An online training method for realizing target detection on unmanned equipment

A technology of target detection and training method, applied in the field of online training of target detection network, can solve the problems of limited sample value, insufficient target detection speed, large calculation amount of traditional methods, etc., to achieve large sample number, accurate real-time detection effect, The effect of high target detection accuracy

Active Publication Date: 2019-04-26
XIDIAN UNIV
View PDF3 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the sample values ​​obtained by offline training are relatively limited, and it is impossible to obtain a more accurate target area according to the current requirements of the user.
In the existing online training technology of target detection, only the method of using traditional target detection algorithm for online training of target detection is considered. The patent proposes an online training method using a target detection algorithm based on artificial threshold selection. This method improves the accuracy of target detection results to a certain extent by increasing the number of detection samples. In terms of selection, the subjective factors are strong, human intervention is required, and the traditional method has a large amount of calculation, which makes the speed of target detection not fast enough, and the detection accuracy is not high enough, and it is impossible to achieve real-time target detection and online training of the network. The effect of high detection results

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
  • An online training method for realizing target detection on unmanned equipment
  • An online training method for realizing target detection on unmanned equipment
  • An online training method for realizing target detection on unmanned equipment

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0021] With the development and progress of artificial intelligence, unmanned equipment occupies an important position in people's production and life. Unmanned equipment often needs functions such as target recognition and obstacle avoidance during routine operation. Therefore, it is necessary to design an image-specific An online training method for object detection networks. In the prior art, the target detection mainly adopts the network training method of offline training, and the sample values ​​obtained by the offline training are relatively limited, so that it is impossible to obtain a relatively accurate target area according to the current requirements of the user. In the existing online training technology of target detection, only the online training method of target detection using the traditional target detection algorithm is considered. Because the traditional target detection method has strong subjective factors in threshold selection, human intervention is requ...

Embodiment 2

[0031] The online training method method of realizing target detection on unmanned equipment is the same as embodiment 1, when designing three types of different target detection networks in step (1), because the YOLOV3 target detection network has accurate detection effect and detection speed is faster, so In the specific network design, the simplified YOLOv3 network is used, and part of its feature extraction network is removed to achieve accurate recognition of small targets. Here, considering that the ReLu function has a good implementation effect, the ReLu function is used as the hidden layer activation function; considering that the Adam optimization algorithm has a relatively fast convergence speed, the Adam optimization algorithm is used for optimization. Because the conditions and purposes of target detection in panoramic video, visible light video and infrared video are different, three different target detection networks are designed according to the characteristics ...

Embodiment 3

[0036] The online training method method of realizing target detection on unmanned equipment is the same as embodiment 1-2, in step (5), select corresponding image database according to new target image and semantic information and expand image database, specifically for unmanned equipment In the process of performing target detection, according to the results of the three types of image detection, the images containing the targets with a confidence degree greater than the set threshold are saved in the corresponding image library, and the expansion and maintenance of the three types of image libraries are completed.

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 an online training method for realizing target detection on unmanned equipment, and solves the problems that target detection and network training cannot be performed in parallel in real time and the detection result is inaccurate in the prior art. The method comprises the following implementation steps: respectively designing a target detection network according to the characteristics of a panoramic video, a visible light video and an infrared video, and carrying out preliminary training; Judging whether target related images and information exist in the information transmitted into the unmanned equipment or not; Selecting a corresponding image library according to the new target image and the information, and expanding the image library; And carrying out online training on each target detection network, finally obtaining new network structure parameters, and replacing a previous network parameter file. According to the invention, an on-line training mode is adopted, target detection and network training can be performed in parallel in real time, the target detection precision is higher, and the method can be used for unmanned equipment in an operation state.

Description

technical field [0001] The invention belongs to the technical field of image processing, and mainly relates to online training of unmanned equipment target detection, in particular to an online training method for realizing target detection on unmanned equipment, which can be used to perform target detection of various types of images on unmanned equipment Online training of detection networks. Background technique [0002] With the development and progress of artificial intelligence, unmanned equipment occupies an important position in people's production and life, and plays an important role in many fields such as remote sensing mapping, disaster rescue, and environmental protection testing. Unmanned equipment often needs functions such as target recognition and obstacle avoidance during routine operation, so it is necessary to design an online training method for image-based target detection networks. Target detection is to detect the target area and perform image segmen...

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): G06K9/00G06F16/51
CPCG06V20/41G06V2201/07
Inventor 张静胡锐桑柳邵旻昊周秦李云松
Owner XIDIAN 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