Real scene three-dimensional processing method and system based on big data

Abstract

The present application relates to the technical field of three-dimensional modeling, in particular to a real scene three-dimensional processing method and system based on big data, which extracts height difference mutation and texture features based on remote sensing big data, fuses transparency and direction continuity information, dynamically adjusts boundary confidence level, and realizes weighted reconstruction of mountain texture layer and three-dimensional expression optimization. In the present application, an elevation grid is established by remote sensing big data and first-order difference is performed to identify the height difference mutation area, so that the mountain boundary identification process is changed from relying on image texture features to quantitative analysis of topographic changes, the boundary state is distinguished in combination with the transparency and gray contrast characteristics of the texture layer, so that the boundary judgment has the continuity and direction recognition ability, the joint evaluation is carried out based on the transparency change trend and the boundary direction fitting error, the dynamic adjustment of the boundary confidence is realized, and then the gray reconstruction is carried out on the layer boundary area according to the confidence adjustment result, and the accuracy of the layer splicing position and direction is enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of three-dimensional modeling, in particular to a real scene three-dimensional processing method and system based on big data. BACKGROUND

[0002] The technical field of three-dimensional modeling relates to the digital modeling of real-world objects or scenes, mainly including real scene data acquisition, model construction and reconstruction, three-dimensional data optimization, etc.

[0003] Among them, the real scene three-dimensional processing usually adopts a sparse point extraction method based on image matching to generate a feature point set, then generates a dense point cloud through dense matching, combines a perspective synthesis method to generate a triangular mesh patch, and finally completes the surface texture mapping of the model through a texture mapping method.

[0004] The prior art adopts an image matching method to extract sparse feature points and generate a dense point cloud in the real scene three-dimensional processing process. When facing complex terrain in mountainous areas, the elevation change cannot be fully expressed in the feature point extraction stage, and the image texture similar area is easy to cause mis-matching or sparse point cloud, thereby causing unclear boundaries or distorted geometry in the triangular mesh patch generation process. In addition, the model surface texture mapping relies on image synthesis and mapping processing, and lacks fine judgment of the transparency and direction continuity of the boundary between layers, especially in the image splicing area, which is easy to cause uneven brightness or direction deviation phenomenon, which is particularly obvious in areas with dramatic terrain undulations or poor texture continuity, causing poor model boundary fusion, visual inconsistency and local reconstruction distortion, which seriously affects the expression integrity and credibility of the three-dimensional model in the application scene. SUMMARY

[0005] The purpose of the present application is to solve the shortcomings in the prior art, and a real scene three-dimensional processing method and system based on big data are proposed.

[0006] In order to achieve the above-mentioned purpose, the present application adopts the following technical scheme: a real scene three-dimensional processing method based on big data, comprising the following steps:

[0007] S1: acquiring the elevation data of the target mountainous area through remote sensing big data and constructing a grid distribution, performing first-order difference on the central elevation value in the adjacent grid, and screening a high difference mutation set;

[0008] S2: acquiring the surface texture image of the mountain through remote sensing big data, constructing a mountain texture layer, judging the transparency of the boundary area of the mountain texture layer based on the high difference mutation set, and constructing a fusion state label set;

[0009] S3: based on the fusion state label set, analyzing the direction continuity of the boundary area of the mountain texture layer, and marking the discontinuous area as a boundary direction abnormal set;

[0010] S4: Re-adjust the confidence level of each boundary region based on the change of the abnormal boundary direction in the abnormal boundary direction set and the boundary region transparency state in the fusion state label set, to obtain a confidence adjustment node set;

[0011] S5: Based on the confidence adjustment node set, the mountain texture layer of the target mountain region is weighted and reconstructed to obtain a mountain real scene three-dimensional processing result.

[0012] As a further scheme of the present application, the height difference mutation set includes mutation region position identification, height difference mutation point number sequence, and difference calculation reference grid number, the fusion state label set includes transparency judgment result mark, fusion boundary attribute classification, and mutation region association relationship, the boundary direction abnormal set includes direction discontinuity section index, abnormal boundary point set number, and direction change trend feature, the confidence adjustment node set includes boundary point confidence level adjustment result, fusion participation identification type, and confidence level grading basis, and the mountain real scene three-dimensional processing result includes fusion boundary gray output result, processed layer data index, and three-dimensional fusion state identification.

[0013] As a further scheme of the present application, the height difference mutation set includes mutation region position identification, height difference mutation point number sequence, and difference calculation reference grid number, the fusion state label set includes transparency judgment result mark, fusion boundary attribute classification, and mutation region association relationship, the boundary direction abnormal set includes direction discontinuity section index, abnormal boundary point set number, and direction change trend feature, the confidence adjustment node set includes boundary point confidence level adjustment result, fusion participation identification type, and confidence level grading basis, and the mountain real scene three-dimensional processing result includes fusion boundary gray output result, processed layer data index, and three-dimensional fusion state identification.

[0014] S111: Obtain the height data of the target mountain region through remote sensing big data and construct a grid distribution, generate a standard regular grid in the mountain range based on a unified projection coordinate system, extract the height value corresponding to the center position in each grid, and generate a grid height value sequence;

[0015] S112: Determine the corresponding relationship of adjacent grid center points according to the arrangement order of the grids in the grid height value sequence, perform first-order difference processing on the height values between adjacent points in sequence, and obtain the adjacent grid height difference change rate;

[0016] S113: Based on the adjacent grid height difference change rate, compare with the mountain region slope threshold value for terrain grading, identify the boundary position of the height difference mutation, and obtain the height difference mutation set.

[0017] As a further scheme of the present application, the height difference mutation set includes mutation region position identification, height difference mutation point number sequence, and difference calculation reference grid number, the fusion state label set includes transparency judgment result mark, fusion boundary attribute classification, and mutation region association relationship, the boundary direction abnormal set includes direction discontinuity section index, abnormal boundary point set number, and direction change trend feature, the confidence adjustment node set includes boundary point confidence level adjustment result, fusion participation identification type, and confidence level grading basis, and the mountain real scene three-dimensional processing result includes fusion boundary gray output result, processed layer data index, and three-dimensional fusion state identification.

