Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Three-dimensional target detection method

A target detection algorithm and detection method technology, applied in the field of 3D target detection, can solve problems such as backward detection accuracy, and achieve the effects of improving speed and accuracy, fine-grained instance segmentation, and high recall rate.

Inactive Publication Date: 2019-10-29
XIAMEN UNIV
View PDF2 Cites 9 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Compared with 2D target detection, 3D target detection has greater challenges and the detection accuracy is relatively backward. Therefore, how to improve the detection speed and detection accuracy is the direction that the industry needs to work on

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
  • Three-dimensional target detection method
  • Three-dimensional target detection method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0029] see figure 1 , the invention discloses a three-dimensional object detection method, which comprises the following steps:

[0030] S1. Using a two-dimensional object detection algorithm to extract a candidate bounding box of an object on the original image.

[0031] The original image is a depth image, and the candidate bounding boxes are 2D bounding boxes. The purpose of extracting candidate bounding boxes is to extract point clouds within their bounding boxes. The two-dimensional target detection algorithm adopts the existing detection algorithm, as long as the purpose of obtaining the two-dimensional bounding box of the target can be achieved, and the present invention does not specifically limit it.

[0032] S2. Convert the depth image region corresponding to the candidate bounding box into a frustum point cloud. This step is achieved through the following sub-steps:

[0033] S21. Extract the depth image region corresponding to the candidate bounding box to obtai...

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 three-dimensional target detection method. The method comprises the following steps: extracting a candidate bounding box of a target on an original image by using a two-dimensional target detection algorithm; converting a depth image region corresponding to the candidate bounding box into a view cone point cloud; carrying out instance segmentation on the view frustum point cloud to obtain interested target point cloud; and regressing the three-dimensional bounding box of the target through the neural network. According to the method, a boundary frame is extracted through the two-dimensional target detection algorithm and then returned to the three-dimensional boundary, the target detection speed and precision can be improved, and the characteristics of the two-dimensional information and the characteristics of the three-dimensional information can be used for detecting the shielded target in the detection process.

Description

technical field [0001] The invention relates to the technical field of target detection, in particular to a three-dimensional target detection method. Background technique [0002] Compared with two-dimensional semantic understanding tasks, three-dimensional semantic understanding tasks are more challenging and meaningful. 3D object detection plays an extremely important role in autonomous driving and augmented reality. On the one hand, three-dimensional target detection can know the approximate size and position of the target, which is very important for automatic driving; on the other hand, when the augmented reality technology needs to combine and interact with the real world and the virtual world, the target The calculation of position change and size of is especially important. On the road of autonomous driving development, a fast and accurate target detection algorithm is urgently needed to ensure the safety of traffic and the reliability of intelligent driving. [...

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
IPC IPC(8): G06K9/00G06K9/32G06K9/34G06N3/04G06N3/08
CPCG06N3/08G06V20/64G06V10/25G06V10/267G06N3/045
Inventor 陈一平林伟生李军王程
Owner XIAMEN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com