Height measurement method based on building shadow restoration

Abstract

The height measurement method based on building shadow recovery relates to the technical field of optical remote sensing image processing, and solves the problem of improving the building height measurement accuracy in high-resolution remote sensing images. The method comprises the following steps: extracting the building shadow in the remote sensing image through a semantic segmentation network, rotating the remote sensing image and the building shadow according to the sun azimuth so that the shadow line is perpendicular to the horizontal direction; dividing the rotated building shadow into a region adjacent to the top of the building, a region adjacent to the side of the building, and a region adjacent to the bottom of the building; recovering the building shadow in the divided regions to obtain the actual length of the recovered shadow; obtaining the actual length sequence of the recovered shadow along the horizontal direction, calculating the height sequence of the building; and calculating the height of the building according to the height sequence of the building. The present application realizes more accurate building height measurement.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of optical remote sensing image processing, and particularly relates to a height calculation method based on building shadow recovery. BACKGROUND

[0002] Building is an important target in geographic condition monitoring, and fast and accurate measurement of building height can bring huge economic value. Building height calculation based on high-resolution remote sensing images is a popular research topic in the fields of photogrammetry and remote sensing. For a long time, many scholars have done a lot of research on related issues. There are two categories, including extraction using stereo pairs and extraction using single remote sensing image shadow. The method of extracting building height using stereo pairs mainly identifies and obtains building height through digital elevation model (DEM) in urban three-dimensional modeling. In the research of obtaining height information using single image, most of them solve the height of the building according to the length of the building shadow. Since the cost of stereo relative data acquisition is high, it is not conducive to large-scale calculation and engineering progress monitoring scene, so the building height information is mainly obtained based on the length of the building shadow.

[0003] So far, in the research of obtaining building height information based on building shadow length by domestic and foreign researchers, there are mainly three ways for shadow extraction: first, the construction of geometric model; second, the texture information in the image; and third, the extraction of shadow through semantic segmentation network. When the height is calculated through the building shadow, the shadow is seriously blocked by the building when the satellite and the sun are on the same side of the building, and the undulating ground causes the shadow line to be not parallel to the building body surface, resulting in a large error in height measurement. The accurate extraction of building height depends more on the accurate extraction and optimization of shadow length. For example, related morphological concepts such as morphological operation, morphological shadow index, and enhanced morphological shadow index are used to improve the recognition accuracy. However, these optimizations are mainly based on the actual shadow area, and the study of the blocked part is limited, so the height calculation accuracy is still not high.

[0004] Therefore, in order to further improve the building height calculation accuracy in high-resolution remote sensing images, a new height calculation method based on building shadow recovery needs to be designed. SUMMARY

[0005] In view of the above problems, the present application provides a height calculation method based on building shadow recovery.

[0006] The technical scheme adopted by the present application to solve the technical problems is as follows:

[0007] The height calculation method based on building shadow recovery comprises the following steps:

[0008] Step 1, extracting the building shadow in the remote sensing image through a semantic segmentation network;

[0009] Step 2, rotating the remote sensing image and the building shadow obtained in step 1 according to the solar azimuth angle, so that the shadow line is perpendicular to the horizontal direction;

[0010] Step 3, dividing the rotated building shadow into region I and region III, or into region I, region II and region III, wherein region I is the region adjacent to the top of the building, region II is the region adjacent to the bottom of the building, and region III is the region adjacent to the side of the building;

[0011] Step 4, restoring the building shadow for the regions obtained in step 3 to obtain the actual length of the restored shadow;

[0012] Step 5, obtaining the actual length sequence of the restored shadow along the horizontal direction, and calculating the height sequence of the building according to the actual length sequence of the restored shadow according to formula (1);

[0013] H=S*tanω1 (1)

[0014] Wherein, S is the actual length of the shadow, ω1 is the solar elevation angle, and H is the height of the building;

[0015] Step 6, calculating the height of the building according to the height sequence of the building.

[0016] The beneficial effects of the present application are:

[0017] The present application makes full use of all shadows, first divides them according to the intersection with the building outline, then restores the shadows for different regions, and finally calculates the height of each column to obtain the final building height. The height measurement method based on building shadow restoration of the present application solves the limitation of building height measurement based on partial shadow blocking, and realizes more accurate building height measurement. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 The flowchart of the height measurement method based on building shadow restoration of the present application.

[0019] Figure 2 The schematic diagram of shadow zoning of step 3 of the height measurement method based on building shadow restoration of the present application.

