Vehicle Event Detection Using Deep Learning Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle systems require manual driver intervention to broadcast decentralized environmental notification messages (DENM), which can divert attention from the road and compromise safety.
Innovation Solution
A vehicle system utilizing a primary deep learning model with a camera sensor to detect events and a secondary deep learning model on a server for validation, allowing auto-triggering of DENM messages without driver interaction, with periodic updates and confirmation through a telematic control unit.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If manual driver intervention is used to broadcast DENM messages, then the driver can control event notification, but driver attention is diverted from the road compromising safety
Solution Approach 1:
The system enables automatic event detection and DENM message broadcasting through the deep learning model without requiring driver intervention. The electronic control unit autonomously processes images from the camera sensor, detects events, and triggers notifications, allowing the system to serve itself rather than relying on manual driver input.
Solution Approach 2:
The patent replaces the manual mechanical interaction (driver pressing buttons or touching screens) with an automated electronic system. The deep learning model and electronic control unit substitute for the driver's manual operations, automatically detecting events and broadcasting DENM messages based on image analysis.
2Reliability
If automatic event detection is implemented using deep learning model, then driver distraction is reduced, but system complexity increases
Solution Approach 1:
The deep learning model acts as an intermediary between the camera sensor and the DENM message broadcasting system. It processes images and determines whether events warrant notification, mediating between raw sensor data and the communication system while maintaining automated operation without direct driver involvement.
Solution Approach 2:
The system segments the event detection process into distinct functional modules: image capture by camera sensor, event detection by deep learning model in the electronic control unit, and notification broadcasting by the telematic control unit. This segmentation allows each component to specialize in its function while working together autonomously.
3Measurement precision
If multiple sensor types are used for event detection, then detection accuracy improves, but system complexity and cost increase
Solution Approach 1:
The patent replaces complex multi-sensor hardware systems with a simplified single-sensor approach enhanced by deep learning software. Instead of using multiple physical sensors (LIDAR, radar, cameras), the system uses only camera sensors but compensates for their limitations through advanced image processing and deep learning algorithms that can detect various event types with high accuracy.
Solution Approach 2:
The system changes the parameter of detection from physical sensor diversity to computational complexity. Rather than adding more sensor types, it enhances the processing capabilities of the existing camera sensor through deep learning models, transforming the approach from hardware complexity to software intelligence.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a vehicle system (1) of a vehicle (2) configured to detect an event (E) and to broadcast said event (E) using a decentralized environmental notification message (DENM), wherein said vehicle system (1) comprises: - at least one camera sensor (10) configured to capture images (I1) of an environment of said vehicle (2), - an electronic control unit (11) configured to : - detect an event (E) using a primary deep learning model (M1) based on said images (I1), - apply an predictability level (A) on said event (E), said predictability level (A) being generated by said primary deep learning model (M1), - transmit said event (E) to a telematic control unit (12) if its predictability level (A) is above a defined level (L1), - said telematic control unit (12) configured to : - receive said event (E) from said electronic control unit (10) and broadcast a related decentralized environmental notification message (DENM) via a vehicle to vehicle communication (V2V) and/or a vehicle to infrastructure communication (V2I), - transmit at least one image (I1) and data details (D) of said event (E) to a server (3), - receive a primary validation information (30) of said event (E) from said server (3), said primary validation information (30) being generated by a secondary deep learning model (M2), and cancel the broadcasting of said decentralized environmental notification message (DENM) if said event (E) is not validated, - if said event (E) is validated, receive an updated instance (M3) of said primary deep learning model (M1) from said server (3) and transmit it to said primary electronic control unit (10) for updating said primary deep learning model (M1).