Detection method for building change by multispectral image

A multi-spectral image and change detection technology, applied in image analysis, image data processing, instruments, etc., can solve problems such as difficult to determine rules and cumbersome process, and achieve the effect of reducing false detection rate, obvious difference, and reducing dependence.

Active Publication Date: 2013-05-01
2 Cites 13 Cited by

AI-Extracted Technical Summary

Problems solved by technology

The use of object-oriented classification is mainly manual participation, and the process is cumbersome.
[0010] Wang Min et al. proposed an image change detection method based on multi-feature evidence fusion. By calculating the structural similarity of the texture,...
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

Method used

As can be seen from Table 2, relative to using pixel-level change detection alone and using texture feature change detection alone, the multispectral image building change detection method proposed by the present invention has a correct rate that has increased by about 15% and 7% respectively, The false detection rate has been significantly reduced, by about 41% and 31% respectively. Since the feature-level change detection is aimed at the change of the object of interest, the false detection rate is much lower than that of the pixel level. Combining the pixel level and the feature level can Reducing the impact of non-changing areas can achieve better detection results. Because the building shadow is only a small part on the image, the method for removing the shadow still has certain limitations; the method of the present invention has a certain improvement in accuracy compared with the multi-feature D-S fusion method, and the false detection rate also decreases to some extent; and object-oriented Compared with the comparison after classification, the correct rate of change detection is increased by 5% after the combination of pixel level and feature level, and the false detection rate is reduced by about 10%. To a certain extent, the change of ground object pairs similar to the building spectrum is reduced detection impact.
The experimental result of comparative analysis above six kinds of methods, can see, use pixel ratio method to carry out the change area range that change detection obtains the largest, it not only has included the change of building, and has included other features around the building From the two time-phase original images, it can be seen that the ground objects such as roads and vegetation around the building have changed greatly, and the change detection using the image pixel information alone cannot obtain the building change information. The effect of extracting texture features for building change detection is significantly better than the pixel ratio method. First, extract the texture features of buildings in two phases, and then compare the extracted building areas pixel by pixel for change detection. This method makes full use of The features of the building on the image are used to detect the change of the building, but there are some areas on the image with similar texture features and buildings. After threshold segmentation, these areas are retained as building areas and participate in the pixel-by-pixel comparison process. , these regions form part of the change detection spurious detection. Detecting changes in building shadows and removing the influence of shadows in images of texture feature changes can reduce the false detection rate to a certain extent. The influence of object shadows should be very small; in areas where buildings are densely arranged, building shadows are blocked by buildings, and the influence of shadows is limited. Effectively improve the detection effect. The comparison method after object-oriented classification first detects the change area by using the pixel ratio method on the image, and then uses the spectral characteristics of the image to classify the ground objects in the change area. Merge into one category, and then perform change detection. Due to the phenomenon of the same object with different spectra and different objects with the same spectrum in remote sensing images, it is easy to cause classification confusio...
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


The invention relates to a detection method for building change by a multispectral image. The detection method comprises the step of detecting the change based on a pixel ratio method to obtain change regions of all ground objects firstly, wherein the change regions are used as a change candidate region of the building. The change detection based on building feature is carried out by using the change regions detected by using a ratio method. False detection easily occurs by adopting one feature, so according to the method, feature change detection is carried out by adopting a mode of sequentially combining texture feature with tone feature to distinguish the change of the building from the change of other ground objects, and thus detection accuracy of the building is increased and the false detection rate is reduced.

Application Domain

Technology Topic

False detectionMultispectral image +4


  • Detection method for building change by multispectral image
  • Detection method for building change by multispectral image
  • Detection method for building change by multispectral image


  • Experimental program(1)