[0020] Figure 3 The four different situations of shadow restoration in region (I) of the height measurement method based on building shadow restoration of the present application. DETAILED DESCRIPTION

[0021] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0022] The flowchart of the height calculation method based on building shadow restoration is as follows: Figure 1 As shown, it includes the following steps:

[0023] Step 1: Extract building shadows from remote sensing images using a semantic segmentation network;

[0024] Step 2: Rotate both the remote sensing image and the building shadows obtained in Step 1 according to the solar azimuth angle, so that the shadow lines on the sides of the buildings are perpendicular to the horizontal direction.

[0025] The remote sensing image is rotated according to the solar azimuth angle so that the direction of its shadow lines (i.e., the length direction of the shadow lines) is perpendicular to the horizontal direction, resulting in the rotated remote sensing image. Similarly, the building shadows are rotated according to the solar azimuth angle so that the direction of their shadow lines (length direction) is perpendicular to the horizontal direction, resulting in the rotated building shadows.

[0026] Step 3: Divide the rotated building shadow into regions I and III, or into regions I, II and III;

[0027] Area I is the area where the building's shadow is adjacent to the top of the building, Area II is the area where the building's shadow is adjacent to the bottom of the building, and Area III is the area where the building's shadow is adjacent to the side of the building;

[0028] Step 4: Based on the area obtained in Step 3, restore the building shadows obtained in Step 3 to obtain the actual length of the restored shadows;

[0029] For each region obtained in step 3, the shadow of the corresponding building is restored to obtain the actual length of the restored shadow;

[0030] Step 5: Based on the results obtained in Step 4, obtain the actual length sequence of the shadow pixel by pixel along the horizontal direction, and calculate the height sequence of the building according to the formula (1) based on the restored actual length sequence of the shadow.

[0031] H=S·tanω1 (1)

[0032] Where S is the actual length of the shadow, ω1 is the solar altitude angle, and H is the building height;

[0033] Step 6: Calculate the building height based on the building height sequence;

[0034] That is, the height sequence of buildings obtained in step 5 is sorted out and the building height is calculated.

[0035] As one embodiment, the implementation process of step 3 above includes the following sub-steps:

[0036] Step 3.1: Calculate the length of the building shadow in the direction perpendicular to the horizontal direction, pixel by pixel.

[0037] Step 3.2: List the sequence of vertical line segments whose lengths are separated from the building's shadow and the top of the building;

[0038] Draw a perpendicular line from a point on the building's shadow to the top of the building in the rotated remotely sensed image. The lengths of the resulting perpendicular lines form a sequence of perpendicular line segment lengths.

[0039] Step 3.3: The region corresponding to the monotonically decreasing or monotonically increasing length of the vertical segments in the vertical segment sequence is considered the region adjacent to the building shadow and the side of the building (Region III). That is, when the length of the vertical line continuously increases or decreases from left to right, it is Region III. The region corresponding to the vertical segment sequence whose lengths are all within a first length range is considered the region adjacent to the bottom of the building shadow (Region II). The first length range is an interval range and includes the maximum value MAX of the vertical segment sequence. For example, the region corresponding to the vertical segment sequence whose lengths are within the range of [MAX-4, MAX] is considered Region II, where MAX and 4 are the number of pixels. The remaining shadow of the building is considered the region adjacent to the top of the building shadow (Region I). That is, if Region II does not exist, the part of the building shadow excluding Region III is Region I; if Region II exists, the part of the building shadow excluding Regions II and III is Region I. The partitioning diagram is shown below. Figure 2 As shown, examples of building shadows obtained under two different lighting conditions are presented, along with the results of shadow partitioning.

[0040] The implementation process of step 4 above includes the following sub-steps:

[0041] Step 4.1: Restore region I using formula (2). The calculation method for this region is as follows: Figure 3 As shown;

[0042]

[0043] Where A is the part of the shadow that is exposed, α1 is the angle between the solar azimuth and the east-west line (latitude), α2 is the angle between the satellite azimuth and the east-west line, ω1 is the solar altitude angle, and ω2 is the satellite altitude angle; β1 is the angle between the shadow line corresponding to the top outline of the building and the east-west line, and -90°<β1≤90°; β2 is the angle between the top outline of the building that intersects with the extension of the shadow line and the east-west line, and -90°<β2≤90°.

[0044] Figure 3 ExampleFigure 2 (a) Four scenarios for shadow restoration in Region I, where N represents the length of the ground obscured by buildings. s This indicates the length of the ground shadow cast by the building.

[0045] Step 4.2: Use formula (3) to restore region III;

[0046] S=A (3)

[0047] Step 4.3: For region II, which is linearly decreasing from region III to region (I), we take formula (2) and formula (3) as two points of the straight line equation and obtain formula (4) as the formula for calculating the shaded length of the region.

