Multi-stage Multimodal Auto-detection of catalytic Converter Theft and Contextual Warning

摘要

This invention comprises a system and a method for auto-detection of catalytic converter theft and contextual warning. The proposed invention uses minimal set of existing sensors in low-trim vehicles. Figure 1 shows the multi-stage multimodal auto-detection of catalytic converter theft and warning system. A trained classifier takes as input acoustic signals from external and/or internal microphones and produces theft indicator as an output based on detecting the sound of the cutting tools used by the thieves (saw, angle grinder, drill, impact drill, etc.). ECUs serving IMU and catalytic converter health status sensor are activated upon detecting abnormal metal cutting sound by the audio-based classifier. The IMU data is used for vibration-based theft detection and jacking up detection based on trained models or calibratable thresholds. Information fusion module fuses acoustic-based theft indicator, vibration-based theft indicator, jacking up-based theft indicator and health status sensor data to generate fused theft estimate. The fused theft indicator is then consolidated using a priori contextual information such as time of the day, location, history records/heatmap of car theft or safety score. Different actions are recommended upon detecting the theft such as activating lights, horns and camera and alerting driver and keep alerting driver if no acknowledgement is received and report incident to the vehicle telematie system and update crowd-sourced heatmap.