Object Detection Using 3D Point Cloud Region Proposals

Solution Overview

Problem

Existing object detection methods generate unreasonable redundant region proposals, leading to waste of computing resources and time during feature extraction.

Innovation Solution

The method involves acquiring a scene image and a three-dimensional point cloud, segmenting the image into sub-regions, merging these sub-regions based on the three-dimensional point cloud to generate more accurate and fewer region proposals, and then performing object detection on these proposals.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If selective search or deep learning methods are used to generate region proposals, then object detection can be performed, but redundant region proposals are generated leading to waste of computing resources and time

Engineering Contradiction:

Improveobject detection accuracyVSAvoidcomputing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary actions by acquiring three-dimensional point cloud data before generating region proposals. This pre-acquired spatial information is used to guide the merging of sub-regions, ensuring that only meaningful region proposals are generated before the actual object detection process begins, thus avoiding redundant computations

Inventive Principle:

Principle #10Preliminary action

2Measurement precision

If selective search or deep learning methods are used to generate region proposals, then object detection can be performed, but redundant region proposals are generated leading to waste of computing resources

Engineering Contradiction:

Improveobject detection accuracyVSAvoidcomputing resource consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent performs preliminary actions by acquiring three-dimensional point cloud data before generating region proposals. This pre-acquired spatial information is used to guide the merging of sub-regions, ensuring that only meaningful region proposals are generated before the actual object detection process begins, thus avoiding redundant computations

Inventive Principle:

Principle #10Preliminary action

3Reliability

If more region proposals are generated to improve detection coverage, then more objects can be detected, but computing time and resource consumption increase

Engineering Contradiction:

Improvedetection coverageVSAvoidobject detection efficiency

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The patent extracts and utilizes three-dimensional spatial information from point cloud data to guide the region proposal generation process. By extracting meaningful spatial relationships, the system can generate fewer but more accurate region proposals that maintain detection coverage while improving efficiency

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent introduces three-dimensional point cloud data as an intermediary element between image segmentation and region proposal generation. This intermediary provides spatial context that enables smarter merging decisions, reducing the number of proposals needed while maintaining reliability

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS12293593B2Object detection method, object detection device, terminal device, and medium

Publication Date: 2025.05.06 GUANGDONG OPPO MOBILE TELECOMMUNICATIONS CORP LTD

AI summary

An object detection method is provided. The method includes: acquiring a scene image of a scene; acquiring a three-dimensional point cloud corresponding to the scene; segmenting the scene image into a plurality of sub-regions; merging the plurality of sub-regions according to the three-dimensional point cloud to generate a plurality of region proposals; and performing object detection on the plurality of region proposals to determine a target object to be detected in the scene image. In addition, an object detection device, a terminal device, and a medium are also provided.