Video Classification via Motion Atom Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video classification methods face challenges in accurately classifying complex and continuous motions in videos due to difficulties in dividing complex motions into simple fragments and varying time points, leading to low precision and multiple classification results.
Innovation Solution
A video classification method that segments videos based on a time sequence to generate motion atoms, forms descriptive vectors using motion phrases that include motion atoms occurring near time points, and uses these vectors to classify videos, incorporating representativeness and coverage parameters to enhance classification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex motions are divided into fragments by time, then classification can be performed, but the division is difficult and results in multiple different classification results with low precision
Solution Approach 1:
The patent segments complex motions into motion atoms based on spatial and temporal characteristics. Motion atoms are generated by dividing videos into frames and extracting motion information from local areas, then clustering these motions to form atomic motion units. This segmentation allows complex continuous motions to be broken down into manageable components that can be systematically compared and classified.
Solution Approach 2:
The patent performs preliminary actions by pre-processing videos to generate motion atoms and building a motion atom library before actual classification. Sample videos are processed in advance to extract motion atoms, which are then stored for comparison against test videos. This preliminary preparation simplifies the actual classification process and improves precision by having ready-made reference units.
2Adaptability or versatility
If time points for dividing complex motions are set differently, then more flexible classification is achieved, but comparison scores become different and results cannot be unified
Solution Approach 1:
The patent changes parameters by using multiple comparison metrics with different weights. Instead of relying on a single comparison score, the system computes multiple scores based on different time point settings and motion atom combinations, then combines them using weight coefficients. This allows the system to adapt to different time divisions while maintaining consistent and reliable classification results through weighted aggregation.
3Ease of manufacture
If motion atoms are used for classification, then simple motion patterns can be identified, but complex continuous motions cannot be properly classified
Solution Approach 1:
The patent merges motion atoms into motion phrases by combining multiple motion atoms that occur in sequence near time points. Motion phrases represent continuous motion patterns and are formed by joining relevant motion atoms based on temporal proximity and semantic coherence. This merging capability allows the system to handle complex continuous motions while maintaining the simplicity of individual motion atoms as building blocks.
Solution Approach 2:
The patent adds a temporal dimension to motion atom comparison by introducing time point information and temporal proximity calculations. Instead of comparing motion atoms in isolation, the system compares them based on their temporal relationships and positions in the video sequence. This dimensional addition enables accurate classification of complex continuous motions while preserving the simplicity of atomic motion units.
Data Source
AI summary
A video classification method and apparatus relate to the field of electronic and information technologies, so that precision of video classification can be improved. The method includes: segmenting a video in a sample video library according to a time sequence, to obtain a segmentation result, and generating a motion atom set; generating, by using the motion atom set and the segmentation result, a motion phrase set that can indicate a complex motion pattern, and generating a descriptive vector, based on the motion phrase set, of the video in the sample video library; and determining, by using the descriptive vector, a to-be-detected video whose category is the same as that of the video in the sample video library. The method is applicable to a scenario of video classification.