Example Embodiment

[0022] Examples:
[0023] 1. First, introduce the principle of pixel level change detection based on the pixel ratio method used in the present invention.
[0024] The most basic change of the building in the remote sensing image is the change of the pixel gray value. Therefore, the pixel gray value can also be regarded as the basic feature of the image. Pixel-level change detection is defined as the direct calculation of the pixel value of the two-phase image to construct a difference image and obtain the change area according to a certain discrimination rule. Commonly used pixel-level change detection methods mainly include pixel difference method, pixel ratio method, and image regression method. The pixel difference method is simple and straightforward to implement, but many small fragments will appear during change detection. The pixel ratio method can reduce the influence of the sun's angle and terrain during image acquisition, but the result of the ratio method is often non-normal distribution. The characteristic of the image regression method is to first establish the regression equation between the images and then subtract, which can reduce the influence of the atmosphere, the incident angle and the environmental difference, but it is necessary to obtain an accurate regression equation and select the appropriate band. The change detection method based on the pixel level of remote sensing image detects the change information of all features. It includes not only the changes of buildings, but also the changes of other features in the area where the building is located. The type of feature change is unknown, so how to effectively Suppressing the false detection of features around buildings is the main problem faced by pixel-level change detection.
[0025] Another important issue based on pixel-level change detection is how to distinguish the changing and non-changing areas of the difference image. By analyzing the difference image and its histogram, selecting an appropriate threshold to accurately distinguish whether each pixel belongs to the changing area or the non-changing area, and dividing the pixels of the difference image with 0 or 1 according to their gray value (where 0 represents Non-change, 1 means change, if it is converted into 8-bit gray value, it will be 0 and 255 respectively), extract the change information, from which the binary image representing the change of the feature of the difference image can be obtained.
[0026] The present invention adopts four methods of pixel ratio method, difference method, image regression method, and principal component analysis method for change detection, and compares these four methods. Based on the three indicators of correct rate, false detection rate and missed detection rate, the ratio method is the best method for building inspection.
[0027] The pixel-level change detection method in the present invention adopts the 3×3 small window matrix pixel ratio method proposed by Tang Puqian et al. The specific method is as follows: replace this pixel with a 3×3 small window matrix centered on a certain pixel in the image, and calculate the sum of all pixels in the small window as the ratio, and set such small window matrices on remote sensing images at different time phases. M1, M2, the ratio of the sum of all pixels in the small window matrix is ​​α, then the formula is:
[0028] α = X i , j = 1 3 M 1 / X i , j = 1 3 M 2 - - - ( 1 )
[0029] Obtain the ratio difference image of two time-phase multispectral remote sensing images. Since the gray value difference of the building before and after the change is large, and the gray value of the unchanged building is small, the threshold value can be set to 0.25, so that Pixels with a ratio α greater than 0.25 are marked as changed pixels and represented by 1; pixels with a ratio α less than 0.25 are marked as constant pixels, represented by 0. In this way, a binary image including all changes in the features is obtained, with 1 representing the changed area and 0 representing the unchanged area.
[0030] An objective evaluation method can be used to analyze the results of change detection. The objective evaluation indicators mainly include three objective evaluation indicators: the correct rate of change detection, the rate of false detection, and the rate of missed detection.
[0031] Correct rate of change detection:
[0032] P td =C td /C t (2)
[0033] Change detection false detection rate:
[0034] P fd =C fd /C d (3)
[0035] Change detection miss rate:
[0036] P ld =C ld /C t (4)
[0037] Where C t To be the sum of the real change pixels in the building area in the remote sensing image, the present invention first performs manual interpretation on two multispectral images of different time phases, gives the positions of all the buildings, and uses eCognition software to segment the images at a suitable scale. After extracting all the buildings and obtaining the building areas of two different time-phase multispectral images, use the method of pixel-by-pixel comparison to count the total number of pixels in the changing area of ​​the building, which is C t.
[0038] C d Is the sum of the detected building area change pixels, C td Is the number of change pixels in the building area detected in the real change pixels, C fd Is the number of falsely detected pixels that are actually unchanged but detected as changed in the building area, C ld It is the number of real changed pixels that are missed in the building area. The value of each index obtained by calculation is shown in Table 1.
[0039] Table 1 Objective evaluation results of change detection
[0040] Evaluation index
[0041] 2. Introduce the principle of feature level change detection adopted by the present invention.
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


no PUM

Description & Claims & Application Information

We can also present the details of the Description, Claims and Application information to help users get a comprehensive understanding of the technical details of the patent, such as background art, summary of invention, brief description of drawings, description of embodiments, and other original content. On the other hand, users can also determine the specific scope of protection of the technology through the list of claims; as well as understand the changes in the life cycle of the technology with the presentation of the patent timeline. Login to view more.
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

Similar technology patents

Classification and recommendation of technical efficacy words

  • Narrow down
  • Reduce the impact

Decellularized grafts from umbilical cord vessels and process for preparing and using same

InactiveUS20050203636A1Narrow downArtificial cell constructsCell culture supports/coatingTissue GraftCell type

Knee brace and method for securing the same

ActiveUS20060135901A1Reduce the impactMinimizing rotational forceNon-surgical orthopedic devicesMedian planeKnee braces
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