Object Movement Analysis via Motion Vector Geometry Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for analyzing object movements within a viewing area require an intermediate step of object formation, which can be time-consuming and inefficient, especially when dealing with dynamic camera movements or large volumes of data.
Innovation Solution
The method bypasses object formation by directly assigning motion vectors to geometry models, using speed diagrams to characterize object movement, and classifying it based on a combination of geometry and speed models, allowing for real-time analysis and reduced data processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If object formation is performed as an intermediate step to analyze object movements, then object tracking accuracy is improved, but processing time and computational complexity increase
Solution Approach 1:
The patent extracts and eliminates the object formation step from the conventional movement analysis pipeline. Instead of forming objects from pixel clusters and then tracking them, the method directly assigns motion vectors to geometry models representing objects, thereby removing the time-consuming intermediate step while preserving tracking capability through direct geometric model comparison
Solution Approach 2:
The patent segments the movement analysis process into independent geometry model evaluations. Each geometry model is evaluated separately against the motion vector field, allowing parallel processing and eliminating the sequential dependency of traditional object formation methods, thus reducing processing time while maintaining accuracy
2Measurement precision
If conventional object formation methods are used to track movements, then object identification accuracy is improved, but device complexity and data processing load increase
Solution Approach 1:
The patent inverts the conventional approach by not forming objects from pixels, but rather by directly comparing motion vectors with pre-defined geometry models. This inversion eliminates the complex object formation process while maintaining identification accuracy through direct geometric matching of motion patterns
Solution Approach 2:
The patent performs preliminary action by pre-defining geometry models that represent possible object shapes and movement patterns. These pre-established models are then directly compared with observed motion vectors, eliminating the need for complex real-time object formation and reducing processing complexity while maintaining identification accuracy
Data Source
Figure 1~2
Figure 3~5
AI summary
For the analysis of object movements using cameras, generally in public places, objects whose movement can be followed and analysed on the basis of the existing trajectories is determined on the basis of grey shade, colour and texture features, and detection of variations and movement vector fields,. The identification of objects technically represents, however, a non-robust working step which often leads to incomplete or false trajectories and therefore to poor analysis results, which is to be avoided in the invention. To this end, according to the invention, the individual movement vectors in the observation region are associated with applied geometry models, and path-time diagrams are drawn up from such a value assignment and compared to speed models. The invention can be applied to the detection of object movements by means of image analyses.