Safety policies implement tiered responses, ranging from recommended cooling times and torque limitations to automatic stop prompts when thresholds are exceeded.These correspond to preventive control strategies for mitigating thermal hazards in vehicles. The mechanical design includes robust, weatherproof camera mounts with vibration damping and heat shields that protect the optics from dirt and overheating, thus ensuring reliable operation on highways and in urban traffic. The architecture is scalable and suitable for passenger cars, trucks, and municipal vehicle fleets. It allows for modular camera configurations and processing levels and can be integrated into existing ADAS systems to provide supplementary sensor data in all weather conditions.

Claims

[1] A system comprising at least one vehicle-mounted thermal imaging camera, a processing unit for analyzing thermal images and data from auxiliary sensors, and a communication interface, wherein the processing unit detects temperature anomalies on the vehicle and predicts the risk of accidents in order to generate real-time warnings to drivers and authorities. [2] System according to claim 1, wherein the processing unit estimates the temperatures of brakes, tires or drive train based on thermal images and classifies the severity of overheating using machine learning models trained on identified component thermal patterns. [3] System according to claim 1, wherein an outward-facing thermal imaging camera with trained detectors enables the detection of vehicles and pedestrians and the estimation of distance in poor visibility conditions to support early collision avoidance. [4] System according to claim 1, wherein the processor links temperature anomalies with kinematic and environmental data to calculate an accident risk assessment and issues graded warning messages with georeferenced telemetry for proactive intervention and emergency response.