Video Indexing via Object Trajectory Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing density of video recording systems results in a time-consuming process for manually filtering large volumes of video data, and existing video indexing technologies face challenges in enhancing indexing efficiency.
Innovation Solution
A video indexing method and device that analyze object trajectory information, calculate similarities between object trajectories, and select representative object snaps to condense video data into indexing images, allowing for efficient browsing and indexing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual filtering is used to process video data, then the user can carefully examine each video segment, but the time required to filter and index video content increases significantly
Solution Approach 1:
The patent extracts key objects from video frames based on trajectory analysis and selects representative frames containing these objects. This extraction approach automatically identifies important video segments without manual filtering, significantly reducing indexing time while maintaining accuracy through object-based selection criteria
Solution Approach 2:
The system performs preliminary trajectory analysis on all objects in the video data before generating the index. By pre-processing and storing trajectory information, the system prepares object movement patterns in advance, enabling rapid identification of key video segments during indexing without time-consuming manual review
2Loss of information
If all video data is stored and processed, then complete video content is available for analysis, but the storage space and processing complexity increase
Solution Approach 1:
The patent extracts only the trajectory information of detected objects from complete video data, storing this condensed representation instead of all raw video frames. This extraction reduces storage requirements and processing complexity while preserving essential movement information needed for accurate video indexing and retrieval
Solution Approach 2:
The system segments video data by identifying and tracking individual objects separately, analyzing their trajectories independently. This segmentation approach simplifies processing by breaking down complex video content into discrete object movements, reducing overall system complexity while maintaining complete video content availability
3Loss of information
If trajectory analysis is performed on all objects, then comprehensive video understanding is achieved, but the computational resources and processing time increase
Solution Approach 1:
The patent applies different processing quality levels to different objects based on their importance. Representative objects whose trajectories significantly deviate from average patterns receive detailed analysis, while similar objects use simplified comparison. This local quality differentiation maintains comprehensive video understanding while improving indexing efficiency by avoiding uniform high-cost processing
Solution Approach 2:
The system performs trajectory analysis on all objects but applies selective detailed comparison only when necessary. By using average trajectory as a reference, the system can quickly dismiss similar objects with partial comparison and reserve comprehensive analysis for exceptional cases, achieving efficient indexing without sacrificing understanding completeness
Data Source
AI summary
A video indexing method includes: analyzing trajectory information of a plurality of objects in video data and storing a plurality of pieces of object trajectory data obtained correspondingly in a storage device; determining whether the storage device satisfies a data removal condition; when the data removal condition is satisfied, performing an object trajectory analysis on the plurality of pieces of object trajectory data to select at least one target object from the objects; and extracting the object trajectory data of the at least one target object from the storage device and generating at least one video indexing image accordingly.