CN104809726A - Change detection method based on multiscale geometrical characteristic vector - Google Patents

Abstract

The invention discloses a change detection method based on a multiscale geometrical characteristic vector. The method includes determining multiscale number and value and conducting multiscale image segmenting; searching for a scale correlation object in scale segmenting results with a basic unit as the inquiry condition, calculating the geometrical characteristics, constructing multiscale geometrical characteristic vector of a detection unit by the geometrical characteristics, and judging the change condition of the unit by measuring change intensity of the geometrical characteristic vector at different phases. The change detection method based on multiscale geometrical characteristic vector is provided. By means of the method, the detection basis is converted from the spectrum space to the scale space, change area detection is achieved through key steps of arithmetic sequence multiscale generation, multiscale geometrical characteristic vector construction, change intensity calculation based on geometrical characteristic vector and the like. The detection method is accurate.

Description

Based on the change detecting method of multiple dimensioned geometric properties vector

Technical field

The present invention relates to a kind of change detecting method, particularly relate to a kind of change detecting method based on multiple dimensioned geometric properties vector.

Background technology

Change detection refers to observes a certain entity or phenomenon at different times, and judges with this process that its state changes.The change Detection Information of precise and high efficiency is significant for research whole world change, spatial database renewal, emergent topographic support etc.(Malaysian is comprised sharp with conventional pixel level change detection algorithm, Sui Hai Gang, Li Pingxiang etc. the remote sensing image change based on kernel function measured similarity detects [J], Lu Miao, Mei Yang, Chen Lijun. utilize ground mulching change detecting method [J] and the J.Chen of spectrum value and shape optimum combination, M.Lu, X.Chen, et al.A spectral gradient difference based approach forland cover change detection [J]) compare, object-based image analysing computer (Object-basedImage Analysis, OBIA) spectrum of image can be fully utilized, texture, several how information, become the trend changing detect delay in recent years.The geometric properties of cutting object is utilized to be then one of important content based on object analysis.But existing method mainly adopts single yardstick to carry out Image Segmentation, to the rule of multiple multi-scale segmentation and applied research less.

Summary of the invention

Object of the present invention is just to provide a kind of change detecting method based on multiple dimensioned geometric properties vector to solve the problem.

The present invention is achieved through the following technical solutions above-mentioned purpose:

Based on a change detecting method for multiple dimensioned geometric properties vector, comprise the following steps:

(1) determine multiple dimensioned quantity and numerical value and carry out image fusion segmentation;

Comprising the multiple dimensioned generation of sequence: sequence is multiple dimensioned is the specific multiple dimensioned set that generates according to certain mathematical programming, build that sequence is multiple dimensioned can use multiple mathematical programming, comprise arithmetic sequence:

S＝{a ₀+i*Δ|i∈[0,n]}

In above formula: S represents the multiple dimensioned set of arithmetic sequence; A0 represents sequence initial value; I is the integer set from 0 to n; △ represents sequence tolerance;

Range scale refers to the scope between the multiple dimensioned set minimum value of arithmetic sequence and maximal value, and after determining range scale, sequence tolerance just decides the level of detail that multiple dimensioned geometric properties vector portrays metric space;

(2) be that search condition searches for yardstick affiliated partner in each multi-scale segmentation result with elementary cell, calculate its geometric properties and build the multiple dimensioned geometric properties vector describing this detecting unit thus;

Build comprising multiple dimensioned geometric properties vector:

Determine arithmetic sequence multiple dimensioned after, need to build multiple dimensioned geometric properties vector, vector cutting object belongs to two-dimentional area pattern in geometry dimension, usable floor area and girth describe the geometric properties of area pattern, describe the geometric properties of cutting object by the shape index in landscape ecology, its computing method are as follows:

$\mathrm{SI}=\frac{P}{4\×\sqrt{A}}$

In above formula, SI represents shape index, and A represents area, and P represents girth;

Calculation procedure comprises:

Each yardstick in using arithmetic sequence multiple dimensioned is split image;

From each multi-scale segmentation object, search package is containing the yardstick affiliated partner of current pixel position;

