CN102800101A - Satellite-borne infrared remote sensing image airport ROI rapid detection method - Google Patents

Abstract

The invention relates to a satellite-borne infrared remote sensing image airport ROI rapid detection method which comprises the following steps of: firstly, converting a remote sensing image from an RGB color space into a Lab color space; then, performing Gaussian filtering of the remote sensing image to avoid noise influence; obtaining the difference between each pixel value and the image pixel average for the filtered image to obtain a remote sensing image differential chart; and finally, performing k-mean clustering segmentation on the remote sensing image differential chart to obtain airport ROI. The experimental result indicates that the method has high speed and high robustness, effectively reduces the difficulty in processing the remote sensing image, and has relatively great value and significance to the real-time detection of a remote sensing image airport target.

Description

A kind of spaceborne infrared remote sensing image airport ROI method for quick

Technical field

The invention belongs to the infrared remote sensing technical field of image processing, be specifically related to a kind of spaceborne infrared remote sensing image airport ROI method for quick,

Background technology

Increasing along with the fast development of satellite remote sensing technology and imaging data, people press for and can carry out Intelligent treatment to remote sensing images, therefrom detect interesting target quickly and accurately.The airport is as one type of specific objective, and its automatic detection has important practical value in fields such as aircraft navigation, military surveillance and precision strikes, receives the many concerns of People more and more.At present; Most detection algorithms are attempted to solve the detection problem on airport through the geometric properties of the linear feature that detects airfield runway, analysis runway; But exist a large amount of highways, river, artificial structure edge etc. to have and the similar linear feature in airport in the remote sensing images, only rely on linear feature can cause a large amount of flase drops, omission.Airport detection method based on the runway characteristic has certain limitation, and problem is detected on the airport that can't effectively solve in the remote sensing images.Also have one type of algorithm directly to adopt the method for image segmentation to obtain the candidate region on doubtful airport, then identification is carried out in the candidate region, further confirm the airport target zone original remote sensing images.Algorithm performance seriously relies on the effect of image segmentation, and speed is slow, efficient is low.

Summary of the invention

The technical matters that solves

For fear of the weak point of prior art, the present invention proposes a kind of spaceborne infrared remote sensing image airport ROI method for quick.

Technical scheme

A kind of spaceborne infrared remote sensing image airport ROI method for quick is characterized in that may further comprise the steps:

Step 1: with the infrared remote sensing image from the RGB color space conversion to the Lab color space, each locations of pixels is [L, a, b] in the Lab color space after the conversion ^TVector;

Step 2: the remote sensing images to the Lab color space after the conversion adopt Gauss operator to carry out convolution algorithm, obtain the image behind the gaussian filtering; The computing formula of said Gauss operator does

$G_{i,j}=\frac{1}{2{\mathrm{\π\σ}}^2}{e}^{-\frac{{(i-\frac{n+1}{2})}^2+{(j-\frac{n+1}{2})}^2}{2{\mathrm{\σ}}^2}}$

Wherein σ is a variance, and n is the dimension of Gauss operator nuclear matrix;

Step 3: each pixel value of the image behind the calculating gaussian filtering and the difference of image pixel mean value obtain the remote sensing images differential chart

S(x,y)＝||I _μ-I _G(x,y)||

Wherein, I _μBe the average of image pixel proper vector, I _G(x y) is corresponding pixel characteristic vector behind the gaussian filtering, || || be Euclidean distance;

Step 4: on the remote sensing images differential chart, make the k mean cluster and cut apart, the result of cutting apart according to the k mean cluster is partitioned on the remote sensing images differential chart and obtains airport ROI.

Said σ=3, n=3, k=2.

Beneficial effect

A kind of spaceborne infrared remote sensing image airport ROI method for quick that the present invention proposes at first transfers remote sensing images to the Lab color space from the RGB color space, then in the Lab color space; Remote sensing images are carried out gaussian filtering; Remove The noise, then ask for the difference of each pixel value and image pixel average in the image, obtain error image; Adopt the k mean cluster to cut apart to error image at last, obtain airport ROI.

Compared with prior art, the present invention is not directly cut apart original remote sensing images and is obtained the airport target candidate region, but it has been carried out some pre-service, cuts apart on the image after pre-service again, obtains airport ROI.The present invention has reduced calculated amount, has improved processing speed and accuracy of detection greatly, and practicality is very strong.