[0018] S211: Obtain the mountain surface texture image through remote sensing big data, call the region image corresponding to the height data, perform geometric correction and brightness equalization processing, complete the map splicing according to the unified coordinate system, and construct the mountain texture layer aligned with the target region;

[0019] S212: based on the spatial coincidence relationship between the mountain texture layer and the elevation mutation set, the corresponding layer boundary region in the mountain texture layer is located, the transparency gradient value and the gray contrast value of each region boundary pixel are extracted, linear interpolation is performed on the two parameters, and the boundary transparency interpolation result is generated;

[0020] S213: according to the numerical relationship between the boundary transparency interpolation result and the average gray value of the layer boundary region, the gray contrast reference value is called as a judgment standard, the region greater than the judgment standard is marked as a transparency contraction state, and the remaining region is marked as a transparency expansion state, and a fusion state label set is generated.

[0021] As a further scheme of the application, the boundary direction anomaly set acquisition step is specifically:

[0022] S311: based on the fusion state label set, the pixel point sequence of the marked boundary region in the mountain texture layer is extracted, the direction angle sequence between adjacent points is constructed according to the boundary point arrangement order, and the boundary direction angle change data is formed;

[0023] S312: according to the change trend of the direction angle of each point in the boundary direction angle change data, the boundary main direction fitting is performed through the Hough transform algorithm, the direction deviation between each boundary point and the fitting curve is calculated, and the boundary direction fitting error is generated.

[0024] S313: based on the boundary direction fitting error, compared with the direction deviation threshold value, the boundary points with error value exceeding the direction deviation threshold value are screened, classified and marked according to the position index of the target boundary point, and the boundary direction anomaly set is acquired.

[0025] As a further scheme of the application, the confidence node set acquisition step is specifically:

[0026] S411: according to the position of the abnormal boundary point in the boundary direction anomaly set, the direction fitting error corresponding to each abnormal point is extracted, and the transparency state information of the boundary region in the same position in the fusion state label set is combined to construct a direction and transparency joint discrimination condition;

[0027] S412: based on the direction fitting error and the transparency state of each boundary point in the direction and transparency joint discrimination condition, the confidence level adjustment rules in the transparency contraction and expansion states are called, the confidence level of the boundary point with the direction fitting error value greater than the error threshold value and the transparency state of contraction is down-regulated, the confidence level of the boundary point with the direction fitting error value less than the error threshold value and the transparency state of expansion is up-regulated, and the boundary point confidence level adjustment result is generated.

[0028] S413: According to the confidence level change of the boundary point in the boundary point confidence level adjustment result, record the boundary point position index and the level value after the confidence level adjustment is completed, and obtain the confidence adjustment node set.

[0029] As a further scheme of the present application, the obtaining step of the mountain real scene three-dimensional processing result is specifically:

[0030] S511: Based on the confidence adjustment node set, extract the confidence level, transparency state and direction fitting error corresponding to each boundary point, locate the fusion boundary area in the mountain texture layer which coincides with the height difference mutation set, and generate boundary fusion information;

[0031] S512: According to the boundary fusion information, call the confidence level as the main fusion weight, the direction fitting error as the position correction factor, and the transparency state to limit the boundary adjustment range, construct the weighted combination parameters of each pixel point in the fusion area, and generate the fusion boundary gray correction result;

[0032] S513: Write the spatial position corresponding to the fusion boundary gray correction result into the corresponding fusion boundary area in the mountain texture layer, obtain the mountain real scene three-dimensional processing result, and use it to correct the layer splicing position and direction trend of the fusion boundary area in the mountain texture layer.

[0033] The real scene three-dimensional processing system based on big data is used to execute the real scene three-dimensional processing method based on big data, and the system comprises:

[0034] The height difference mutation extraction module obtains the elevation data of the target mountain area through remote sensing big data and constructs a grid distribution, performs first-order difference on the central elevation value in the adjacent grid, and selects a height difference mutation set;

[0035] The texture fusion discrimination module obtains the mountain surface texture image through remote sensing big data, constructs a mountain texture layer, judges the transparency of the boundary area of the mountain texture layer based on the height difference mutation set, and constructs a fusion state label set;

[0036] The boundary direction analysis module analyzes the direction continuity of the boundary area of the mountain texture layer based on the fusion state label set, and marks the discontinuous area as a boundary direction abnormal set;

[0037] The confidence level adjustment module adjusts the confidence level of each boundary area according to the change of the abnormal boundary direction in the boundary direction abnormal set and the transparency state of the boundary area in the fusion state label set, and obtains a confidence adjustment node set;

[0038] The layer reconstruction output module performs weighted reconstruction on the mountain texture layer of the target mountain area based on the confidence node set, and obtains a mountain real scene three-dimensional processing result.

[0039] Compared with the prior art, the application has the advantages and positive effects that:

[0040] In the application, the elevation grid is established through remote sensing big data and first-order difference is performed to identify the high-difference mutation area, so that the mountain boundary identification process is changed from relying on image texture features to quantitative analysis of terrain changes, the boundary state is distinguished in combination with the transparency and gray contrast characteristics of the texture layer, the boundary judgment has continuous and directional recognition ability, the confidence degree is dynamically adjusted based on the joint evaluation of the transparency change trend and the boundary direction fitting error, the identification accuracy of the abnormal area is improved, and then the gray reconstruction is performed on the layer boundary area according to the confidence adjustment result, so as to fuse the transparency state and direction correction information of the boundary area, enhance the accuracy of the layer splicing position and direction, and significantly improve the problems of boundary blur and inconsistent direction caused by image splicing through the whole-process terrain change-based boundary analysis and texture layer optimization strategy, improve the realness of the three-dimensional real scene model while optimizing the continuity and accuracy of the fusion area, and ensure that the mountain three-dimensional expression has higher matching rationality and visual consistency in structure and texture. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 It is a workflow schematic diagram of the application;

[0042] Figure 2 It is a flowchart of step S1 of the application;

[0043] Figure 3 It is a flowchart of step S2 of the application;

[0044] Figure 4 It is a flowchart of step S3 of the application;

[0045] Figure 5 It is a flowchart of step S4 of the application;

[0046] Figure 6 It is a flowchart of step S5 of the application. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical scheme and advantages of the application clearer, the application will be further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the application and do not limit the application.

