Remote sensing image aircraft target detection method based on bounding box correction algorithm

A technology of remote sensing images and aircraft targets, which is applied in the field of aircraft target detection in remote sensing images, can solve problems such as inaccurate positions of classification label confidence boxes, inability to solve the problem of multi-frame false alarms, and damage to the integrity of aircraft targets. Achieve the effect of reducing false alarms, reducing missed detections, and obtaining accurate results

Active Publication Date: 2020-02-07
XIDIAN UNIV
View PDF3 Cites 13 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005]The traditional non-maximum suppression algorithm NMS method is based on the classification score, only the prediction frame with the highest score can be left, but in most cases the intersection and IoU It is not strongly correlated with the classification score, and many boxes with high confidence in the classification label are not very accurate
[0006]In addition, due to the segmentation of the remote sensing image, the aircraft target at the edge of the segmentation may be se

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
  • Remote sensing image aircraft target detection method based on bounding box correction algorithm
  • Remote sensing image aircraft target detection method based on bounding box correction algorithm
  • Remote sensing image aircraft target detection method based on bounding box correction algorithm

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0026] Aircraft target detection in remote sensing images is an important application of target detection in remote sensing images. Accurate positioning of aircraft targets is of great value in both military and civilian fields. In the military field, it can detect and strike enemy ground aircraft targets in a timely and effective manner, and win the initiative of the war; in the civilian field, it has the functions of realizing airport automation management and other functions.

[0027] The traditional non-maximum suppression NMS method used in the overlapping segmentation post-processing of aircraft target detection in remote sensing images is based on classification scores, and only the highest-scoring prediction frame can be left, but in most cases IoU and scores are not strongly correlated. If the processing simply selects the highest-scoring box after overlapping segmentation for aircraft object detection, many of the highest-scoring boxes will not be located very accurat...

Embodiment 2

[0045] The remote sensing image aircraft target detection method based on the bounding box correction algorithm is the same as embodiment 1, and the bounding box correction is carried out to the aircraft target detection result in the test sample in step (6), including the following steps:

[0046] (6a) Coordinate conversion: Convert the coordinates in the overlapping segmented image to the coordinates in the complete remote sensing image, and convert the coordinates in the upper left corner of the bounding box in the target detection result of each overlapping slice to the coordinate system of the segmented image The coordinates and the coordinates of the vertex in the lower right corner of the bounding box in the small image coordinate system are respectively converted into the bounding box position coordinates in the corresponding complete remote sensing image test sample, and the coordinate conversion is completed to obtain the bounding box detection result based on the comp...

Embodiment 3

[0052] The remote sensing image aircraft target detection method based on the bounding box correction algorithm is the same as embodiment 1-2, and the overlapping correction method described in step (6b) refers to: at first set an upper limit threshold T and a lower limit threshold K; if score If the intersection ratio iou of the highest frame and a frame is greater than the set lower threshold K, then the frame and the highest-scoring frame are considered to be marks on the same target; then the score is found from the frame screened by the lower threshold K The best frame, that is, the frame with the second best overall score, if the intersection ratio iou between this frame and the highest scoring frame is smaller than the upper limit threshold T set by the present invention, it is considered that the frame with the highest score at the beginning needs to be evaluated. Correction, that is, combined with the box with the second best score for correction, the specific method i...

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 remote sensing image airplane target detection method based on a bounding box correction algorithm, and solves the problems that a traditional non-maximum suppression algorithm directly selects a bounding box with the highest score, and the detection bounding box is inaccurate, and a segmented airplane is marked by multiple boxes during detection, so that false alarm is caused. The method comprises the following steps: generating a training and testing sample set; preprocessing a training sample; overlapping and slicing the test samples to obtain preprocessed test samples; training an aircraft target detection model; detecting an aircraft target in the slice; carrying out bounding box correction on a detection result in the test sample by using a bounding box correction method; and generating an airplane target detection result graph. According to the invention, overlapping correction is carried out on the intersecting boundary frames on the same target by using an overlapping correction method, so that a more accurate detection frame is obtained. Fusion improvement is performed on the parallel adjacent boundary frames on the same target by using an adjacent fusion method, and multi-frame false alarms on the same target are removed in a detection result. The method is used for remote sensing image airplane target detection.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to aircraft target detection in the technical field of image target detection, in particular to an aircraft target detection method for remote sensing images based on a bounding box correction algorithm. The invention can be used for detecting aircraft targets in remote sensing images. Background technique [0002] In recent years, computer vision technology represented by target detection and recognition has made remarkable progress. Applying target detection technology to optical remote sensing images has become one of the research hotspots in remote sensing science and computer vision technology. [0003] Aircraft is a high-value military equipment and vehicle. Using target detection technology to automatically detect ground aircraft targets on remote sensing images is of great significance in both military and civilian fields. In the military field, timely and ef...

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): G06K9/00G06K9/62
CPCG06V20/13G06F18/214
Inventor 侯彪周育榕焦李成马文萍马晶晶杨淑媛
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