Robot Object Detection Using Edge and Template Score Fusion

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing robot systems struggle to accurately identify and interact with objects in complex environments due to inefficiencies in template matching techniques, particularly in distinguishing object edges and appearances.

Innovation Solution

A computing system that performs attention-based template matching by using a processing circuit to analyze image and edge information, generating edge bitmaps and matching scores to plan robot interactions based on the degree of match between image and model templates.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional template matching is used for object recognition, then the system is simpler to implement, but the accuracy of object identification deteriorates in complex environments

Engineering Contradiction:
Improveobject recognition accuracyVSAvoidtemplate matching complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the template matching process into two distinct components: area-based matching (comparing overall object regions) and edge-based matching (comparing object boundaries). This segmentation allows each component to specialize in different aspects of object recognition, with edge-based matching providing robustness against appearance variations while area-based matching captures overall object presence, thereby improving overall accuracy without excessive complexity

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes the matching parameters by introducing edge detection algorithms (such as Canny edge detection) and using edge bitmaps as additional matching criteria. By transforming the matching process to operate on edge features rather than just pixel intensities, the system achieves better invariance to lighting changes, color variations, and partial occlusions, improving recognition accuracy in complex environments

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If edge-based template matching is implemented, then the accuracy of object identification is improved, but the computational complexity increases

Engineering Contradiction:
Improveobject identification accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent performs edge detection on both the input image and template images before the matching process. By pre-processing and storing edge bitmaps alongside the original images, the system avoids redundant edge detection computations during matching operations. This preliminary action significantly reduces computational complexity during runtime while maintaining the accuracy benefits of edge-based matching

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent merges area-based matching scores and edge-based matching scores into a combined matching score. This combination allows the system to leverage the strengths of both approaches: area-based matching provides robustness to overall object presence while edge-based matching provides sensitivity to object shape and boundary features. The merged approach achieves high accuracy without requiring extremely complex individual matching components

Inventive Principle:
Principle #5Merging (Combining)

3Reliability

If multiple template matching scores are calculated, then the reliability of object recognition is improved, but the processing time increases

Engineering Contradiction:
Improveobject recognition reliabilityVSAvoidprocessing time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent applies different matching strategies to different parts of the object recognition process: area-based matching is applied to overall object regions for quick initial assessment, while edge-based matching is applied to object boundaries for precise shape verification. This local differentiation allows the system to spend computational resources where they are most needed, improving reliability without uniformly increasing processing time across all matching operations

Inventive Principle:
Principle #3Local quality

Data Source

PatentUS12558790B2Method and computing systems for performing object detection
Publication Date: 2026.02.24 MUJIN INC
  • US12558790B2 patent drawing
  • US12558790B2 patent drawing
  • US12558790B2 patent drawing

AI summary

A computing system including a communication interface and a processing circuit. The communication interface communicates with a robot and a camera having a field of view. The processing circuit performs obtaining image information based on objects within the field of view and determines a first template matching score which indicates a degree of match between the image information and an model template. The processing circuit further determines image edge information based on the image information and determines a second template matching score which indicates a degree of match between the image edge information and a template. The processing circuit additional determines an overall template matching score based on the first template matching score and the second template matching score.