[0048] In the description of the present application, it should be understood that the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship shown in the drawings, which are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation of the present application. In addition, in the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise specifically limited.

[0049] Referring to Figure 1 The present application provides a technical solution: a real scene three-dimensional processing method based on big data, comprising the following steps:

[0050] S1: Obtain the elevation data of the target mountainous area by remote sensing big data and construct a grid distribution, perform first-order difference on the central elevation value in the adjacent grid, and screen a high difference mutation set;

[0051] S2: Obtain the mountain surface texture image by remote sensing big data, construct a mountain texture layer, judge the transparency of the boundary area of the mountain texture layer based on the high difference mutation set, and construct a fusion state label set;

[0052] S3: Based on the fusion state label set, analyze the direction continuity of the boundary area of the mountain texture layer, and mark the discontinuous area as a boundary direction anomaly set;

[0053] S4: According to the change of the abnormal boundary direction in the boundary direction anomaly set and the transparency state of the boundary area in the fusion state label set, adjust the confidence level of each boundary area, and obtain a confidence adjustment node set;

[0054] S5: Based on the confidence adjustment node set, the mountain texture layer of the target mountainous area is weighted and reconstructed to obtain a mountain real scene three-dimensional processing result;

[0055] The high difference mutation set includes mutation region position identification, high difference mutation point number sequence, difference calculation reference grid number, the fusion state label set includes transparency judgment result mark, fusion boundary attribute classification, mutation region association relationship, the boundary direction anomaly set includes direction discontinuity section index, abnormal boundary point set number, direction change trend feature, the confidence adjustment node set includes boundary point confidence level adjustment result, fusion participation identification type, confidence level grading basis, and the mountain real scene three-dimensional processing result includes fusion boundary gray output result, processed layer data index and three-dimensional fusion state identification.

[0056] Referring to Figure 2 The acquisition step of the high difference mutation set is specifically:

[0057] S111: Obtain elevation data of the target mountainous area through remote sensing big data and construct a grid distribution. Generate a standard regular grid within the mountainous area based on a unified projection coordinate system, extract the elevation value corresponding to the center position in each grid, and generate a grid elevation value sequence.

[0058] To acquire elevation data of a target mountainous area and construct a grid distribution using remote sensing big data, the first step is to define the boundary of the target area. Taking a typical mountainous region as an example, such as a mountainous area with significant topographic relief, digital elevation model (DEM) data of the area is acquired using a remote sensing image platform. A sampling interval of 30 meters is selected as the resolution. Geographic information processing (GIS) software is used to uniformly set the coordinate system to a common projection system, and a standardized regular grid is constructed with an interval of 50 meters. The entire mountainous area is divided into continuous grid units, with each grid corresponding to a square area. Then, the coordinates of the center position of each grid are extracted, and the corresponding elevation value is queried in the DEM. For example, when the center point is located on a ridgeline, the elevation reading is 1280 meters, and when it is located in a valley, it is 960 meters. This extraction process relies on coordinate matching methods. The software's built-in spatial indexing function is used to quickly locate and read the elevation data. After the elevation values ​​corresponding to all grid center points are arranged in spatial order, a complete elevation value sequence is constructed and recorded in the database.

[0059] S112: Based on the arrangement order of the grids in the grid elevation value sequence, determine the correspondence between the center points of adjacent grids, and perform first-order difference processing on the elevation values ​​between adjacent points in sequence to obtain the rate of change of elevation difference between adjacent grids;

[0060] Based on the obtained grid elevation value sequence, the spatial relationship between adjacent grids is determined by arranging them in row and column order. Horizontally adjacent grids are grouped together, and vertically adjacent grids are also grouped together. Then, the elevation values ​​of the center points of each group of adjacent grids are differentially processed: the elevation of the next grid is subtracted from the elevation of the previous grid, and then divided by the distance between the two points to obtain the rate of change of elevation per unit distance. For example, in two adjacent grids, the elevation of the first grid is 1010 meters, and the second is 1035 meters, with a distance of 50 meters between them. The difference is 25 meters, and the rate of change is 25 divided by 50, resulting in a value of 0.5. This value indicates that the elevation changes by 0.5 meters per meter in that direction. If the value is negative, it indicates a decrease in elevation. The rate of change between all adjacent grids is calculated in this way and recorded in both the horizontal and vertical directions, forming a complete rate of change dataset.

[0061] S113: Based on the elevation difference change rate between adjacent grids, compare it with the slope threshold of mountainous areas used for terrain classification to identify the boundary location of elevation difference abrupt changes and obtain the set of elevation difference abrupt changes;

[0062] After obtaining the height difference change rate of all adjacent grids, it is necessary to compare with the slope threshold defined in the terrain classification to judge whether there is mutation. The threshold is usually set by experience, for example, in typical mountain analysis, it is set to 0.3, which means that the height change in unit distance is greater than 15 meters, which is a mutation. Read each group of adjacent grid change rate data in turn, compare it with the threshold, when a group of change rate value is greater than 0.3, it is judged as a height difference mutation boundary point, and the corresponding spatial position of the point is recorded as a mutation point, for example, in a pair of adjacent grids, the change rate is 0.38, which is greater than the preset threshold, so this group of points is classified as a mutation point, then all the mutation point positions are summarized, visualized by using graphics software, and the mutation boundary line is marked by superimposing on the map, and multiple continuous mutation points are connected to form a linear distribution, and the position change of the mountain contour is further judged.

