Cascade target recognition method and system based on deep learning

A deep learning and target recognition technology, applied in neural learning methods, character and pattern recognition, instruments, etc., can solve problems such as poor handling and poor detection of small targets, and achieve the effect of reducing interference and improving detection accuracy.

Active Publication Date: 2020-01-14
STATE GRID INTELLIGENCE TECH CO LTD
View PDF7 Cites 12 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Its disadvantage is that we only pay attention to the features of the last layer in the deep network, but the detailed information can improve the detection accuracy to a certain extent
Therefore, the detection effect of small targets in a large field of view is poor.
At the same time, images in many scenes are subject to outdoor collection and complex lighting conditions, and will be affected by light and blur, making it more difficult to handle.

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
  • Cascade target recognition method and system based on deep learning
  • Cascade target recognition method and system based on deep learning
  • Cascade target recognition method and system based on deep learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0041] In the first embodiment, an image of power transmission and transformation is used as an input, and the target is a defect of power transmission and transformation equipment, which is described as an example.

[0042] In the cascade target recognition method based on deep learning figure 1 Showcasing the cascaded detection structure of the predictive architecture, figure 2 Shows the structure of the parallel target detection logic reasoning mechanism for image preprocessing, image 3 It shows the logic reasoning mechanism structure of nth level model parallel object detection.

[0043] The specific method includes the following steps:

[0044] In the example of the present disclosure, the power transmission line image of the UAV is used as the sample source. UAV is used to collect images of typical transmission line defects.

[0045] The disclosed example only takes the component defect at the bolt and pin of the sample defect category as an example. Manual select...

Embodiment 2

[0064] The difference from the above-mentioned embodiment is that the input data of this embodiment is the pre-determined regional inspection image collected by the drone, and the recognition target to be recognized is a regional disaster (such as debris flow, fire, etc.).

[0065] Obtain the inspection image of the area to be observed, and perform image preprocessing. The preprocessing methods include image rotation, image mirroring, image defogging, image enhancement, image deblurring and other methods.

[0066] Perform image augmentation.

[0067] Image expansion can choose to traverse and use all image processing methods for sample expansion, or use random probability to select image processing methods for sample expansion.

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 provides a cascade target recognition method and system based on deep learning, and the method comprises the steps: obtaining a to-be-detected sample of an inspection image, marking a target detection sample, and expanding the number of samples; fusing multiple features of the image enhancement data to carry out a multi-stage deep learning detection algorithm, realizing significant equipment detection for a target with a large proportion, and eliminating noise interference of a complex background on the detection algorithm. Multi-stage deep learning algorithm detection is carriedout by fusing multiple features of image enhancement data, so that the detection accuracy of deep learning in small target detection is improved, and the influence of image quality on the detection algorithm is reduced.

Description

technical field [0001] The disclosure belongs to the field of artificial intelligence, and specifically relates to a method and system for cascading target recognition based on deep learning. Background technique [0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art. [0003] In recent years, target detection has been widely used in military, medical, transportation, security and other fields. However, the detection of small target areas under a large field of view has always been a difficult problem in target detection. Small targets have fewer pixels and no obvious features. Therefore, compared with large targets, the detection rate of small targets is low. When the proportion of the target in the original image is very small, the recognition algorithm often makes misjudgments due to noise interference in other areas, especially in inspection images with complex backgroun...

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): G06K9/00G06K9/36G06K9/40G06K9/62G06N3/08
CPCG06N3/08G06V20/00G06V10/20G06V10/30G06F18/214G06F18/24
Inventor 刘广秀王万国许玮慕世友周大洲李建祥王振利刘丕玉张旭刘越贾亚军李勇郭锐赵金龙李振宇许荣浩
Owner STATE GRID INTELLIGENCE TECH CO 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