Vehicle Event Detection via Selective Video Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Dashcams face challenges in efficiently detecting and reporting events such as dangerous driving due to the continuous recording of images and video, which consumes significant network and memory resources, leading to delayed and inaccurate event reporting.
Innovation Solution
Implementing a system that uses machine learning models and image processing techniques to detect events in real-time, allowing for alerts and notifications without storing non-event footage, thereby reducing data transmission and storage needs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If continuous recording of images and video is implemented, then event detection capability is improved, but network and memory resource consumption increases significantly
Solution Approach 1:
The patent extracts and processes only relevant portions of continuous video data by detecting events in real-time and selectively capturing only event-related frames. This eliminates the need to store and transmit entire continuous video streams, significantly reducing memory and network resource consumption while maintaining event detection capability.
Solution Approach 2:
The system performs preliminary event detection on continuous video streams before full recording or transmission occurs. By using machine learning models to pre-analyze video content and identify events, the system can selectively record only relevant segments, avoiding waste of resources on non-event footage.
2Reliability
If continuous recording and storage of all footage is implemented, then no event data is lost, but data transmission and storage needs increase significantly
Solution Approach 1:
The patent extracts only event-related video segments from continuous recording by using object detection and event classification algorithms. This extraction process identifies and isolates portions of video containing events of interest, eliminating the need to transmit and store entire continuous video streams.
Solution Approach 2:
The system changes the recording parameter from continuous full-resolution video capture to event-triggered selective capture. By monitoring video content in real-time and detecting event parameters (such as object presence, motion patterns, or specific scenarios), the system dynamically adjusts recording behavior to capture only relevant data.
3Loss of time
If real-time event detection is implemented, then timely alerts can be provided, but processing complexity increases
Solution Approach 1:
The patent segments the complex event detection task into multiple specialized machine learning models, each trained to detect specific event types or objects. This segmentation allows parallel processing of different video streams or event categories, reducing overall processing complexity while maintaining real-time detection capability.
Solution Approach 2:
The system introduces an intermediary layer of event detection algorithms that act as mediators between raw video input and final event reporting. This intermediary layer pre-processes video data, identifies potential events, and filters out non-event content before full analysis or recording occurs, simplifying the overall processing pipeline.
Data Source
AI summary
One or more devices may detect an indication of relevant multiple images of an environment of a vehicle. The indication may include at least one of input from a user of the vehicle, a set of instructions, or a threshold measurement by a sensor. The one or more devices may mark the multiple images of the environment as relevant. The one or more devices may identify corresponding types of one or more objects shown in the multiple images. The one or more devices may determine whether the multiple images show an occurrence of an event with respect to the one or more objects. Detection of the event may be based on determining corresponding locations of the one or more objects in relation to each other or to the vehicle. The one or more devices may perform, based on whether the multiple images show the occurrence of the event, one or more actions.