[0063] Please refer to Figure 3 The acquisition step of the fusion state label set is specifically:

[0064] S211: Obtain the mountain surface texture image through remote sensing big data, call the regional image corresponding to the elevation data, perform geometric correction and brightness equalization processing, complete the map splicing according to the unified coordinate system, and construct the mountain texture layer aligned with the target area;

[0065] To obtain the mountain surface texture image, first select the image range consistent with the target elevation area, obtain the multi-spectral image of the corresponding area through remote sensing satellite, and call the digital elevation model corresponding to the area for matching processing. After loading in the GIS platform, geometric correction is performed. During the correction process, ground control points are set by selecting landmark features such as road intersections and peaks in the actual geographical position to ensure that the image coordinates are accurately projected in the unified coordinate system, complete the geometric deformation correction, then perform brightness equalization processing on each image block, adjust the overall brightness and contrast of the image using the histogram equalization algorithm, eliminate the light differences caused by different times and angles of shooting, then splice the maps in a spatial superposition manner according to the coordinate boundaries, use the overlapping area pixels between the splicing points for weighted fusion, construct a mountain texture layer covering the complete target area, and ensure that the layer is aligned with the digital elevation data in terms of pixel accuracy and coordinate reference.

[0066] S212: Based on the spatial coincidence relationship between the mountain texture layer and the height difference mutation set, locate the corresponding layer boundary area in the mountain texture layer, extract the transparency gradient value and gray contrast value of each boundary pixel, perform linear interpolation on the two parameters, and generate the boundary transparency interpolation result;

[0067] After the completion of the mountain texture layer construction, the system loads the spatial coordinate information of all the boundary points in the previously obtained elevation mutation set, retrieves the image pixel area corresponding to these coordinate points in the layer, marks these boundary areas as the feature extraction range, and in the image analysis tool, sequentially extracts two key parameters of each boundary pixel, namely the transparency gradient value and the gray contrast value. The method for obtaining the transparency gradient value is as follows: select the current pixel, calculate the transparency difference value with the four adjacent pixels above, below, left and right respectively, and take the absolute value to obtain the average. In the example, the transparency of the current pixel is 60, and the adjacent pixels are 55, 63, 58 and 61, respectively. The difference values are 5, 3, 2 and 1, respectively. The absolute value is taken to obtain the transparency gradient value: (5+3+2+1) ÷ 4=2.75.

[0068] The method for obtaining the gray contrast value is similar. The gray value of the current pixel is extracted and the difference between the gray values of the adjacent pixels is calculated to obtain the degree of gray change of the current pixel and the surrounding pixels. For example, the gray value of the current pixel is 130, and the adjacent pixels are 125, 140, 128 and 135, respectively. The corresponding difference values are 5, 10, 2 and 5, respectively. The gray contrast value is: (5+10+2+5) ÷ 4=5.5.

[0069] Subsequently, the transparency gradient value and the gray contrast value are subjected to linear interpolation processing. In order to balance the role of these two features in image boundary feature recognition, it is necessary to set the interpolation weight coefficient. Here, according to the actual image recognition test experience, the transparency gradient weight is set to 0.4 and the gray contrast weight is set to 0.6. The linear interpolation calculation is as follows: interpolation value=(transparency gradient value x transparency weight)+(gray contrast value x gray weight), interpolation value=(2.75 x 0.4)+(5.5 x 0.6), interpolation value=1.1+3.3=4.4. The result indicates that the transparency interpolation value of the current boundary pixel is 4.4, which is written into the layer attribute table. The above extraction and interpolation process is repeated for all boundary pixels to form a complete boundary transparency interpolation layer, which is completely aligned with the original image position in the layer structure and used for subsequent judgment of the region fusion state change.

[0070] S213: According to the numerical relationship between the boundary transparency interpolation result and the average gray value of the layer boundary region, the gray contrast reference value is called as the judgment standard, the region greater than the judgment standard is marked as the transparency contraction state, and the remaining region is marked as the transparency expansion state, and a fusion state label set is generated;

[0071] After generating the boundary transparency interpolation layer, the system compares the interpolation value of each pixel in the layer with the average gray value of the boundary region where the pixel is located. To this end, first, each group of boundary pixel regions is divided by block, for example, a 50-meter grid or an equilateral boundary aggregation block is taken as a group unit, and the average of the gray values of all boundary pixels in each block region is calculated. Then, the interpolation value of each pixel calculated before is compared with the average gray value of the region. The basis for judgment is the gray contrast reference value set by the system, which is usually based on the overall gray statistical range of the target layer. For example, through the analysis of the gray distribution of the whole image, it is found that the common contrast mutation demarcation point is about 30, so the reference value is set to 30. The judgment rule is: if the difference between the interpolation result and the average gray value is greater than the reference value, it is marked as transparency contraction state; otherwise, it is marked as transparency expansion state. The state of each pixel or region will be written into the fusion state label table in the form of marking, recorded as "contraction" or "expansion" state respectively, and a one-to-one correspondence between the label and the pixel position of the layer is established. The system performs the above processing on the boundary regions of the whole image in batches, and finally outputs a set of fusion state label data, which describes the transparency space change attribute of the boundary region and can be used as the basis for subsequent layer superposition or feature synthesis.

[0072] Referring to Figure 4 The acquisition steps of the boundary direction anomaly set are as follows:

[0073] S311: Based on the fusion state label set, the pixel point sequence of the marked boundary region in the mountain texture layer is extracted, the direction angle sequence between adjacent points is constructed according to the arrangement order of the boundary points, and the boundary direction angle change data is formed;

[0074] First, all the marked boundary areas in the mountain texture layer are identified, the boundary pixels in these areas are extracted in turn to form a continuous pixel sequence, the extraction order is determined according to the boundary direction, and the pixels are connected in a clockwise path from the starting point to the ending point to ensure that the point sequence is continuous in space without breakpoints. The system automatically removes duplicate or isolated pixels to ensure the logical integrity of the closed or open boundary line. After the extraction is completed, the system reads the row and column positions of each adjacent pixel pair in the image coordinates, judges the connection direction between the current point and the next point, and takes the angle formed by the reference east direction (i.e. horizontal right) as the direction angle value. The direction angle ranges from 0 degrees to 360 degrees, and is measured in a counterclockwise direction. For example, moving from the current pixel to the right side pixel, the direction angle is 0 degrees; moving upward is 90 degrees; moving to the left is 180 degrees; and moving downward is 270 degrees. If the current pixel position is a point in the image, the next pixel is offset to the right and up, and the angle between 0 degrees and 90 degrees is considered a "small angle"; otherwise, if the direction angle is between 180 degrees and 270 degrees, it means that the direction is offset to the left and down, which is a "large angle". For example, if the current pixel is located at row 100 and column 150 in the image, the next pixel is at row 99 and column 153, the system judges that the moving direction is right and up, and the direction angle is about 30 degrees, which is a small angle; if the next pixel is at row 103 and column 147, the direction is offset to the left and down, and the system identifies the direction angle as about 210 degrees, which is a large angle. In this way, the system judges the angle of all adjacent pixel groups and records the direction angles in sequence according to the original boundary path, and finally forms a complete boundary direction angle sequence dataset. This dataset provides basic information about the direction change of each point on the boundary in the image space.

