Image acquisition method and device for simulating crowded scene and visual processing method

A technology for crowding scenes and simulating images, applied in the field of image processing, it can solve the problems of insufficient data, missed or false detection of robots, and highly similar data distribution, and achieves the effect of ensuring quality, good generalization performance and robustness.

Active Publication Date: 2021-01-08
UBTECH ROBOTICS CORP LTD
View PDF9 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] Taking the robot application as an example, when training the network model of the actual robot, it often encounters the situation of insufficient data or highly similar data distribution, which causes the robot to perform poorly in some specific scenarios.
For example, take a robot tracking a football as an example, especially when a part of the football is blocked by sports personnel, etc., the robot is often prone to missing or false detections

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
  • Image acquisition method and device for simulating crowded scene and visual processing method
  • Image acquisition method and device for simulating crowded scene and visual processing method
  • Image acquisition method and device for simulating crowded scene and visual processing method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0068] Please refer to figure 1 , this embodiment proposes an image acquisition method for simulating a crowded scene, which can be applied to a scene where partial information of a simulated target object is blocked or overlapped. Considering the limited number of target scene images, this embodiment uses a series of image processing operations to process the limited target scene images to obtain more simulated images in crowded scenes. This method will be described below.

[0069] Step S110, performing enhancement processing on the target scene image according to a preset rule to obtain a first preset number of enhanced images, wherein a preset ratio of multiple enhanced images all contain the target object.

[0070] Usually, the number of target scene images includes multiple pieces, and a larger number of enhanced images can be obtained by using these target scene images for data enhancement processing. The first preset quantity can be set according to actual training ne...

Embodiment 2

[0106] Please refer to Figure 5 , similar to the image acquisition method in the above-mentioned embodiment 1, this embodiment proposes an image acquisition method, which is not limited to simulate images in crowded scenes where object information is blocked, but can also be used in other occasions, such as Training image augmentation when migrating from a simulation scene to a real scene, etc. Due to the roughness of the simulation data, it is impossible to accurately simulate the texture and other information of the real target target. If the image acquisition method is used for data expansion based on part of the real scene data, it can make up for the defects in the simulation data, thereby reducing the number of algorithm problems. The time consumption when the simulation scene is migrated to the real scene, etc.

[0107] Step S210, performing an enhancement on the target scene image to obtain multiple first images. Wherein, the primary enhancement includes performing ...

Embodiment 3

[0118] Please refer to Image 6 , this embodiment proposes a visual processing method, which mainly applies the above-mentioned image acquisition method to a terminal device including a camera, such as a robot. The visual processing method will be described below.

[0119] In step S310, the target scene image is captured by the photographing device, and a simulated image is generated based on the target scene image using the above-mentioned image acquisition method.

[0120] Generally, when the uses of the terminal devices are different, the neural network models adopted by the terminal devices are also different. Considering that the number of target scene images in the actual scene collected by the terminal device may be limited, when training the model, the training data may be insufficient or the distribution is highly similar, which will affect the training effect of the model. Therefore, this implementation The example proposes to enrich the diversity of training data ...

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 embodiment of the invention provides an image acquisition method and device for simulating a crowded scene and a visual processing method, and the method comprises the steps: carrying out the enhancement of a target scene image according to a preset rule, and then sequentially carrying out cutting, splicing and overlapping on all enhanced images according to corresponding cutting, splicing andoverlapping rules, so that an aliasing image containing the information of the target object is used as a simulation image for simulating a crowded scene. According to the technical scheme, the problem that the number of actual sample images is limited during model training in various visual application scenes is well solved, especially for a crowded scene in which information of a target objectis shielded; in addition, the simulation image obtained through the method can be compatible with the context information of the global scene image and the local information of the target object in the scene image, so that the model training quality and the like can be well ensured.

Description

technical field [0001] The present application relates to the technical field of image processing, in particular to an image acquisition method, device and visual processing method for simulating crowded scenes. Background technique [0002] Visual information is one of the most intuitive input information for human beings. With the vigorous development of technologies such as smartphones and short video applications, the scale of visual image data continues to grow. Deep learning technology, especially convolutional neural network technology, is constantly improving and iterating with the growth of visual image datasets, but complex convolutional neural network models will bring a higher risk of network overfitting: the model is overfitting The features in the data set lead to poor performance of the model when migrating to the real scene. Therefore, the quality of visual image data is one of the important factors that directly affect the final performance of computer vis...

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): G06K9/00G06T7/246G06T5/00G06T5/50G06T3/40G06T7/10
CPCG06T7/246G06T5/007G06T5/50G06T3/4038G06T7/10G06T2207/20132G06T2207/10016G06T2207/20081G06T2207/20084G06V20/53Y02T10/40
Inventor 王阳赵明国
Owner UBTECH ROBOTICS CORP LTD
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