Object Detection System Using 2D Camera and Trained Learning Model

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

There is a need for a low-cost and high-accuracy system to detect left-behind objects in transport containers to prevent improper shipment and ensure efficient container reuse.

Innovation Solution

A left-behind object detection system utilizing a combination of two-dimensional and three-dimensional cameras, where a first detection device uses a learning model trained on two-dimensional image data to detect objects, and a second detection device uses three-dimensional image data for accurate detection, with a reporting device alerting workers of any left-behind items, and a switch to manage information flow for detection results.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a three-dimensional camera is used for object detection, then detection accuracy is improved, but system cost increases

Engineering Contradiction:

Improvedetection accuracyVSAvoidsystem cost

Core Design Contradiction:

Measurement precisionVSEase of manufacture

Solution Approach 1:

The patent uses a two-dimensional camera to capture images and a learning model to create a virtual representation of the container interior, copying the functionality of expensive three-dimensional detection systems through software-based image analysis rather than hardware-based 3D scanning

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent replaces the mechanical/optical three-dimensional camera system with a computational approach using two-dimensional images and machine learning algorithms, substituting physical detection hardware with software-based analysis to achieve the same detection accuracy at lower cost

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Ease of manufacture

If a learning model is trained and used for detection, then system cost is reduced, but detection accuracy may be compromised

Engineering Contradiction:

Improvesystem costVSAvoiddetection accuracy

Core Design Contradiction:

Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent performs preliminary training of the learning model using three-dimensional image data as ground truth before deployment. The model is pre-trained and stored in memory, so during actual operation it can accurately classify objects using only two-dimensional images without requiring real-time 3D scanning

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent transforms the detection problem from three-dimensional space to two-dimensional image analysis by training a learning model that processes 2D images. The model learns to infer 3D object characteristics from 2D projections, effectively solving a 3D detection problem using 2D data

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

Data Source

PatentUS11587309B2Object detection system, object detection device, and object detection method

Publication Date: 2023.02.21 TOSHIBA TEC KK

AI summary

An object detection system includes a first detection device configured to control a first camera to acquire a first image of a container and detect an object in the container based on the first image, and a second detection device configured to control a second camera to acquire a second image of the container. The second image is a more detailed or accurate image than the first image for detecting the object in the container based on the second image. The first detection device is further configured to train a learning model for detecting the object in the container based on the first image using detection results from the second detection device as an indication of the correct result, and then use the trained learning model for detecting the object in the container based on the first image.