Calculate the geometric properties of all yardstick affiliated partners, and be combined into multiple dimensioned geometric properties vector;

(3) change intensity by measuring different phase geometric properties vector judges the situation of change of this unit;

Calculate comprising the change intensity based on multiple dimensioned geometric properties vector:

From the shape difference of multiple dimensioned geometric properties vector, use related coefficient to calculate the change intensity of the multiple dimensioned geometric properties vector of different phase, computing method are shown below:

$\begin{array}{c}\mathrm{CI}({\mathrm{MSGFV}}_{t_1},{\mathrm{MSGFV}}_{t_2})=1-\mathrm{Corelate}({\mathrm{MSGFV}}_{t_1},{\mathrm{MSGFV}}_{t_2})\\ =1-\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}({\mathrm{MSGFV}}_{t_1}^i-\stackrel{\‾}{{\mathrm{MSGFV}}_{t_1}})({\mathrm{MSGFV}}_{t_2}^i-\stackrel{\‾}{{\mathrm{MSGFV}}_{t_2}})}{\sqrt{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{({\mathrm{MSGFV}}_{t_1}^i-\stackrel{\‾}{{\mathrm{MSGFV}}_{t_1}})}^2}\sqrt{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{({\mathrm{MSGFV}}_{t_2}^i-\stackrel{\‾}{{\mathrm{MSGFV}}_{t_2}})}^2}}\end{array}$

In above formula, CI represents change intensity, MSGFV represents multiple dimensioned geometric properties vector, Corelate represents related coefficient, and t1 represents the 1st phase, and t2 represents the 2nd phase, n represents multiple dimensioned number, when the pattern curve of two phase multiple dimensioned geometric properties vectors is close, its correlation coefficient value is comparatively large, and change intensity value is then less; If instead shape difference is large, then correlation coefficient value is less, and change intensity value is then larger.

Beneficial effect of the present invention is:

The invention provides a kind of change detecting method based on multiple dimensioned geometric properties vector, the foundation detected is transformed into metric space from spectral space by the method, the committed steps such as the structure by the multiple dimensioned generation of arithmetic sequence, multiple dimensioned geometric properties vector, the change intensity calculating based on geometric properties vector realize the mensuration of region of variation, and change detecting method of the present invention is more accurate.

Accompanying drawing explanation

Fig. 1 is the change detection principle schematic based on multiple dimensioned geometric properties vector in the present invention;

Fig. 2 is the segmentation result schematic diagram of ETM image in 2002 in the present invention;

Fig. 3 and Fig. 4 is the change intensity result schematic diagram of three kinds of variation characteristics.

Embodiment

Below in conjunction with accompanying drawing, the invention will be further described:

As shown in Figure 1, a kind of change detecting method based on multiple dimensioned geometric properties vector of the present invention, comprises the following steps:

(1) determine multiple dimensioned quantity and numerical value and carry out image fusion segmentation;

Comprising the multiple dimensioned generation of sequence: sequence is multiple dimensioned is the specific multiple dimensioned set that generates according to certain mathematical programming, build that sequence is multiple dimensioned can use multiple mathematical programming, comprise arithmetic sequence:

S＝{a ₀+i*Δ|i∈[0,n]}

In above formula: S represents the multiple dimensioned set of arithmetic sequence; A0 represents sequence initial value; I is the integer set from 0 to n; △ represents sequence tolerance;

Range scale refers to the scope between the multiple dimensioned set minimum value of arithmetic sequence and maximal value, and after determining range scale, sequence tolerance just decides the level of detail that multiple dimensioned geometric properties vector portrays metric space;

(2) be that search condition searches for yardstick affiliated partner in each multi-scale segmentation result with elementary cell, calculate its geometric properties and build the multiple dimensioned geometric properties vector describing this detecting unit thus;

Build comprising multiple dimensioned geometric properties vector:

Determine arithmetic sequence multiple dimensioned after, need to build multiple dimensioned geometric properties vector, vector cutting object belongs to two-dimentional area pattern in geometry dimension, usable floor area and girth describe the geometric properties of area pattern, describe the geometric properties of cutting object by the shape index in landscape ecology, its computing method are as follows:

$\mathrm{SI}=\frac{P}{4\×\sqrt{A}}$

In above formula, SI represents shape index, and A represents area, and P represents girth;

Calculation procedure comprises:

Each yardstick in using arithmetic sequence multiple dimensioned is split image;

From each multi-scale segmentation object, search package is containing the yardstick affiliated partner of current pixel position;

Calculate the geometric properties of all yardstick affiliated partners, and be combined into multiple dimensioned geometric properties vector;

(3) change intensity by measuring different phase geometric properties vector judges the situation of change of this unit;

Calculate comprising the change intensity based on multiple dimensioned geometric properties vector:

From the shape difference of multiple dimensioned geometric properties vector, use related coefficient to calculate the change intensity of the multiple dimensioned geometric properties vector of different phase, computing method are shown below:

$\begin{array}{c}\mathrm{CI}({\mathrm{MSGFV}}_{t_1},{\mathrm{MSGFV}}_{t_2})=1-\mathrm{Corelate}({\mathrm{MSGFV}}_{t_1},{\mathrm{MSGFV}}_{t_2})\\ =1-\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}({\mathrm{MSGFV}}_{t_1}^i-\stackrel{\‾}{{\mathrm{MSGFV}}_{t_1}})({\mathrm{MSGFV}}_{t_2}^i-\stackrel{\‾}{{\mathrm{MSGFV}}_{t_2}})}{\sqrt{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{({\mathrm{MSGFV}}_{t_1}^i-\stackrel{\‾}{{\mathrm{MSGFV}}_{t_1}})}^2}\sqrt{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{({\mathrm{MSGFV}}_{t_2}^i-\stackrel{\‾}{{\mathrm{MSGFV}}_{t_2}})}^2}}\end{array}$

In above formula, CI represents change intensity, MSGFV represents multiple dimensioned geometric properties vector, Corelate represents related coefficient, and t1 represents the 1st phase, and t2 represents the 2nd phase, n represents multiple dimensioned number, when the pattern curve of two phase multiple dimensioned geometric properties vectors is close, its correlation coefficient value is comparatively large, and change intensity value is then less; If instead shape difference is large, then correlation coefficient value is less, and change intensity value is then larger.

Embodiment

Using ENVI EX 4.7 to realize the multi-scale division of remote sensing image, is 0 to 100 to the segmentation range scale of survey region.In order to careful description metric space, select numerical value less 5 as the multiple dimensioned tolerance of arithmetic sequence.Therefore, the computing method that arithmetic sequence is multiple dimensioned are by formula:

S＝{a ₀+i*Δ|i∈[0,n]}

Be converted into formula:

S＝{5+i*5|i∈[0,19]}

＝{5,10,15,L,95,100}

Fig. 2 shows the result that ENVI EX 4.7 pairs of test blocks ETM+ image in 2002 carries out 5,50,100 multi-scale segmentation.

The MSGFV change detecting method utilizing the present invention to propose calculates the change intensity of survey region.As shown in Figure 3, comprise area, change intensity image that girth, shape index three kinds of geometric properties calculate, the possibility of region representation change brighter in figure greatly, otherwise then less.Using statistics method sets change threshold, and by the mean value of change intensity image, this threshold value adds that 1.5 times of standard deviations obtain, what change intensity was greater than this threshold value is region of variation, and what be less than this threshold value is invariant region, and the result of three kinds of change detections as shown in Figure 4.The part of the main integrated distribution of region of variation and city expansion along the line in the Weihe River, close with the result of image visualization interpretation.

Adopt stratified random smapling method to have chosen constant sample 35 polygons (238 pixels) and change sample 16 polygons (128 pixels) in test block by high resolution image and other auxiliary information, then use confusion matrix to calculate overall accuracy and the Kappa coefficient (as shown in the table) of three kinds of change testing results.

