A Recognition Method for Unknown Targets in Multispectral Images Based on Generalized Evidence Theory

A multi-spectral image and evidence theory technology is applied in the field of multi-spectral image recognition of unknown targets based on generalized evidence theory, achieving the effects of good real-time performance, simple calculation, and elimination of misjudgments

CN107967449BActive Publication Date: 2019-07-05NORTHWESTERN POLYTECHNICAL UNIV
3 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Publication Date
2019-07-05

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

Based on multi-spectral images, the present invention provides a method for realizing unknown target recognition under surface-space background by using generalized evidence theory, and relates to the fields of target recognition and image processing. The invention establishes a triangular fuzzy model for the target, cloud and sky, classifies the pixel points after widening the model, judges whether there is an unknown target according to the pixel point classification result, and updates the current triangular fuzzy number model with the pixel point classification result. The present invention uses the generalized evidence theory to classify the pixel points. This classification method better integrates the image information of different bands, can realize the discrimination of unknown targets, and has the advantages of simple calculation and good real-time performance; the random interference elimination method proposed by the present invention , which well excludes the misjudgment caused by random interference.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The invention relates to the fields of target recognition and image processing, and is a method for realizing recognition of unknown targets in multispectral images based on generalized evidence theory. Background technique

[0002] In the battlefield environment, the main objects of target recognition are aircraft targets that accidentally enter and invade our airspace, including non-cooperative targets and hostile targets. From the perspective of machine learning, when dealing with pattern recognition problems, researchers need to establish a template database for all possible target categories, and select all target category samples for training. However, it is impossible to build a complete database of non-cooperative and hostile targets. When the aircraft target is an unknown target outside the database, that is, a non-cooperative target or a hostile target, due to the incompleteness of its database, it is easy to misjudge the target as other kno...

Examples

Embodiment Construction

[0025] The present invention will be further described below in conjunction with accompanying drawings and examples.

[0026] Step 1: Input a frame of 25-band multi-spectral image taken at the angle facing the sky and the gray minimum, median and Maximum value, set up a corresponding triangular fuzzy number model according to the gray minimum, median, and maximum values ​​of the cloud, sky and current target imaging of the input, the current target is a currently interested tracking target (such as a defense target), The identification frame is In the identification frame, C means cloud, S means sky, T means current target, Representing an unknown target, the method for establishing the C, S, T triangular fuzzy number model is:

[0027] Input the minimum gray value of 131, the median value of 172.5 and the maximum value of 214 in band 1 cloud imaging, respectively as the minimum value, median value and maximum value of the band 1 cloud triangular fuzzy number model, then the...