Motion Trajectory Image Conversion for Robust Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for processing motion trajectories in human-machine verification rely on artificial rules and discrete statistics, leading to incomplete feature extraction and unstable results due to over-fitting and loss of implicit information.
Innovation Solution
Converting motion trajectories into images using position and time elements, performing image processing to obtain feature vectors, and using these vectors for human-machine identification or verification, which enhances feature extraction and stability by avoiding artificial rule-based modeling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If artificial rules and discrete statistics are used for feature extraction, then the extraction process is simple, but the feature quality is incomplete and loses implicit information
Solution Approach 1:
The patent copies the motion trajectory data into an image format, preserving all original information including implicit features. By representing trajectory points as image pixels with position and time encoding, the system maintains complete trajectory information without loss during the extraction process
Solution Approach 2:
The patent replaces manual rule-based feature extraction with automated image processing algorithms. Convolutional neural networks and other image analysis techniques automatically extract features from the trajectory image, eliminating the need for hand-crafted rules while preserving all implicit information
2Reliability
If functional approximation and curve fitting are used, then the motion trajectory can be modeled, but the results are unstable due to difficulty in selecting curve type and sensitivity to abnormal points
Solution Approach 1:
The patent enables the system to automatically process trajectory data through image encoding and automated feature extraction. The image processing algorithms self-adapt to different trajectory patterns without requiring manual curve selection, making the process reliable and stable across various scenarios
Solution Approach 2:
By copying trajectory data into image format, the system preserves the exact trajectory shape and characteristics without approximation. This eliminates curve fitting errors and sensitivity to abnormal points, as the image directly represents the original trajectory data
3Adaptability or versatility
If artificial rules are used for feature extraction, then the extraction process is controllable, but the rules have no universality due to strong relation with specific scenario
Solution Approach 1:
The patent creates a universal feature extraction system by encoding all motion trajectories as images with standardized position and time elements. This image-based representation can be processed by the same algorithms across different scenarios, making the system universally applicable without scenario-specific rule adjustments
Solution Approach 2:
The patent replaces manual rule definition with automated image processing. The system automatically extracts features from trajectory images using algorithms that adapt to different scenarios without requiring human intervention, achieving universality across various application contexts
Data Source
Figure 1~2
AI summary
A feature processing method and device for a motion trajectory, and a computer storage medium. The method comprises: a client acquires a motion trajectory generated by a user behavior to obtain a string of ordered point set, and each trajectory point in the ordered point set comprises a position element and a time element; use information in the position element and the time element to convert the motion trajectory into an image; and by performing image processing on the image, obtaining one or more feature vectors in the motion trajectory.