3D Camera Object Detection Using Depth Profile Validation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for detecting objects in warehouses using industrial trucks with 3D cameras lack precision and reliability, particularly in partially automated or driverless systems.

Innovation Solution

A method involving a 3D camera attached to an industrial truck captures image data, projects pixels onto a projection plane to create a two-dimensional dataset, and uses template matching to recognize objects, with a secondary depth validation step to improve detection accuracy by comparing depth profiles with predefined patterns at different distances.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If only two-dimensional image data is used for object detection, then the detection process is simpler and faster, but the accuracy and reliability of object recognition deteriorates

Engineering Contradiction:
Improvedetection speedVSAvoidobject detection accuracy
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent transitions from two-dimensional image data to three-dimensional data by incorporating depth information. The 3D camera captures depth values for each pixel, and depth profiles are extracted and compared with reference depth profiles. This dimensional enhancement allows the system to maintain fast processing while significantly improving object detection accuracy by validating both 2D visual appearance and 3D structural characteristics.

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

2Measurement precision

If depth information is fully utilized in the initial detection phase, then object recognition accuracy improves, but the computational complexity and processing time increase

Engineering Contradiction:
Improveobject detection precisionVSAvoidprocessing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the object detection process into two distinct phases: initial detection using 2D image data for rapid identification, and validation using depth profiles for precision confirmation. This segmentation allows the system to process only relevant depth information for objects that pass the initial 2D screening, thereby maintaining high precision while reducing overall computational complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system performs preliminary object identification using 2D image data before engaging in more computationally intensive depth profile analysis. This preliminary action filters out non-matching objects early in the process, so that full 3D depth validation is applied only to potential matches, optimizing the balance between precision and processing complexity.

Inventive Principle:
Principle #10Preliminary action

3Ease of operation

If template matching is performed without depth validation, then the detection process is more straightforward, but false positive recognition increases

Engineering Contradiction:
Improvedetection process simplicityVSAvoidfalse positive rate
Core Design Contradiction:
Ease of operationVSReliability

Solution Approach 1:

The patent introduces depth profile comparison as an intermediary validation step between 2D template matching and final object recognition. The depth profile acts as a mediator that verifies the three-dimensional structural consistency of detected objects, effectively filtering out false positives that may arise from 2D template matching alone while maintaining overall process simplicity.

Inventive Principle:
Principle #24Intermediary (Mediator)

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

This approach significantly enhances the reliability of object detection by preventing false positives and allowing precise position determination, even when objects are partially visible or at varying distances, thereby improving the accuracy and reliability of object recognition.

Implementation Method 1

The 3D camera can in particular time-of-flight camera

Methodology Applied
Scientific EffectTime of flight: Time of Flight

Data Source

PatentEP3316181B1Method for detecting objects in a bearing and industrial truck with a device for detecting objects in a storage location
Publication Date: 2021.07.28 JUNGHEINRICH AG
  • EP3316181B1 patent drawingFigure 1
  • EP3316181B1 patent drawingFigure 2a~4c

AI summary

A method for detecting objects in a warehouse comprising the following steps: • Capturing image data with a 3D camera mounted on a forklift, where the image data includes a multitude of pixels and distance information is assigned to each pixel, • Creating a two-dimensional dataset by projecting pixels onto a projection plane, • Comparing the two-dimensional dataset with a predefined pattern representing an object and preliminary detection of an object at a position on the projection plane if a match is found, • Creating another dataset comprising a depth profile along a line at the position on the projection plane, • Comparing the further dataset with a predefined depth profile pattern assigned to the object and final detection of the object if a match is found.