The change testing result of the multiple dimensioned geometric properties vector calculation of Shape-based interpolation index is obviously better than the result of other two kinds of geometric properties, therefore using shape index result as the net result in this test block, therefore the present invention has better accuracy of detection.

Embodiment recited above is only be described the preferred embodiment of the present invention; not scope of the present invention is limited; do not departing under the present invention designs spiritual prerequisite; the various distortion that the common engineering technical personnel in this area make technical solution of the present invention and improvement, all should fall in protection domain that claims of the present invention determine.

Claims (9)

1., based on a change detecting method for multiple dimensioned geometric properties vector, it is characterized in that: comprise the following steps:

(1) determine multiple dimensioned quantity and numerical value and carry out image fusion segmentation;

(2) be that search condition searches for yardstick affiliated partner in each multi-scale segmentation result with elementary cell, calculate its geometric properties and build the multiple dimensioned geometric properties vector describing this detecting unit thus;

(3) change intensity by measuring different phase geometric properties vector judges the situation of change of this unit.

2. a kind of change detecting method based on multiple dimensioned geometric properties vector according to claim 1, it is characterized in that: in step (1), comprise the multiple dimensioned generation of sequence: sequence is multiple dimensioned is the specific multiple dimensioned set that generates according to certain mathematical programming, build that sequence is multiple dimensioned can use multiple mathematical programming.

3. a kind of change detecting method based on multiple dimensioned geometric properties vector according to claim 2, is characterized in that: the mathematical programming building the multiple dimensioned use of sequence comprises arithmetic sequence:

S＝{a ₀+i*Δ|i∈[0,n]}

In above formula: S represents the multiple dimensioned set of arithmetic sequence; A0 represents sequence initial value; I is the integer set from 0 to n; △ represents sequence tolerance.

4. a kind of change detecting method based on multiple dimensioned geometric properties vector according to claim 2, is characterized in that: range scale refers to the scope between the multiple dimensioned set minimum value of arithmetic sequence and maximal value.

5. a kind of change detecting method based on multiple dimensioned geometric properties vector according to claim 2, is characterized in that: after determining range scale, and sequence tolerance just decides the level of detail that multiple dimensioned geometric properties vector portrays metric space.

6. a kind of change detecting method based on multiple dimensioned geometric properties vector according to claim 1, it is characterized in that: in step (2), build comprising multiple dimensioned geometric properties vector: determine arithmetic sequence multiple dimensioned after, need to build multiple dimensioned geometric properties vector, vector cutting object belongs to two-dimentional area pattern in geometry dimension, usable floor area and girth describe the geometric properties of area pattern, describe the geometric properties of cutting object by the shape index in landscape ecology.

7. a kind of change detecting method based on multiple dimensioned geometric properties vector according to claim 5, is characterized in that: the computing method describing the geometric properties of cutting object are as follows:

In above formula, SI represents shape index, and A represents area, and P represents girth.

8. a kind of change detecting method based on multiple dimensioned geometric properties vector according to claim 6, is characterized in that: calculation procedure comprises:

Each yardstick in multiple dimensioned with arithmetic sequence is split image;

From each multi-scale segmentation object, search package is containing the yardstick affiliated partner of current pixel position;

Calculate the geometric properties of all yardstick affiliated partners, and be combined into multiple dimensioned geometric properties vector.

9. a kind of change detecting method based on multiple dimensioned geometric properties vector according to claim 1, is characterized in that: in step (3), and the change intensity comprised based on multiple dimensioned geometric properties vector calculates:

From the shape difference of multiple dimensioned geometric properties vector, use related coefficient to calculate the change intensity of the multiple dimensioned geometric properties vector of different phase, computing method are shown below:

In above formula, CI represents change intensity, MSGFV represents multiple dimensioned geometric properties vector, Corelate represents related coefficient, and t1 represents the 1st phase, and t2 represents the 2nd phase, n represents multiple dimensioned number, when the pattern curve of two phase multiple dimensioned geometric properties vectors is close, its correlation coefficient value is comparatively large, and change intensity value is then less; If instead shape difference is large, then correlation coefficient value is less, and change intensity value is then larger.