In-Vehicle AI Passenger Detection for Seating Configuration Checks

The patent combines multiple detection approaches (camera-based AI detection with traditional weight and seatbelt sensors) into a unified system. The camera system captures images of the vehicle interior and uses AI models to detect passenger presence and seating status, while traditional sensors continue to provide complementary data. This merging of detection methods resolves the contradiction by achieving high accuracy through multi-modal sensing while maintaining system simplicity through integrated processing.

The patent introduces camera-based visual detection as an intermediary between the physical seating state and the detection system. Instead of relying directly on weight sensors that produce false readings, the system uses cameras to capture visual evidence of passenger seating, which then serves as an intermediate verification layer. This intermediary approach eliminates false positives/negatives while the AI processing keeps the overall system manageable in complexity.

The patent implements feedback mechanisms where the AI model continuously analyzes camera images and compares detected seating status with sensor data. When discrepancies are detected (false positives or negatives), the system uses feedback loops to re-evaluate and correct detections. The processor receives ongoing feedback from both camera-based AI detection and traditional sensors, adjusting its determination of actual seating status to eliminate false readings and improve reliability.

The system performs preliminary detection using cameras and AI models before finalizing seating status determination. The camera system proactively captures images and the AI model pre-analyzes seating conditions, providing preliminary detection results that are then verified against traditional sensor data. This preliminary action prevents false positives and negatives by establishing an accurate baseline detection before final safety decisions are made.