Description of drawings

Fig. 1 is realization flow figure of the present invention;

Fig. 2 is original infrared remote sensing image;

Fig. 3 is the figure as a result after the remote sensing images gaussian filtering among the present invention;

Fig. 4 carries out difference result calculated figure to Fig. 3;

Fig. 5 is the figure as a result that Fig. 4 is cut apart.

Embodiment

Combine embodiment, accompanying drawing that the present invention is further described at present:

The step of a kind of spaceborne infrared remote sensing image airport ROI method for quick is following:

Step 1: with the infrared remote sensing image from the RGB color space conversion to the Lab color space, each locations of pixels of the image of conversion is [L, a, b] ^TVector.Concrete steps are following:

Step a: to the XYZ color space, conversion formula does from the RGB color space conversion

$\left[\begin{array}{c}X\\ Y\\ Z\end{array}\right]=\left[\begin{array}{ccc}0.412& 0.358& 0.180\\ 0.213& 0.715& 0.072\\ 0.019& 0.119& 0.950\end{array}\right]\left[\begin{array}{c}R\\ G\\ B\end{array}\right]$

Step b: to the Lab color space, conversion formula does from the XYZ color space conversion

$L=\left\{\begin{array}{cc}166*{(Y/Y_n)}^{1/3}-16& Y/Y_n>0.008856\\ 903.3*(Y/Y_n)& Y/Y_n\≤0.008856\end{array}\right.$

a＝500*(f(X/X _n)-f(Y/Y _n))

b＝200*(f(Y/Y _n)-f(Z/Z _n))

Wherein: X _n, Y _n, Z _nBe the tristimulus values of white light;

$f\left(t\right)=\left\{\begin{array}{cc}{t}^{1/3}& t>0.008856\\ 7.787t+16/116& t\≤0.008856\end{array}\right.$

Step 2: adopt Gauss operator to carry out convolution algorithm to the remote sensing images that convert the Lab color space into, obtain the image behind the gaussian filtering, discrete Gauss operator nuclear matrix computing formula does

$G_{i,j}=\frac{1}{2{\mathrm{\π\σ}}^2}{e}^{-\frac{{(i-\frac{n+1}{2})}^2+{(j-\frac{n+1}{2})}^2}{2{\mathrm{\σ}}^2}}$

Wherein σ is a variance, and n is the dimension of gaussian kernel matrix;

Step 3: for the image behind the Gaussian Blur, ask the difference of each pixel value and image pixel mean value, obtain the remote sensing images differential chart;

S(x,y)＝||I _μ-I _G(x,y)||

Wherein, I _μBe the average of image pixel proper vector, I _G(x y) is corresponding pixel characteristic vector behind the gaussian filtering.|| || be Euclidean distance.

Step 4: on the remote sensing images differential chart, make the k mean cluster and cut apart, concrete steps are following:

Step a: set 2 initial barycenter arbitrarily and promptly cut apart gray-scale value f ₁, f ₂

Step b: be assigned to gray-scale value to each pixel p of remote sensing images differential chart with it in the cluster of nearest that barycenter representative, the background assignment is 0, and the target assignment is 255;

Step c: according to formula New barycenter f after the dispensed ₁, f ₂, N wherein _jBe number of pixels in the j class, the gray-scale value of g (x) representation class interior pixel p, j=1,2;

Steps d: repeating step b and step c are up to f ₁And f ₂No longer change, just can on the remote sensing images differential chart, be partitioned into airport ROI.

The hardware environment of specific embodiment is: Intel Duo 2 double-core 2.93G computing machines, 2.0GB internal memory, 512M video card, the software environment of operation is: Matlab R2011a, Windows XP.We have realized the method that the present invention proposes with Matlab software.Original remote sensing images have been selected an infrared remote sensing image that resolution is 400*400 for use.

Practical implementation of the present invention is following:

1, color space conversion: to the Lab color space, each locations of pixels of the image of conversion is [L, a, b] to just original infrared remote sensing image 2 from the RGB color space conversion ^TVector.Concrete switch process is:

1) from the RGB color space conversion to the XYZ color space, conversion formula does

$\left[\begin{array}{c}X\\ Y\\ Z\end{array}\right]=\left[\begin{array}{ccc}0.412& 0.358& 0.180\\ 0.213& 0.715& 0.072\\ 0.019& 0.119& 0.950\end{array}\right]\left[\begin{array}{c}R\\ G\\ B\end{array}\right]$

2) from the XYZ color space conversion to the Lab color space, conversion formula does

$L=\left\{\begin{array}{cc}166*{(Y/Y_n)}^{1/3}-16& Y/Y_n>0.008856\\ 903.3*(Y/Y_n)& Y/Y_n\≤0.008856\end{array}\right.$

a＝500*(f(X/X _n)-f(Y/Y _n))

b＝200*(f(Y/Y _n)-f(Z/Z _n))

Wherein: X _n, Y _n, Z _nBe the tristimulus values of white light.

