Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

A Fast Target Detection Method in Remote Sensing Image Based on Rotation Anchor Point Clustering

A remote sensing image and detection method technology, which is applied in the field of remote sensing image target detection, can solve the problems of cumbersome parameter adjustment in the training process, poor anchor point pertinence, and large anchor point redundancy, so as to reduce the number of anchor points and improve detection Accuracy and detection speed, the effect of easy algorithm

Active Publication Date: 2022-06-17
HARBIN ENG UNIV
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The detection method of horizontal anchor points can easily lead to missed detection in the case of dense objects, thereby reducing the detection accuracy; and the manual design of anchor points is not very specific, and the anchor point redundancy is large, which directly leads to the need to increase a large number of different sizes and widths and heights Compared with the anchor point, it is used to cover detection targets of different sizes and shapes. The training process is cumbersome to adjust parameters, which greatly limits the detection speed.

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
  • A Fast Target Detection Method in Remote Sensing Image Based on Rotation Anchor Point Clustering
  • A Fast Target Detection Method in Remote Sensing Image Based on Rotation Anchor Point Clustering
  • A Fast Target Detection Method in Remote Sensing Image Based on Rotation Anchor Point Clustering

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0042] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0043] The present invention first finds a suitable rotation anchor point through an algorithm based on k-means clustering, and then uses a two-stage detection algorithm to detect a specific target to improve the slow detection speed of remote sensing image targets. as follows:

[0044] (1) After preprocessing the input image, the feature information of the image is extracted by the deep convolutional neural network as the backbone network, and output to the next link as the feature map;

[0045] (2) Use the k-means clustering algorithm to cluster the labeled frame data in the training set image, and use the scale, width and height as prior information on the feature map to count the k most representative anchor points, It performs rotation processing to obtain the rotation anchor point that is most likely to cover the target;

[0046] (3) Perf...

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 a remote sensing image target rapid detection method based on rotation anchor point clustering. Firstly, a rotation anchor point is designed based on a k-means clustering algorithm to obtain a series of rotation anchor points; Binary classification and coordinate rough regression, combined with post-processing of rotation non-maximum suppression, obtain positive and negative sample information and streamlined high-quality proposals; finally, multi-scale rotation RoI pooling is performed on the proposal to obtain fixed RoIs containing regions of interest Length vector, input the vector to the fully connected layer (FC) for classification and coordinate regression of specific categories, and use INMS post-processing again to obtain the final detection result of the target. The invention can effectively reduce the redundancy of anchor points, improve the detection speed and detection accuracy of remote sensing image targets, the algorithm is easy to implement, the parameters are easy to adjust, and has the advantages of mathematical explainability. The method has broad application prospects and good economic benefit.

Description

technical field [0001] The invention relates to a method for rapid detection of remote sensing image targets, in particular to a rapid detection method for remote sensing image targets based on rotation anchor point clustering, and belongs to the field of remote sensing image target detection. Background technique [0002] Since the United States launched the first earth resources satellite in 1972, remote sensing technology has received unprecedented attention around the world. The remote sensing image data has the characteristics of high precision, large coverage, and clear spectral resolution, and is favored by researchers. Object detection is an important part in the field of image processing. With the continuous development of remote sensing technology, whether in the military or civilian fields, the demand for detecting specific targets from remote sensing images is increasing day by day. essential technology. [0003] However, the rapid development of remote sensing...

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 Patents(China)
IPC IPC(8): G06V20/13G06V10/25G06V10/762G06V10/764G06V10/82G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06V20/13G06V10/25G06V10/462G06V2201/07G06N3/045G06F18/23213G06F18/24
Inventor 杨志钢黎明李泳江柳晴川杨远兰
Owner HARBIN ENG UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products