[0048]

[0049] Where γ is the horizontal pixel position, γ1 is the minimum value of γ in the area where the building shadow is adjacent to the side of the building, and γ2 is the maximum value of γ in the area where the building shadow is adjacent to the side of the building.

[0050] The implementation process of step 6 above includes the following sub-steps:

[0051] Step 6.1: Round the building height sequence to the nearest integer, that is, round non-integer height values ​​to the nearest integer.

[0052] Step 6.2: After rounding, remove noise points from the height sequence, that is, remove height values ​​that appear less than a threshold. For example, the threshold is 5% of the total number of elements in the height sequence. Elements (height values) that appear less than 5% of the time are considered noise.

[0053] Step 6.3: Extract the level number of the maximum value in the height sequence after noise reduction;

[0054] Step 6.4: Obtain the original height value (the height value before rounding) according to the horizontal number, and use the original height value as the measured height of the building.

[0055] This invention makes full use of all shadows. First, the building is divided into zones according to the intersection of the shadows and the building's top outline. Then, the shadows of different zones are restored, and the height of each zone is calculated column by column to finally deduce the final building height. This invention's height calculation method based on building shadow restoration overcomes the limitations of measuring building height when partial shadows are obscured, achieving more accurate building height measurement.

[0056] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A height calculation method based on building shadow restoration, characterized in that, Includes the following steps: Step 1: Extract building shadows from remote sensing images using a semantic segmentation network; Step 2: Rotate both the remote sensing image and the building shadows obtained in Step 1 according to the solar azimuth angle so that the shadow lines are perpendicular to the horizontal direction. Step 3: Divide the rotated building shadow into regions I and III, or regions I, II and III. Region I is the area where the building shadow is adjacent to the top of the building, region II is the area where the building shadow is adjacent to the bottom of the building, and region III is the area where the building shadow is adjacent to the side of the building. Step 4: For the areas obtained in Step 3, restore the shadows of the buildings to obtain the actual length of the restored shadows; Step 5: Obtain the actual length sequence of the restored shadow pixel by pixel along the horizontal direction, and calculate the height sequence of the building according to formula (1) based on the actual length sequence of the restored shadow. （1） in, This is the actual length of the shadow. The solar altitude angle, The height of the building; Step 6: Calculate the building height based on the building height sequence; Step 3 includes the following steps: Step 3.1: Calculate the length of the building shadow in the direction perpendicular to the horizontal direction, pixel by pixel. Step 3.2: List the sequence of vertical line segments whose lengths are separated from the building's shadow and the top of the building; Step 3.3: Divide the building shadow into regions based on the vertical segment length sequence. The region corresponding to the segment with monotonically decreasing or monotonically increasing vertical length in the vertical segment sequence is regarded as Region III. The region corresponding to the segment with vertical length within a first length range in the vertical segment sequence is regarded as Region II. The first length range is an interval range and includes the maximum value MAX of the vertical segment sequence. The shadow of the building other than Region III and Region II is regarded as Region I. Step 4 includes the following steps: Step 4.1: For the area where the building's shadow is adjacent to the top of the building, use formula (2) to restore it; （2） in, The part that is exposed by the shadow. It is the angle between the solar azimuth and the east-west line. The angle between the satellite azimuth and the east-west line. The solar altitude angle, The satellite's elevation angle; The angle between the top outline segment intersecting the extension of the shaded line and the east-west line, and ; Step 4.2: Use formula (3) to restore the area where the building's shadow is adjacent to the bottom of the building; （3） Step 4.3: Use formula (4) to restore the area where the building's shadow is adjacent to the building's side. （4） in, The horizontal pixel position. The area where the building's shadow is adjacent to the building's side. The minimum value, The area where the building's shadow is adjacent to the building's side. The maximum value.

2. The height calculation method based on building shadow restoration as described in claim 1, characterized in that, The implementation process of step 6 includes the following sub-steps: Step 6.1: Round the building height sequence to the nearest integer, that is, round non-integer height values ​​to the nearest integer. Step 6.2: Remove noise points from the height sequence obtained in Step 6.1; Step 6.3: Extract the level number of the maximum value in the height sequence after noise reduction; Step 6.4: Obtain the original height value in the height sequence of the building according to the horizontal number, and use the original height value as the measured height of the building.

3. The height calculation method based on building shadow restoration as described in claim 2, characterized in that, Step 6.2 specifically involves removing height values ​​from the height sequence that appear less than a threshold, where the threshold is 5% of the total number of elements in the height sequence.

Images