[0075] S312: According to the trend of the direction angle of each point in the boundary direction angle change data, the main direction of the boundary is fitted by the Hough transform algorithm, the direction deviation between each boundary point and the fitted curve is calculated, and the boundary direction fitting error is generated;

[0076] After the boundary direction angle sequence is constructed, the system evaluates the direction change trend of each boundary pixel to extract the main direction of the overall boundary by performing a weighted Hough transform fitting analysis. The core of this algorithm is based on the mapping relationship between the image space coordinates and the angle space, combined with the transparency interpolation value of each pixel in the fusion state label, and the weight factor is assigned through normalization processing. The basis for setting this weight is that the higher the transparency value, the more significant the change of the image boundary, and a higher response weight should be given; otherwise, a lower weight is given. The setting method is linear normalization, which maps all transparency interpolation values to between 0 and 1, so that the weight can be directly used for weighted processing in the continuous angle space.

[0077] The Hough transform accumulation formula is as follows:

[0078] ;

[0079] where, : the vote value of the accumulator at position, unitless, representing the weighted accumulation intensity; : the shortest distance between the line and the origin in the image, unit: pixel; : the direction of the line in the angle space, unit: degree, range from 0 to 180; : the image coordinate of the i-th boundary pixel, unit: pixel; : the normalized transparency interpolation weight of the point, range from 0 to 1, unitless; : the unit impulse function (also known as the pulse function), which outputs 1 when the expression inside is 0, and 0 otherwise, used to achieve exact matching accumulation. The function of this accumulation function is to map all possible line parameters of each boundary point in the image at a specific angle to the parameter space, and then

[0080] statistically strengthen the voting mechanism using the transparency-derived weight in two-dimensional space, so as to find the direction with the maximum accumulation value in the parameter space as the main direction. Suppose the transparency interpolation values of five boundary pixel points are: 4.4, 3.9, 4.1, 4.7, and 5.0. The system identifies that the minimum value is 3.5 and the maximum value is 5.0, and the normalization process is as follows: For the interpolation value of 4.4, the normalized weight is:

[0081]

[0082] ;

[0083] ;

[0084] In the weighted Hough transform process, suppose the coordinates of the point are 100 pixels horizontally and 150 pixels vertically, unit: pixel, and the current traversal angle is 60 degrees. Using the trigonometric function to calculate:

[0085] pixel;

[0086] Add the weight 0.6 to the accumulator at position , and repeat the execution for all pixels and angles. Finally, select the angle position with the maximum accumulation value as the main direction. Suppose that among all angles, the accumulation value of 63 degrees is the maximum, then the system sets 63 degrees as the main direction of the boundary.

[0087] ​​Next, the system calculates the absolute value of the difference between the original direction angle and the main direction angle of each boundary point, i.e. the direction deviation. For example, if the direction angle of a certain boundary point is 58 degrees, the deviation is:

[0088] ;

[0089] In this way, the system completes the calculation of the direction fitting error of all points, obtaining a complete error sequence, which is used for comparison with the deviation threshold in the next stage to screen out abnormal points with prominent direction changes.

[0090] By extracting the direction angle and coordinate information of each boundary pixel point, and combining the transparency interpolation value for normalization processing, each point is assigned a corresponding weight value, which is used to enhance the influence of sudden points in the weighted Hough transform. Then, by traversing the angle and distance combinations in the angle space, the possible corresponding straight line directions of each point in the space are counted in a weighted cumulative manner, and finally a two-dimensional accumulator matrix containing the cumulative intensity of all directions is formed, from which the direction with the highest cumulative value is selected as the main direction. This main direction represents the trend result that most boundary points in the image tend to in space. Then the system calculates the angle difference between each boundary point and the main direction to generate the direction fitting error, which describes whether the boundary point deviates from the main trend. The purpose of the formula calculation result is to provide a numerical and comparable measure of direction deviation, so that subsequent numerical threshold can be used to screen out boundary points with obvious deflection.

[0091] S313: Based on the boundary direction fitting error, compare with the direction deviation threshold, screen out the boundary points with error value exceeding the direction deviation threshold, classify and mark according to the position index of the target boundary point, and obtain the boundary direction anomaly set;

[0092] First, a unified direction deviation threshold is set, which is usually set according to the allowed direction fluctuation range of the boundary structure. For example, 5 degrees is used as the judgment limit, which means that if the deviation between the direction of any boundary point and the main direction is greater than 5 degrees, it is considered as a direction abnormal point. Then the system reads the direction fitting error value of each boundary point in turn and compares it with the preset threshold value. If the direction error of a certain point is 8 degrees, which is higher than the 5-degree threshold, it is determined as an abnormal point. During the screening process, all points that meet the conditions are extracted, and they are numbered and indexed according to their row and column coordinates in the image. At the same time, a classification marking structure is established to record the position order, for example, adjacent or continuous abnormal points are classified into the same class to form a boundary direction abnormal segment. This abnormal set is registered as the "boundary direction anomaly set" by the system and stored separately as a data collection, which contains attributes such as image position, error value, and classification of the abnormal point, for subsequent spatial structure analysis, boundary reconstruction, or abnormal region positioning processing.

