Vehicle Video Storage Using Object-Based Frame Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern vehicles face challenges in efficiently compressing video data captured by automotive cameras due to the high processing demands of existing compression techniques, which can interfere with real-time data processing required for driving automation systems (DAS) functions.
Innovation Solution
A method and system that determines object classes and instances within video data captured by automotive cameras, uses frame distances based on these objects to reduce the number of frames stored, and generates object lists for excluded frames, reusing processing already done for DAS functions to minimize additional data processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If video data is compressed using traditional compression techniques, then storage size is reduced, but processing effort increases and interferes with real-time DAS functions
Solution Approach 1:
The patent applies preliminary action by performing object detection and frame selection during the existing visual perception processing pipeline for DAS functions, before the video data needs to be stored. By identifying important frames and objects during the perception task (which is already being performed for driver assistance), the system prepares the video data for efficient storage without requiring separate compression processing later. This resolves the contradiction by using the already-necessary DAS processing to prepare the data, eliminating the need for additional compression effort.
Solution Approach 2:
The patent applies universality by making the visual perception processing serve dual purposes: (1) enabling real-time DAS functions for driver safety, and (2) preparing video data for efficient storage. The same object detection and frame analysis performed for DAS purposes are simultaneously used to identify which frames to store and which to compress, making the processing multi-functional. This resolves the contradiction by eliminating the need for separate compression processing while achieving both real-time DAS performance and efficient storage.
2Loss of information
If all video frames are stored, then complete video record is preserved, but memory footprint increases
Solution Approach 1:
The patent applies local quality by treating different video frames differently based on their importance to DAS functions. Instead of uniform storage or compression, the system identifies frames containing important objects or events (determined during visual perception processing) and preserves these in full quality, while applying different handling to less important frames. This resolves the contradiction by maintaining complete information for critical moments while reducing memory footprint for non-critical portions of the video record.
Solution Approach 2:
The patent applies discarding and recovering by selectively excluding certain frames from full storage based on their relevance to DAS functions. During visual perception processing, frames that do not contain important objects or events are identified and excluded from the stored video sequence. However, the system recovers the ability to reconstruct or retrieve information about these excluded frames through the object detection data and metadata generated during DAS processing. This resolves the contradiction by reducing memory footprint through selective discarding while maintaining information completeness through recovery via metadata.
Data Source
AI summary
A method for storing video data captured by automotive cameras installed in a vehicle. One or more object classes and object instances are determined within automotive sensor data, which include the video data. The determination of the one or more object classes and object instances is performed as part of one or more visual perception tasks enabling one or more driving automation system (DAS) features. Based on the determined object classes and object instances, the plurality of frames and one or more frame distances, a reduced plurality of frames is determined. For each frame not included in the reduced plurality of frames, a corresponding object list is generated, which identifies the one or more object classes and object instances determined within a corresponding frame as well as their corresponding positions. Then, the reduced plurality of frames and the object lists are stored.