Object Recognition Across Pose Changes Using Adjacent Frame Matching

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional object recognition technologies struggle with low accuracy when recognizing objects that undergo form changes, such as a cat transitioning from a standing to a lying position, often misidentifying them as different objects.

Innovation Solution

The method employs an object recognition algorithm that analyzes the similarity and correlation between adjacent frames of images, using pixel movement speed and direction to determine if objects are the same, leveraging a stored object matching template with feature information for accurate recognition across varying forms.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional object recognition technology is used to recognize objects in fixed forms, then recognition accuracy for static objects is improved, but recognition accuracy for objects with changing forms deteriorates

Engineering Contradiction:
Improverecognition accuracyVSAvoidadaptability to form changes
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The system pre-calculates and stores multiple pose templates for each object type before recognition occurs. When an object is detected, the system compares the detected object against these pre-prepared templates to determine both identity and pose, eliminating the need for real-time pose estimation and improving both speed and accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system transitions from static single-pose recognition to dynamic multi-pose recognition by incorporating pose detection capabilities. The recognition system adapts to different object poses by selecting appropriate pre-calculated templates, enabling accurate recognition regardless of object orientation or configuration.

Inventive Principle:
Principle #15Dynamics

2Productivity

If object recognition is performed on each frame independently, then processing speed is improved, but recognition accuracy for moving objects deteriorates

Engineering Contradiction:
Improveprocessing speedVSAvoidrecognition accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The system uses detection results from previous frames as feedback to guide current frame detection. By tracking object positions and poses across frames and using this information to inform subsequent detections, the system maintains high processing speed while improving accuracy through temporal consistency and motion prediction.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP3855358B1Object recognition method, terminal device, computer readable storage medium and computer program product
Publication Date: 2026.02.25 HUAWEI TECH CO LTD
  • EP3855358B1 patent drawingFigure 1
  • EP3855358B1 patent drawingFigure 2
  • EP3855358B1 patent drawingFigure 3~4

AI summary

An object recognition method and a terminal device are provided. The method includes: A terminal device recognizes a first target object in a first frame of image. The terminal device recognizes a second target object in a second frame of image adjacent to the first frame of image. If a similarity between the first target object and the second target object is greater than a preset similarity, and a moving speed is less than a preset speed, the terminal device determines that the first target object and the second target object are a same object. In this method, the terminal device can recognize whether objects in adjacent frames of images are a same object, to help improve object recognition accuracy.