[0093] Please refer to Figure 5The acquiring step of the confidence adjustment node set is specifically as follows:

[0094] S411: According to the abnormal boundary point positions in the abnormal boundary direction set, the direction fitting error corresponding to each abnormal point is extracted, and the transparency state information of the boundary region at the same position in the fusion state label set is combined to construct a direction and transparency joint discrimination condition;

[0095] The position coordinates of each abnormal point in the image are read one by one, and the corresponding direction fitting error value is extracted, and the transparency state information corresponding to the position of the point in the fusion state label set is simultaneously searched, and the system combines the two elements to construct a direction and transparency joint discrimination condition, and each boundary point is analyzed as a combination. The joint discrimination condition defined by the system takes a single boundary point as the smallest unit, takes the direction error value as the first judgment element, and takes the transparency state as the second judgment element. For example, the direction error of a certain boundary point is 7 degrees, and the transparency state is “shrinkage”, so the joint state of the point is (high error, shrinkage); the error of another point is 3 degrees, and the transparency state is “expansion”, and the joint state is (low error, expansion); all points are assigned similar joint attribute labels for calling in subsequent confidence level adjustment rules. The system records the direction error value, transparency label state and point position coordinates for each point, and establishes a comprehensive data structure for judgment.

[0096] S412: Based on the direction fitting error and the transparency state of each boundary point in the direction and transparency joint discrimination condition, the confidence level adjustment rules in the transparency shrinkage and expansion states are called, the confidence level of the boundary point with a direction fitting error value greater than the error threshold and a transparency state of shrinkage is adjusted downward, the confidence level of the boundary point with a direction fitting error value less than the error threshold and a transparency state of expansion is adjusted upward, and the boundary point confidence level adjustment result is generated;

[0097] The confidence level adjustment rules in the call transparency state are invoked to process the boundary points in different combination states. The rules are set as follows: when the direction error is greater than the threshold value and the transparency state is "shrink", the confidence level is lowered; when the direction error is less than the threshold value and the transparency state is "expand", the confidence level is raised. During the processing, the system first sets the error threshold value, for example, 5 degrees, and then checks all the boundary points in turn according to the joint conditions. If the direction error of a point is 6.8 degrees and the state is "shrink", since 6.8 is greater than 5 degrees, the condition is met, and the confidence level of the point is lowered from the initial set value 80 to 70. If the error of another point is 3.2 degrees and the state is "expand", since 3.2 is less than 5 degrees, the condition is met, and the system raises the confidence level of the point from the original value 65 to 75. If the boundary point does not meet any rule, its confidence level remains unchanged. The adjustment range of the level can be controlled according to the system set range, for example, both the raising and lowering are taken as 10 as the change unit, and the level values before and after the modification are recorded, and finally the boundary point confidence level adjustment result table is formed.

[0098] S413: According to the confidence level change of the boundary points in the boundary point confidence level adjustment result, the position index and level value of the boundary points whose confidence level has been adjusted are recorded to obtain the confidence adjustment node set;

[0099] The information of all the boundary points whose confidence level has changed is recorded, including the image coordinate index, the original level value, the adjusted level value and the adjustment direction of the point. Each record represents a processed "confidence adjustment node", and is written into the confidence adjustment node set in a structured way. The system arranges the set in the original spatial order of the points to ensure that the specific point can be quickly indexed during the subsequent layer superposition, for example, the original level of a point is 80, and after being lowered, it is 70, the position index is row number 105 and column number 212, and the system stores this information in the confidence adjustment node set entry.

[0100] Please refer to Figure 6 The acquisition steps of the mountain real scene three-dimensional processing result are as follows:

[0101] S511: Based on the confidence adjustment node set, the confidence level, transparency state and direction fitting error corresponding to each boundary point are extracted, the fusion boundary area coinciding with the high difference mutation set in the mountain texture layer is located, and the boundary fusion information is generated;

[0102] First, the system reads three key attribute information corresponding to each boundary point: confidence level, transparency status, and orientation fitting error. This information has been calculated and labeled in previous processing steps and stored in the dataset in a structured form. The system extracts the information of these points one by one in spatial order and locates the corresponding layer region by combining their spatial coordinates in the image. Then, the system loads the location data of the elevation change set and performs spatial overlap matching with the fusion state region in the mountain texture layer to identify the intersection of the two types of regions—that is, the boundary segment with a clear fusion state and belonging to the elevation change region. In this intersection region, the system compares each point with the confidence adjustment point set and extracts the three attributes of all points falling into this region as the basic attributes for boundary fusion in subsequent processing. The system uniformly classifies these points and their associated attributes into the "Boundary Fusion Information" structure, forming a list of boundary points within the fusion processing range and recording the corresponding pixel position coordinates.

[0103] S512: Based on the boundary fusion information, the confidence level is used as the main fusion weight, the orientation fitting error is used as the position correction factor, and the transparency status is used to limit the boundary adjustment range. The weighted combination parameters of each pixel in the fusion area are constructed to generate the grayscale correction result of the fusion boundary.

[0104] In the constructed boundary fusion information structure, the system extracts the confidence level of each boundary pixel within the fusion region. Direction fitting error With transparency status The three parameters represent the pixel's participation attributes in the blending weight, position correction, and adjustment range, respectively. The blending weight, primarily based on confidence level, determines the intensity of the pixel's influence on the result during grayscale correction; the orientation error, as a correction factor, is used for fine-tuning the pixel's spatial orientation; and the transparency state, as a logical gate, limits whether the pixel is allowed to participate in the final layer adjustment.

[0105] The system defines a grayscale correction value for each pixel at the fusion boundary. Its original grayscale value is The fusion correction adopts the following weighted combination formula:

[0106] ;

[0107] in, : pixel Original grayscale value (unit: dimensionless, range 0–255); : pixel The grayscale value after fusion correction (unit: dimensionless). : pixel Confidence level (unit: confidence score, e.g., 0–100); Integration Area The total confidence level sum of all participating pixels (unit: confidence score); : pixel The direction fitting error (unit: angle, unit "degree"); : The maximum direction error range set for normalization (unit: degree); : Transparency state gating function, defined as follows: if the transparency state is "shrink", then ; if the transparency state is "expand", then ; : Fusion adjustment coefficient (unit: dimensionless), used to control the overall adjustment range of gray scale. The adjustment coefficient is used to control the overall adjustment range in the fusion gray scale correction process, which is the outermost proportional factor in the weighting formula. Its setting is mainly based on the following three factors: image gray scale dynamic range: since the pixel gray scale value of the mountain texture layer is usually between 0 and 255, the adjustment coefficient must be limited to a reasonable value range to avoid the value after gray scale correction exceeding the effective display range. For example, when the confidence level of the participating fusion is high and the error is low, the increment after adjustment should not make the gray scale value overflow the image expressible range, so the adjustment coefficient is usually set between 10 and 30 to ensure that the adjustment range is locally controllable. Boundary visual transition requirements: if the original gray scale difference of the image boundary area is small, a smaller adjustment coefficient (such as 10 or 15) can be used to ensure that the fusion correction is as smooth as possible; if there is a clear joint or high difference mutation in the boundary, a larger adjustment coefficient (such as 25 or 30) should be set to enhance the consistency of the boundary and increase the fusion response strength. Fusion scene complexity: when the confidence level distribution of the fusion area is concentrated (for example, most points are concentrated in the high confidence range), the system can set a larger adjustment coefficient to make the high confidence points dominate the correction process; if the confidence level distribution is dispersed, the adjustment coefficient needs to be reduced to avoid local over-adjustment leading to boundary structure damage.

[0108] Let the original gray scale value of a certain fusion pixel point be , the confidence level be , the total confidence sum in the area be , the direction fitting error be , the maximum direction error be set to , the transparency state be "shrink", i.e. , and the fusion adjustment coefficient be .

[0109] Substitute into the formula to calculate as follows:

[0110] ;

[0111] The final corrected gray scale value is 129.6, which is kept as an integer by the system and written into the layer as 130.

[0112] Firstly, three key parameters of each pixel, including confidence level, direction fitting error and transparency state, are extracted by boundary fusion information and introduced into the weighted gray correction calculation; the confidence level is used as the fusion main weight to control the influence proportion of the pixel on the final correction value, the direction error is introduced by the normalized form to be used as the adjustment factor of the spatial trend deviation, and the transparency state is determined by the gating method to ensure that the gray change direction is consistent with the layer structure adjustment logic; finally, the system integrates the three factors into the weighted formula, and the calculation result obtained is the adjusted gray value of each boundary pixel, which is used to correct the gray difference and structure deviation of the boundary area in the original mountain texture layer, and ensure the transition consistency and direction continuity of the boundary connection area.

[0113] S513: write the spatial position corresponding to the fusion boundary gray correction result into the corresponding fusion boundary area of the mountain texture layer to obtain the mountain real scene three-dimensional processing result, which is used to correct the layer splicing position and direction trend of the fusion boundary area in the mountain texture layer;

[0114] After completing the fusion boundary gray correction calculation, each pixel gray value corresponding to the correction result is written into the fusion boundary area of the mountain texture layer in turn, and the writing operation is indexed by the pixel spatial position to ensure that each correction value covers the original layer position corresponding to it. The writing process does not change the overall structure of the layer, but only replaces and updates the boundary area in the original graph which overlaps with the confidence adjustment node and the high difference mutation. In the execution process, the system uses the coordinate list recorded by the boundary fusion information as the positioning basis to perform single-point writing operation on each correction value, and synchronously binds the transparency and direction information during writing to avoid mismatch of internal information of the layer. After all the updates are completed, the system performs a coordinate alignment verification and direction continuity detection on the entire layer to ensure that the introduction of new gray values does not damage the original direction logic of the layer boundary. The finally output mountain texture layer is the fusion result containing the latest gray correction data, which has the processing characteristics of adjusted spatial splicing position and corrected direction trend, and the system identifies it as the mountain real scene three-dimensional processing result.

[0115] The real scene three-dimensional processing system based on big data is used to execute the real scene three-dimensional processing method based on big data, and the system comprises:

[0116] The high difference mutation extraction module obtains the elevation data of the target mountain area through remote sensing big data and constructs a grid distribution, performs first-order difference on the center elevation value in the adjacent grid, and screens the high difference mutation set.

[0117] The texture fusion discrimination module acquires a mountain surface texture image through remote sensing big data, constructs a mountain texture layer, judges the transparency of the boundary region of the mountain texture layer based on a height difference mutation set, and constructs a fusion state label set;

[0118] The boundary direction analysis module analyzes the direction continuity of the boundary region of the mountain texture layer based on the fusion state label set, and marks the discontinuous region as a boundary direction anomaly set;

[0119] The confidence level adjustment module adjusts the confidence level of each boundary region according to the change of the abnormal boundary direction in the boundary direction anomaly set and the boundary region transparency state in the fusion state label set, and obtains a confidence adjustment node set.

[0120] The layer reconstruction output module performs weighted reconstruction on the mountain texture layer of the target mountain region based on the confidence adjustment node set, and obtains a mountain real scene three-dimensional processing result.

[0121] The above is only a preferred embodiment of the present application, and does not limit the form of the present application, any skilled person in the art can use the disclosed technical content to make changes or modifications as equivalent embodiments applied to other fields, but any simple modification, equivalent change and modification made according to the technical essence of the present application to the above embodiments without departing from the technical solution content of the present application still belongs to the protection scope of the technical solution of the present application.