$f\left(t\right)=\left\{\begin{array}{cc}{t}^{1/3}& t>0.008856\\ 7.787t+16/116& t\≤0.008856\end{array}\right.$

2, gaussian filtering: adopt Gauss operator that the remote sensing images that convert the Lab color space into are carried out the convolutional filtering computing, discrete gaussian kernel matrix computations formula does

$G_{i,j}=\frac{1}{2{\mathrm{\π\σ}}^2}{e}^{-\frac{{(i-\frac{n+1}{2})}^2+{(j-\frac{n+1}{2})}^2}{2{\mathrm{\σ}}^2}}$

Wherein σ is a variance, and n is the dimension of gaussian kernel matrix.Select σ=3 in this example for use, n=3 calculates the gaussian kernel matrix and it is carried out normalization, and the discrete gaussian kernel matrix that obtains after the normalization does

$\left[\begin{array}{ccc}0.1070& 0.1131& 0.1070\\ 0.1131& 0.1196& 0.1131\\ 0.1070& 0.1131& 0.1070\end{array}\right]$

Fig. 3 is Fig. 2 is adopted the figure as a result after this gaussian kernel is carried out convolutional filtering.

3, calculated difference figure: for the image behind the gaussian filtering 3, ask the difference of each pixel value and image pixel mean value, obtain the remote sensing images differential chart.

S(x,y)＝||I _μ-I _G(x,y)||

Wherein, I _μBe the average of image pixel proper vector, I _G(x y) is corresponding pixel characteristic vector behind the gaussian filtering.|| || be Euclidean distance.In this example, Fig. 4 carries out difference result calculated figure to Fig. 3.

4, cut apart: on remote sensing images differential chart 4, make the k mean cluster and cut apart, concrete steps are following:

Step a: set 2 initial barycenter arbitrarily and promptly cut apart gray-scale value f ₁, f ₂

Step b: be assigned to gray-scale value to each pixel p of remote sensing images differential chart with it in the cluster of nearest that barycenter representative, the background assignment is 0, and the target assignment is 255.

Step c: according to formula New barycenter f after the dispensed ₁, f ₂, N wherein _jBe number of pixels in the j class, the gray-scale value of g (x) representation class interior pixel p, j=1,2.

Steps d: repeating step b and step c are up to f ₁And f ₂No longer change, just can on the remote sensing images differential chart, be partitioned into airport ROI, obtain Fig. 5 as a result.

Claims (2)

1. spaceborne infrared remote sensing image airport ROI method for quick is characterized in that may further comprise the steps:

Step 1: with the infrared remote sensing image from the RGB color space conversion to the Lab color space, each locations of pixels is [L, a, b] in the Lab color space after the conversion ^TVector;

Step 2: the remote sensing images to the Lab color space after the conversion adopt Gauss operator to carry out convolution algorithm, obtain the image behind the gaussian filtering; The computing formula of said Gauss operator does

$G_{i,j}=\frac{1}{2{\mathrm{\π\σ}}^2}{e}^{-\frac{{(i-\frac{n+1}{2})}^2+{(j-\frac{n+1}{2})}^2}{2{\mathrm{\σ}}^2}}$

Wherein σ is a variance, and n is the dimension of Gauss operator nuclear matrix;

Step 3: each pixel value of the image behind the calculating gaussian filtering and the difference of image pixel mean value obtain the remote sensing images differential chart

S(x,y)＝||I _μ-I _G(x,y)||

Wherein, I _μBe the average of image pixel proper vector, I _G(x y) is corresponding pixel characteristic vector behind the gaussian filtering, || || be Euclidean distance;

Step 4: on the remote sensing images differential chart, make the k mean cluster and cut apart, the result of cutting apart according to the k mean cluster is partitioned on the remote sensing images differential chart and obtains airport ROI.

2. a kind of spaceborne infrared remote sensing image according to claim 1 airport ROI method for quick is characterized in that: said σ=3, n=3, k=2.

Images