3D target detection method without post-processing operation

A target detection and 3D technology, applied in the field of computer vision, can solve problems such as structural incompatibility, and achieve the effect of saving time and overhead

Active Publication Date: 2021-06-29
ZHEJIANG UNIV
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, current mainstream 3D object detectors are not structurally adapted to this strategy

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • 3D target detection method without post-processing operation
  • 3D target detection method without post-processing operation
  • 3D target detection method without post-processing operation

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0030] The present invention will be further described in detail below with reference to the accompanying drawings and examples, and the following examples are intended to facilitate the understanding of the present invention, but will not afford any limits.

[0031] Such as figure 1 As shown, a 3D target detection method without post-processing operation, including the following steps:

[0032] Step 1, in the target area, the K 3D candidate box and 1 object embedded feature are initialized. Candidates and object embedded features are available. As the training is carried out, the object embedding feature encodes the general characteristics of the object to be detected, indicating that e∈R 1×C Where C represents the characteristic dimension;

[0033] Step 2, we use the feature extraction model PointNet used by the existing Votenet to extract the input point cloud samples. Specifically, refer to "Deep Hough Voting for 3D Object Detection in PointClouds". Point cloud sample s∈R n×3 ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses a 3D target detection method without post-processing operation. The method comprises the following steps: (1) initializing K 3D candidate frames and an object embedding feature; (2) carrying out feature extraction on an input point cloud sample to obtain point features; (3) extracting K 3D candidate frame features from the point features; (4) screening and extracting 3D candidate frame features by using object embedded features to obtain K features; (5) enabling the K features to exchange feature information by using a self-attention model to obtain K proposal features; (6) predicting K prediction results according to the proposal features, and training after matching the prediction results with the annotation information one by one; (7) replacing the K 3D candidate frames in the step (1) with the 3D candidate frames of the K prediction results predicted in the step (6), and replacing the object embedding in the step (1) with the feature proposal obtained in the step (5); and repeating the steps (3)-(7) for multiple times to obtain a detection result. According to the invention, the problem of redundant prediction of the existing 3D target detector can be solved.

Description

Technical field [0001] The present invention belongs to the field of computer visual, in particular, to a 3D target detection method that does not require post-processing operations. Background technique [0002] 3D target detection is a widely used technique for unmanned, indoor object detection, robotic navigation, and the like. The 3D target detection task is input to point cloud data, and the output prediction results include the location of the 3D box, the category of the object in the 3D box, and the confidence of the 3D box. [0003] In recent years, the detection accuracy of the 3D target detector has greatly improved, including: "DeepHough Voting for 3D Object Detection in Point Clouds", "Ahierarchical GraphNetwork for 3D Object Detection On Point Clouds", "H3DNET: 3D Object Detectionusing Hybrid Geometric Primitives, "MLCVNET: MULTI-Level Context Votenet for3D Object Detection". [0004] However, the prediction results of these detectors have a large amount of redundanc...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/00G06K9/62
CPCG06V20/64G06V2201/07G06F18/214
Inventor 刘子立蔡登徐国栋杨鸿辉何晓飞
Owner ZHEJIANG UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap