Optimization algorithm for rapidly collecting Z value for large-data-volume oblique photography aerial flight

Abstract

The invention discloses an optimization algorithm for rapidly collecting Z values for large-data-volume oblique photography aerial flight, and relates to the technical field of natural resource data processing optimization. Comprising the following steps of 1, performing live-action three-dimensional and digital surface model construction on millions of unmanned aerial vehicle flight collection, and performing block resource division according to computer resources; 2, performing quadtree indexing and storage according to computer block resources; 3, reading and storing the land type pattern spots of the collected Z according to different resources; according to the method, the extraction speed and the extraction efficiency of DLTB data of millions of land-type pattern spots are high for a large-data-volume inclined unmanned aerial vehicle flight image; the required time is shorter for oblique photography and land type pattern spots with the same data size, the extraction result is reliable, accurate and objective, and all procedures do not need manual intervention; the generated data can be seamlessly fit with live-action three-dimensional data when a two-dimensional three-dimensional one-picture platform is constructed.

Description

Technical Field

[0001] The present invention relates to the technical field of natural resource data processing optimization, and in particular to an optimization algorithm for fast Z value acquisition for large-scale oblique photography aerial photography. Background Art

[0002] When constructing a real-life 3D platform, most existing land-type patches cannot be fitted with oblique photography mesh data because they do not have geometric Z. The present invention mainly solves the problem of fitting land-type patches with real-life 3D. During flight, the number of maps of a survey area generally ranges from a few hundred to several million. SHP data is common data in surveying and mapping, and can be represented as land-type patches, contour lines, topographic maps, vector housing data, etc. This article takes existing land-type patches as an example. Fitting existing land-type survey patches (DLTB) data with real-life 3D is an urgent problem to be solved. The present invention introduces quadtree fast indexing and quadtree storage to quickly build a key-value correspondence with land-type patches, and reconstructs the geometric Z of each patch obtained. After reconstruction, the DLTB with Z can fit the oblique photography Mesh data well, and the processing solution is fast, simple and convenient. The Z value usually represents elevation or altitude.

[0003] Land type map data survey, but how to deal with these large quantities of DLTB data, which are tens of thousands of data or hundreds of thousands of data, how to quickly collect Z values ​​and construct land type survey maps with geometric Z is the key difficulty solved by the present invention. The amount of traditional land type map data is huge. According to the existing processing software, it takes more than ten days to collect the Z values ​​of 300,000 land type maps and reconstruct the geometry. If it is loaded directly through the web program, it is relatively slow and difficult to put into practical application. The patented method can quickly construct the geometric Z value, and construct the output results of the improved quadtree retrieval and storage algorithm, and accelerate the actual application technology of real-scene three-dimensional application and land type map in two-dimensional and three-dimensional stereoscopic one map. Summary of the invention

[0004] The purpose of the present invention is to solve the problems mentioned in the above background technology, and to propose an optimization algorithm for the rapid collection of Z values ​​for large-scale oblique photography aerial flights. Taking land-type spots as an example, the present invention generates real-scene three-dimensional and digital surface models DSM for drone aerial images, as well as two-dimensional shp data of stock historical land-type spots. When constructing a real-scene three-dimensional platform, the two-dimensional shp data cannot fit with the real-scene three-dimensional data. To solve the problem of not fitting, the first thing to be solved is the elevation Z value problem of the real-scene three-dimensional data and the DLTB data. The Z value shp geographic entity data with three-dimensional DLTB is quickly collected and constructed from the stock two-dimensional data, and fit with the real-scene three-dimensional. For the problem of fitting large-scale DLTB and real-scene three-dimensional, the program uses optimized quadtree retrieval and quadtree index storage to speed up retrieval and storage efficiency, and then obtains the geometric xy information corresponding to each spot, and collects the geometric Z value of each spot corresponding to the real-scene three-dimensional data. After collection, the DLTB with Z value is reconstructed, and the data storage efficiency is accelerated and seamlessly fit.

[0005] The purpose of the present invention can be achieved by the following technical solution: An optimization algorithm for fast acquisition of Z values ​​for large-scale oblique photography, comprising the following steps:

[0006] Step 1: Construct real-life 3D and digital surface models for millions of images collected by drone flights, and divide them into blocks according to computer resources;

[0007] Step 2: Perform quadtree indexing and storage according to computer block resources;

[0008] Step 3: Read and store the land type patches collected by Z according to different resources;

[0009] Step 4: Store and read the land class patches according to the above steps, and then obtain the geometric coordinate points according to whether the patch geometry is within the entire map sheet, and then traverse and obtain the geometric Z value of the land class patch, and reconstruct it for practical application.

[0010] As a preferred implementation of the present invention, the step 1 of dividing the block resources according to the computer resources includes:

[0011] S11: Obtain digital surface model DSM based on UAV flight data;

[0012] S12: Obtain the image size and divide and store it into blocks, storage resources;

[0013] S13: Store into N blocks according to computer resources.

[0014] As a preferred implementation of the present invention, the quadtree indexing and storage according to computer block resources in step 2 includes:

[0015] S21: storing the digital surface model DSM into N blocks, and storing the blocks according to the determined number of map partitions N;

[0016] S22: Perform quadtree indexing according to reasonable resources: read the xy pixel value of each image block, insert a tree node at the position point, determine whether the number of position points in the current node exceeds the maximum capacity, if so, perform node classification, add the position point to the child node, and then re-execute S22; if not, execute step S23;

[0017] S23: After the quadtree is divided, the quadtree is stored, and the storage includes geometric position positioning xy information and pixel value information;

[0018] S24: The block resources and storage depth information reasonably divided according to the computer resources are stored, waiting for subsequent reasonable storage and calculation of resources.

[0019] As a preferred implementation of the present invention, the step 3 of reading and storing the land type spots collected by Z according to different resources includes:

[0020] S31: Obtain the land type map DLTB, read the land type map DLTB, and store it in the database data gdb according to the number of partitioned map sheets N;

[0021] S32: Read the corresponding spots in each image and store and traverse them separately. If all the spots are in the partitioned map, read the geometric node coordinates of each feature class in the map, locate and traverse a single map to obtain; if the spots are stored in different maps across regions, read the geometric node coordinates of each feature class in the map, locate and traverse the map to obtain;

[0022] S33: Read the geometric node coordinates of each feature class in the map, find the corresponding map and obtain the Z value of each feature;

[0023] S34: Retrieve N differently stored digital surface model DSM sheets and obtain the Z value of each geometry;

[0024] S35: storing the geometric position Z information of each spot and reconstructing the spot;

[0025] S36: Reconstruct the partition map patch into a land type map patch;

[0026] S37: Generate three-dimensional solid land type map result data.

[0027] As a preferred implementation of the present invention, in step 4, traversal is performed again to obtain the geometric Z value of the land type patch and to reconstruct it for practical application, including:

[0028] S41: import the actual shp into the library data gdb for subsequent data collection;

[0029] S42: Read and directly retrieve and store the same resource, and perform calculations based on subsequent needs;

[0030] S43: For resources that are not in the same block, the block resources involved in the land type map are obtained, and the geometry Z is read for storage;

[0031] S44: Produce geographic entities according to different geometries and geometric rules for practical applications.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] 1. The present invention has a fast speed and high efficiency in extracting large amounts of oblique drone aerial images and millions of land type patches DLTB data; the same amount of oblique data and land type patches, which takes 48 hours to extract using existing software, can be solved in just 6 hours using this algorithm.

[0034] 2. The extraction results of the present invention are reliable, accurate and objective, and the entire process can be solved without human intervention.

[0035] 3. The data generated by the present invention can be seamlessly combined with the real-scene three-dimensional data when constructing a two-dimensional and three-dimensional one-picture platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0037] Figure 1 It is the algorithm flow chart of the present invention;

[0038] Figure 2 A flow chart for dividing computer resources into blocks according to the present invention;

[0039] Figure 3 A flowchart of quadtree indexing and storage for the present invention;

[0040] Figure 4 The flowchart of the present invention is to read and store the land type patches collected by Z according to four different resources. DETAILED DESCRIPTION

[0041] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] Example 1

[0043] See also Figure 1-Figure 4 As shown, an optimization algorithm for fast Z value acquisition of large data volume oblique photography is applied to a processor or other equipment with data processing capabilities, including:

[0044] Step 1: Construct real-life 3D and digital surface models for millions of images collected by drone flights, and divide them into blocks according to computer resources;

[0045] Step 1: Obtain the DSM digital surface model based on the UAV flight data;

[0046] Step 2: Get the image size and store it as a block storage resource;

[0047] Step 3: Store into N blocks according to computer resources.

[0048] Step 2: Perform quadtree indexing and storage according to computer block resources;

[0049] Step 1: According to the DSM storage, it is divided into N blocks and divided reasonably according to the quadtree resources;

[0050] Step 2: Perform quadtree partitioning based on reasonable resources; including:

[0051] S21: storing the digital surface model DSM into N blocks, and storing the blocks according to the determined number of map partitions N;

[0052] S22: Perform quadtree indexing according to reasonable resources: read the xy pixel value of each image block, insert a tree node at the position point, determine whether the number of position points in the current node exceeds the maximum capacity, if so, perform node classification, add the position point to the child node, and then re-execute S22; if not, execute step S23;

[0053] S23: After the quadtree is divided, the quadtree is stored, and the storage includes geometric position positioning xy information and pixel value information;

[0054] S24: storing the block resources and storage depth information reasonably divided according to the computer resources, waiting for subsequent reasonable storage and calculation of resources;

[0055] Step 3: Read and store the land type patches collected by Z according to different resources;

[0056] Get the land class spots DLTB, read the land class spots DLTB, and store them in the library data gdb according to the number of partitioned sheets N; read the corresponding spots in each image and store and traverse them separately. If all the spots are in the partitioned sheets, read the geometric node coordinates of each feature class in the sheet, locate and traverse a single sheet to obtain them; if the spots are stored in different sheets across regions, read the geometric node coordinates of each feature class in the sheet, locate and traverse the sheet to obtain them; read the geometric node coordinates of each feature class (element class) in the sheet, find the corresponding sheet and obtain the Z value of each feature; retrieve N differently stored digital surface model DSM sheets and obtain the Z value of each geometry; store the geometric position Z information of each spot and reconstruct the spot; reconstruct the partitioned spots into a land class spot; generate three-dimensional solid land class spot result data.

[0057] Step 4: Store and read the land class patches according to the above steps, and then obtain the geometric coordinate points according to whether the patch geometry is within the entire map sheet, and then traverse and obtain the geometric Z value of the land class patch, and reconstruct it for practical application.

[0058] Step 1: According to the actual shp, import the library data gdb and collect the data later;

[0059] Step 2: Read and directly retrieve and store the same resource, and perform calculations based on subsequent needs;

[0060] Step 3: For resources that are not in the same block, obtain the block resources involved in the land type map patch, read the geometry Z for storage;

[0061] Step 4: Produce geographic entities according to different geometries and geometric rules for practical applications.

[0062] When the present invention is in use, the drone collects millions of images during flight, and constructs real-life three-dimensional and digital surface models. Then the program divides the resources into blocks according to computer resources: the resources that can be stored are constructed according to the maximum computing resources, and the number of geometric blocks N to be stored is calculated. Then, the drone flight DSM data is subsequently stored in N blocks; after the index N blocks are established, the DLTB data is stored and read, and then the Z value is collected after the image spots are read and the geometric xy is obtained, and then reconstructed. The extraction speed of large-scale tilted drone flight images and millions of land-type image spots DLTB data is fast and the extraction efficiency is high; the same amount of tilted data and land-type image spots require 48 hours to extract using existing software, but can be solved in only 6 hours using this algorithm. The extraction results are reliable, accurate, and objective, and all programs can be solved without human intervention; the generated data can be seamlessly fitted with real-life three-dimensional data when constructing a two-dimensional and three-dimensional stereoscopic one-picture platform.

[0063] Example 2

[0064] Based on Example 1, an optimization algorithm for fast Z value acquisition for large-scale oblique photography flight also includes:

[0065] Step 5: Share the data stored in the land type map and Z value, and collect the shared access times and access times of each land type map at the same time; when the shared access times are greater than the set threshold, sort all the access times according to the chronological order, calculate the time difference between two adjacent access times to get the access interval time, and calculate the average of all the access interval time to get the average access time; identify the maximum access interval time and the minimum access interval time and mark them as the maximum access interval time and the minimum access interval time respectively, and divide the value of the maximum access interval time by the value of the minimum access interval time to get the access interval ratio; normalize the shared access times, the average access time and the access interval ratio and mark the values ​​of the three as GX1, TP2 and FG3 respectively; substitute them into the preset spot analysis model:

[0066] Output the patch update value DLTB corresponding to the land class patch g ; Wherein, YC is the set number of times threshold, TS4 is the number of days the land type map has been stored, e1, e2, e3 and e4 are preset weight factors, and their sizes are custom settings, such as 1.87, 1.67, 1.35, and 1.88 respectively. Compare the patch update value with the set update threshold. When the patch update value is greater than the set update threshold, generate a re-optimization instruction for the area corresponding to the land type map and send it to the corresponding drone. The drone will fly over the area corresponding to the land type map to collect images, and execute steps one to four, and then update and store the optimized structure. This is to facilitate regular data updates and maintenance to ensure the timeliness and accuracy of the data.

[0067] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. An optimization algorithm for fast Z value acquisition of large data volume oblique photography, characterized in that: The following steps are involved: Step 1: Construct real-life 3D and digital surface models for millions of images collected by drone flights, and divide them into blocks according to computer resources; Step 2: Perform quadtree indexing and storage according to computer block resources; Step 3: Read and store the land type patches collected by Z according to different resources; Step 4: Store and read the land class patches according to the above steps, and then obtain the geometric coordinate points according to whether the patch geometry is within the entire map sheet, and then traverse and obtain the geometric Z value of the land class patch, and reconstruct it for practical application.

2. The optimization algorithm for fast acquisition of Z values ​​for large-scale oblique photography according to claim 1, characterized in that: The block resource division according to computer resources described in step 1 includes: S11: Obtain digital surface model DSM based on UAV flight data; S12: Obtain the image size and divide and store it into blocks, storage resources; S13: Store into N blocks according to computer resources.

3. The optimization algorithm for fast Z value acquisition of large data volume oblique photography according to claim 1, characterized in that: The quadtree indexing and storage according to computer block resources described in step 2 include: S21: storing the digital surface model DSM into N blocks, and storing the blocks according to the determined number of map partitions N; S22: Perform quadtree indexing according to reasonable resources: read the xy pixel value of each image block, insert a tree node at the position point, determine whether the number of position points in the current node exceeds the maximum capacity, if so, perform node classification, add the position point to the child node, and then re-execute S22; if not, execute step S23; S23: After the quadtree is divided, the quadtree is stored, and the storage includes geometric position positioning xy information and pixel value information; S24: The block resources and storage depth information reasonably divided according to the computer resources are stored, waiting for subsequent reasonable storage and calculation of resources.

4. The optimization algorithm for fast Z value acquisition of large data volume oblique photography according to claim 1, characterized in that: The reading and storage of the land type patches collected by Z according to different resources described in step 3 include: S31: Obtain the land type map DLTB, read the land type map DLTB, and store it in the database data gdb according to the number of partitioned map sheets N; S32: Read the corresponding spots in each image and store and traverse them separately. If all the spots are in the partitioned map, read the geometric node coordinates of each feature class in the map, locate and traverse a single map to obtain; if the spots are stored in different maps across regions, read the geometric node coordinates of each feature class in the map, locate and traverse the map to obtain; S33: Read the geometric node coordinates of each feature class in the map, find the corresponding map and obtain the Z value of each feature; S34: Retrieve N differently stored digital surface model DSM sheets and obtain the Z value of each geometry; S35: storing the geometric position Z information of each spot and reconstructing the spot; S36: Reconstruct the partition map patch into a land type map patch; S37: Generate three-dimensional solid land type map result data.

5. The optimization algorithm for fast Z value acquisition of large data volume oblique photography according to claim 1, characterized in that: In step 4, press traverse again, obtain the geometric Z value of the land type patch, and reconstruct it for practical application, including: S41: import the actual shp into the library data gdb for subsequent data collection; S42: Read and directly retrieve and store the same resource, and perform calculations based on subsequent needs; S43: For resources that are not in the same block, the block resources involved in the land type map are obtained, and the geometry Z is read for storage; S44: Produce geographic entities according to different geometries and geometric rules for practical applications.

Images