Object Tracking in Image Sequences via Intermediate Position Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection systems in sequences of images, such as video, face difficulties in accurately detecting and tracking objects, especially when they are obscured or in different orientations, leading to incomplete identification and processing.
Innovation Solution
A computer-implemented method and system that detects initial and subsequent positions of objects in a sequence of images, estimates intermediate positions through interpolation, and performs operations like blurring on selected objects without affecting others, using a face detection module, clustering module, estimation module, and blurring module to enhance object detection and privacy protection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If object detection systems process sequences of images frame-by-frame, then detection accuracy for individual frames is maintained, but completeness and continuity of object tracking deteriorates
Solution Approach 1:
The system performs preliminary detection of objects in initial frames before estimating their positions in intermediate frames. This allows the system to establish baseline object positions and track them through interpolation across subsequent frames, ensuring continuous tracking without requiring re-detection in every single frame.
Solution Approach 2:
The patent introduces intermediate position estimation as a mediator between initial and subsequent object detections. By calculating intermediate positions based on motion vectors or trajectory models, the system bridges gaps between detected frames, maintaining tracking continuity even when objects are obscured or detection is missed in intermediate frames.
2Reliability
If the system estimates intermediate positions of objects, then tracking continuity is improved, but computational complexity increases
Solution Approach 1:
The system applies partial estimation by only calculating intermediate positions for objects that were detected in both initial and subsequent frames. Rather than attempting to estimate positions for all objects in all frames, the system focuses computational resources on tracking continuous object trajectories, reducing overall computational complexity while maintaining tracking continuity for tracked objects.
Solution Approach 2:
The patent changes the parameter approach from detecting all objects in all frames to estimating position parameters (coordinates, size) of objects across time. By representing objects as parameters that change over time rather than re-detecting them each frame, the system reduces computational complexity while maintaining tracking continuity.
3Productivity
If the system performs operations on all detected objects, then processing completeness is maintained, but processing efficiency deteriorates
Solution Approach 1:
The system applies local quality by performing operations selectively on specific objects based on their properties or importance. Rather than uniformly processing all detected objects, the system identifies and prioritizes objects that require processing (such as faces or objects of interest) while skipping operations on less relevant objects, improving efficiency without sacrificing completeness for important targets.
Solution Approach 2:
The patent applies partial action by performing operations only on a subset of detected objects rather than all objects. The system determines which objects warrant processing based on detection confidence, object type, or other criteria, thereby improving processing efficiency while maintaining completeness for the most important objects.
4Measurement precision
If the system detects objects in obscured or different orientations, then detection accuracy improves, but detection reliability deteriorates
Solution Approach 1:
The system uses intermediate position estimation as a mediator to maintain reliable tracking of objects that may be obscured or appear in different orientations. By predicting where objects should be based on their trajectories rather than relying solely on direct detection, the system maintains reliable tracking even when detection is uncertain due to obscuration or orientation changes.
Solution Approach 2:
The patent implements feedback mechanisms where the system continuously refines object position estimates based on detected positions in subsequent frames. When objects are detected in different orientations or partially obscured, the feedback loop adjusts the trajectory model and position predictions, improving both detection accuracy and reliability through iterative refinement.
Data Source
AI summary
The subject matter of this specification can be implemented in, among other things, a computer-implemented method including detecting positions of objects of a specific type within an ordered sequence of images. The method includes estimating one or more intermediate positions of one or more intermediate instances of an object in one or more intermediate images within the ordered sequence of images between an initial image and a subsequent image based on an initial position of an initial instance of the object in the initial image and a subsequent position of a subsequent instance of the object in the subsequent image. The method includes providing a list of the objects for presentation. The method includes receiving a selection of the object from the list. The method includes performing an operation on the initial instance, the intermediate instances, and the subsequent instance of the object.