Claims

1. A method for real-time three-dimensional processing based on big data, characterized in that, The method comprises the following steps: S1: obtaining elevation data of a target mountain area by remote sensing big data and constructing a grid distribution, performing first-order difference on the central elevation value in adjacent grids, and screening a high difference mutation set; S2: obtaining a mountain surface texture image by remote sensing big data, constructing a mountain texture layer, judging the transparency of the boundary region of the mountain texture layer based on the high difference mutation set, and constructing a fusion state label set; The acquisition step of the fusion state label set is specifically: S211: obtaining a mountain surface texture image by remote sensing big data, calling a regional image corresponding to the elevation data, performing geometric correction and brightness equalization processing, completing map splicing according to a unified coordinate system, and constructing a mountain texture layer aligned with the target region; S212: based on the spatial coincidence relationship between the mountain texture layer and the high difference mutation set, positioning the corresponding layer boundary region in the mountain texture layer, extracting the transparency gradient value and gray contrast value of each regional boundary pixel, performing linear interpolation on the two parameters, and generating a boundary transparency interpolation result; S213: according to the numerical relationship between the boundary transparency interpolation result and the average gray value of the layer boundary region, calling a gray contrast reference value as a judgment standard, marking the region greater than the judgment standard as a transparency contraction state, and marking the remaining region as a transparency expansion state, and generating a fusion state label set; S3: based on the fusion state label set, analyzing the direction continuity of the boundary region of the mountain texture layer, and marking the discontinuous region as a boundary direction anomaly set; S4: according to the change of the abnormal boundary direction in the boundary direction anomaly set and the transparency state of the boundary region in the fusion state label set, readjusting the confidence level of each boundary region to obtain a confidence adjustment node set; The acquisition step of the confidence adjustment node set is specifically: S411: according to the position of the abnormal boundary point in the boundary direction anomaly set, extracting the direction fitting error corresponding to each abnormal point, and combining the transparency state information of the boundary region at the same position in the fusion state label set, constructing a direction and transparency joint discrimination condition; S412: based on the direction fitting error and transparency state of each boundary point in the direction and transparency joint discrimination condition, calling the confidence level adjustment rules in the transparency contraction and expansion states, executing confidence level down for the boundary point with a direction fitting error value greater than an error threshold and a transparency state of contraction, executing confidence level up for the boundary point with a direction fitting error value less than the error threshold and a transparency state of expansion, and generating a boundary point confidence level adjustment result; S413: according to the confidence level change of the boundary point in the boundary point confidence level adjustment result, recording the position index and level value of the boundary point whose confidence level adjustment has been completed, and obtaining a confidence adjustment node set; S5: based on the confidence adjustment node set, performing weighted reconstruction on the mountain texture layer of the target mountain area to obtain a mountain real scene three-dimensional processing result.

2. The big data based real scene three-dimensional processing method according to claim 1, characterized in that, The height difference mutation set includes mutation region position identification, height difference mutation point number sequence, and difference calculation reference grid number, the fusion state label set includes transparency judgment result mark, fusion boundary attribute classification, and mutation region association relationship, the boundary direction anomaly set includes direction discontinuous section index, abnormal boundary point set number, and direction change trend feature, the confidence adjustment node set includes boundary point confidence level adjustment result, fusion participation identification type, and confidence level grading basis, and the mountain real scene three-dimensional processing result includes fusion boundary gray output result, processed layer data index, and three-dimensional fusion state identification.

3. The big data based real scene three-dimensional processing method according to claim 1, characterized in that, The acquisition step of the height difference mutation set is specifically: S111: Obtain the height data of the target mountain area through remote sensing big data and construct a grid distribution, generate a standard regular grid in the mountain range based on a unified projection coordinate system, extract the height value corresponding to the center position in each grid, and generate a grid height value sequence; S112: Determine the corresponding relationship of adjacent grid center points according to the arrangement order of the grids in the grid height value sequence, perform first-order difference processing on the height values between adjacent points in sequence, and obtain the adjacent grid height difference change rate; S113: Based on the adjacent grid height difference change rate, compare with the mountain area slope threshold value for terrain grading, identify the boundary position of the height difference mutation, and obtain the height difference mutation set.

4. The big data based real scene three-dimensional processing method according to claim 1, characterized in that, The acquisition step of the boundary direction anomaly set is specifically: S311: Based on the fusion state label set, extract the pixel point sequence of the marked boundary region in the mountain texture layer, construct the direction angle sequence between adjacent points according to the boundary point arrangement order, and form the boundary direction angle change data; S312: According to the change trend of the direction angle of each point in the boundary direction angle change data, perform boundary main direction fitting through the Hough transform algorithm, calculate the direction deviation between each boundary point and the fitting curve, and generate the boundary direction fitting error; S313: Based on the boundary direction fitting error, compare with the direction deviation threshold value, filter the boundary points whose error values exceed the direction deviation threshold value, classify and mark them according to the position index of the target boundary points, and obtain the boundary direction anomaly set.

5. The big data based real scene three-dimensional processing method according to claim 1, characterized in that, The acquisition step of the mountain real scene three-dimensional processing result is specifically: S511: Based on the confidence adjustment node set, extract the confidence level, transparency state and direction fitting error corresponding to each boundary point, locate the fusion boundary region in the mountain texture layer which coincides with the height difference mutation set, and generate boundary fusion information; S512: According to the boundary fusion information, call the confidence level as the main fusion weight, the direction fitting error as the position correction factor, and the transparency state for limiting the boundary adjustment range, construct the weighted combination parameters of each pixel point in the fusion region, and generate the fusion boundary gray correction result; S513: Write the spatial position corresponding to the fusion boundary gray correction result into the corresponding fusion boundary region in the mountain texture layer, obtain the mountain real scene three-dimensional processing result, and use it to correct the layer splicing position and direction trend of the fusion boundary region in the mountain texture layer.

6. A real-time three-dimensional processing system based on big data, characterized by, The large data-based real scene three-dimensional processing method according to any one of claims 1-5, the system comprises: a high difference mutation extraction module, which acquires elevation data of a target mountain area through remote sensing big data and constructs a grid distribution, performs first-order difference on the central elevation value in adjacent grids, and screens a high difference mutation set; a texture fusion discrimination module, which acquires a mountain surface texture image through remote sensing big data, constructs a mountain texture layer, judges the transparency of a boundary region of the mountain texture layer based on the high difference mutation set, and constructs a fusion state label set; a boundary direction analysis module, which analyzes the direction continuity of the boundary region of the mountain texture layer based on the fusion state label set, and marks a discontinuous region as a boundary direction anomaly set; a confidence level adjustment module, which adjusts the confidence level of each boundary region according to the change of the abnormal boundary direction in the boundary direction anomaly set and the transparency state of the boundary region in the fusion state label set, and obtains a confidence adjustment node set; a layer reconstruction output module, which performs weighted reconstruction on the mountain texture layer of the target mountain area based on the confidence adjustment node set, and obtains a mountain real scene three-dimensional processing result.

Images