Video Output System with Recognition-Based Event Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video output systems for automatic driving vehicles struggle to efficiently present and record video data from cameras, as passengers find it difficult to grasp the surrounding environment from directly presented videos, and the amount of recorded data becomes overwhelming.
Innovation Solution
A video output system that includes a camera with an image sensor, a recognition unit, a comparison unit, and a camera calculation unit, which processes captured video to identify specific objects or environments and outputs only the video meeting predetermined recording conditions, along with identification and additional information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If captured video is directly presented to passenger, then passenger can see surrounding environment, but passenger cannot grasp recognition situation and information becomes overwhelming
Solution Approach 1:
The system extracts only the essential recognition results (detected objects, their types, and positions) from the full video data, separating the critical information from the redundant visual content. This allows passengers to understand the surrounding environment through concise textual/visual overlays rather than processing entire video streams.
Solution Approach 2:
The video information is segmented into distinct components: full video capture for system processing, extracted recognition results for passenger display, and selective recording of specific events. This segmentation enables different uses of the same source data at different levels of detail.
2Reliability
If captured video is directly recorded, then complete surrounding environment is recorded, but amount of recorded data becomes enormous
Solution Approach 1:
The system extracts only video segments that contain significant events (object detections, collisions, pedestrian crossings) for permanent recording, rather than recording all video continuously. This extraction based on recognition results dramatically reduces recorded data volume while maintaining reliability for critical events.
Solution Approach 2:
Instead of continuous recording, the system uses periodic event-triggered recording based on recognition results. Video is recorded only when specific conditions are met (object detection, collision risk, pedestrian in path), creating a periodic recording pattern that reduces overall data volume while capturing all essential events.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A video output system (100) provided with a camera (1) and an interface device (2), wherein the interface device (2) is provided with a condition storage unit (22) which specifies conditions under which a video is to be selected by the camera, and a video output unit (21) which outputs a captured video received from the camera (1), and wherein the camera (1) is provided with: an image sensor (5); a video generation unit (6) which generates captured videos on the basis of the output of the image sensor (5); a condition acquisition unit (8) which acquires said conditions from the interface device (2); a recognition/comparison unit (7, 9) which determines whether or not each captured video generated by the video generation unit (6) meets the acquired conditions, and if it is determined that a captured video meets the conditions, generates identification information on the basis of recognition results of a subject to be recognized that is included in the captured image determined to meet the conditions, and that is associated with the conditions; and a camera calculation unit (10) which outputs, to the interface device (2), at least each captured video determined by the recognition/comparison unit (7, 9) as meeting the conditions, from among the captured videos generated by the video generation unit (6), and the identification information associated with that captured video.