Hierarchical Snapshot Sampling for Efficient Object Re-Identification

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Surveillance systems face computational challenges in accurately tracking and re-identifying objects across multiple cameras due to the high computational intensity of existing descriptor extraction methods.

Innovation Solution

A hierarchical sampling method that clusters snapshots based on similarity and extracts descriptors of varying complexity levels, reducing the number of snapshots while maintaining re-identification accuracy by using neural networks to estimate snapshot quality and select representative frames.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If complex descriptors are applied to all snapshots for accurate re-identification, then identification accuracy is improved, but computational cost increases significantly

Engineering Contradiction:
Improvere-identification accuracyVSAvoidcomputational cost
Core Design Contradiction:
Measurement precisionVSPower

Solution Approach 1:

The patent segments the snapshot processing into two distinct groups: representative snapshots that receive complex descriptor processing, and non-representative snapshots that use simpler methods. This segmentation is achieved through clustering snapshots based on similarity metrics, selecting one representative from each cluster, and applying computationally intensive complex descriptors only to these representatives while using simpler descriptors for the remaining snapshots.

Inventive Principle:
Principle #1Segmentation

2Measurement precision

If all snapshots are processed with high computational methods, then identification accuracy is improved, but processing time increases

Engineering Contradiction:
Improvere-identification accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary clustering and representative snapshot selection before applying complex descriptor processing. By pre-identifying which snapshots are representative based on simpler similarity metrics, the system prepares the data structure in advance, allowing complex computations to be focused only on the necessary subset of snapshots, thereby reducing overall processing time while maintaining accuracy.

Inventive Principle:
Principle #10Preliminary action

3Productivity

If the number of snapshots processed is reduced for efficiency, then computational cost decreases, but identification accuracy may deteriorate

Engineering Contradiction:
Improveprocessing efficiencyVSAvoidre-identification accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies local quality by differentiating the processing quality applied to different snapshots. Representative snapshots, which are deemed most informative for identification, receive high-quality complex descriptor processing, while non-representative snapshots use simpler, lower-quality processing. This localized quality assignment ensures that computational resources are concentrated on the most critical snapshots, maintaining overall identification accuracy while improving efficiency.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP3800578B1Hierarchical sampling for object identification
Publication Date: 2026.05.06 JOHNSON CONTROLS TYCO IP HLDG LLP
  • EP3800578B1 patent drawingFigure 1
  • EP3800578B1 patent drawingFigure 2
  • EP3800578B1 patent drawingFigure 3

AI summary

Aspects of the present disclosure include methods, systems, and non-transitory computer readable media that perform the steps of receiving a first plurality of snapshots, generating a first plurality of descriptors each associated with the first plurality of snapshots, grouping the first plurality of snapshots into at least one cluster based on the plurality of descriptors, selecting a representative snapshot for each of the at least one cluster, generating at least one second descriptor for the representative snapshot for each of the at least one cluster, wherein the at least one second descriptor is more complex than the first plurality of descriptors, and identifying a target by applying the at least second descriptor to a second plurality of snapshots.