Video Processing Apparatus for Non-Overlapping Object Association

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing tracking systems struggle to accurately associate objects in images, particularly when the target object is not directly overlapping with the specific object, such as a person's lower body and their head, limiting the flexibility in object relationships.

Innovation Solution

A video processing apparatus with a target detection unit, specific object detection unit, and association unit that detects and associates regions of tracking targets with specific objects based on relative positions, allowing for non-overlapping associations and improved tracking accuracy through PTZ camera control.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If object association is based on image region overlap, then association accuracy is improved for overlapping objects, but association capability is lost for non-overlapping objects such as body parts

Engineering Contradiction:
Improveassociation accuracyVSAvoidassociation capability
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent transitions from two-dimensional image region overlap to three-dimensional spatial relationship analysis by introducing depth information and relative position calculations. This allows association of objects that are separated in the image plane but maintain spatial relationships in 3D space, such as body parts at different depths or positions.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The patent changes the association parameters from binary overlap detection to continuous relative position measurement. By calculating distances and angular relationships between object centers, the system can associate objects based on their spatial configuration rather than requiring pixel-level overlap, enabling body part association and other non-overlapping relationships.

Inventive Principle:
Principle #35Parameter changes

2Device complexity

If tracking uses only the target object region, then tracking simplicity is maintained, but tracking accuracy deteriorates when the target object is partially occluded or small

Engineering Contradiction:
Improvetracking simplicityVSAvoidtracking accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent introduces associated objects as intermediary elements that facilitate tracking. When the primary target object is occluded or becomes difficult to detect, the system uses the detected positions of associated objects (such as other body parts) to infer and maintain the target's position, ensuring continuous tracking without requiring direct detection of the target in every frame.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Ease of operation

If PTZ control is based on target object position only, then control simplicity is maintained, but control accuracy deteriorates when the target is small or at image edges

Engineering Contradiction:
Improvecontrol simplicityVSAvoidcontrol accuracy
Core Design Contradiction:
Ease of operationVSManufacturing precision

Solution Approach 1:

The patent merges the position information of multiple associated objects to determine the optimal PTZ control target. By calculating the centroid or weighted average position of the target object and its associated objects, the system achieves more stable and accurate camera control, especially when the primary target is small, low-contrast, or positioned at image edges where detection accuracy may be reduced.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentUS10438360B2Video processing apparatus, video processing method, and storage medium
Publication Date: 2019.10.08 CANON KK
  • US10438360B2 patent drawing
  • US10438360B2 patent drawing
  • US10438360B2 patent drawing

AI summary

A video processing apparatus is directed to appropriately associating a plurality of objects in an image. A target detection unit detects a position of a region of a tracking target included in a video image. A specific object detection unit detects a position of a specific object included in the video image. The specific object has a predetermined feature. An association unit associates the region of the tracking target with the specific object based on a relative position between the position of the region of the tracking target detected by the target detection unit and the position of the specific object detected by the specific object detection unit.