Test and evaluation method for wide-area three-dimensional map construction technology

Abstract

The invention discloses a wide-area three-dimensional map construction technology test and evaluation method. The method comprises the following steps: setting assessment indexes of a three-dimensional map construction method according to actual application requirements; the assessment indexes comprise quantization ranges of multiple assessment factors, and the assessment factors correspond to key image processing steps in the three-dimensional map construction method; constructing a test data set, wherein adjacent images in the test data set are overlapped; obtaining a to-be-evaluated three-dimensional map construction method, and extracting a key image processing step; based on the constructed test data set, constructing a three-dimensional map by using the three-dimensional map construction method; in the map construction process, key image processing steps in the three-dimensional map construction method are monitored in real time, corresponding quantized value calculation is executed, and a quantized value corresponding to each assessment factor is obtained; and based on the quantized value corresponding to each assessment factor of the three-dimensional map construction method and a set assessment index, judging whether the three-dimensional map construction method meets assessment requirements or not.

Description

Technical Field

[0001] The invention belongs to the technical field of computer image processing and relates to a wide-area three-dimensional map construction technology testing and evaluation method. Background Art

[0002] With the rapid development of various technologies in the field of computer image processing, research on wide-area three-dimensional map construction is also increasing. Wide-area three-dimensional maps rely on a rich source of image features. Remote sensing images collected by different types of sensors have their own characteristics, which can complement each other's information. However, in order to realize the application of this technology in actual combat environments, there are certain requirements for the performance of algorithms such as wide-area feature extraction and registration, and map resolution.

[0003] Compared with plane feature extraction and two-dimensional map construction, there is no unified evaluation benchmark and test process for the test and evaluation of wide-area three-dimensional map construction methods, and it is impossible to measure and compare different construction methods, which is not conducive to the selection of wide-area three-dimensional map construction methods for combat use. Therefore, how to propose a test and evaluation method for wide-area three-dimensional map construction technology is an urgent problem that technicians in this field need to solve. Summary of the invention

[0004] The purpose of the present invention is to provide a wide-area three-dimensional map construction technology testing and evaluation method for testing and evaluating different three-dimensional map construction methods, thereby providing support for the selection of algorithms when constructing wide-area three-dimensional maps.

[0005] In order to achieve the above tasks, the present invention adopts the following technical solutions:

[0006] A wide-area three-dimensional map construction technology test and evaluation method comprises the following steps:

[0007] The evaluation index of the three-dimensional map construction method is set according to the actual application requirements; the evaluation index includes the quantitative range of multiple evaluation factors, and the evaluation factors correspond to the key image processing steps in the three-dimensional map construction method;

[0008] Construct a test data set; the images in the test data set are composed of real aerial scene images of the flight platform, the flight platform needs to keep the camera angle unchanged during the flight process of acquiring the image, the image resolution reaches above the preset resolution, and the height of the flight platform remains unchanged; there is overlap between adjacent images in the test data set;

[0009] Obtaining a 3D map construction method to be evaluated, and extracting key image processing steps in the 3D map construction method;

[0010] Based on the constructed test data set, a three-dimensional map is constructed using the three-dimensional map construction method; during the map construction process, key image processing steps in the three-dimensional map construction method are monitored in real time, corresponding quantitative value calculations are performed, and the quantitative value corresponding to each assessment factor is obtained;

[0011] Based on the quantitative value corresponding to each assessment factor of the three-dimensional map construction method and the set assessment indicators, it is determined whether the three-dimensional map construction method meets the assessment requirements.

[0012] Furthermore, the key image processing steps of the three-dimensional map construction method include: image feature point detection, image registration, two-dimensional map construction, and multi-source information fusion.

[0013] Furthermore, the assessment factors of the set assessment indicators include: feature point repetition rate, feature point registration accuracy, registration time, two-dimensional map construction time, and wide area information grid resolution.

[0014] Furthermore, the quantitative range of the assessment indicators is:

[0015] Feature point repetition rate ≥ 70%, feature point registration accuracy ≥ 80%, registration time ≤ 1 second, two-dimensional map construction time ≤ 3s, wide area information grid resolution ≤ 1m.

[0016] Furthermore, the feature point repetition rate refers to the feature point repetition rate between the two images to be matched during the image feature point detection process; the feature point registration accuracy represents whether the feature point registration in the image to be registered is accurate during the image registration process; the registration time refers to the average registration time of the images to be registered during the image registration process; the two-dimensional map construction time refers to the time for generating a two-dimensional map; the wide-area information grid resolution refers to the grid resolution of the three-dimensional wide-area map generated by the multi-source information fusion algorithm.

[0017] Furthermore, the key image processing step in the real-time monitoring three-dimensional map construction method performs corresponding quantitative value calculation to obtain the quantitative value corresponding to each assessment factor, including:

[0018] For the image feature point detection and image registration steps, the feature point repetition rate, feature point registration accuracy and registration time are calculated as quantitative values:

[0019] The feature point repetition rate is:

[0020]

[0021] Among them, Det represents the number of feature points detected jointly in the overlapping area of ​​the two images to be registered in the test data set, and K1 and K2 represent the number of feature points detected in the overlapping area of ​​the two images respectively;

[0022] Feature point registration accuracy:

[0023]

[0024] Among them, True_Match represents the number of pairs of feature points correctly matched by the two images to be registered using the image registration algorithm, and Match represents the total number of pairs of feature points matched;

[0025] The time required for the two images to be registered to complete the registration is recorded as the registration time;

[0026] The feature point repetition rate, feature point registration accuracy and registration time need to be tested multiple times in succession, and the average value is taken as the final quantitative value;

[0027] For the two-dimensional map construction step, the time from a registered image to the generation of a two-dimensional map using the image is recorded; the average value of the time recorded through multiple tests is used as the quantitative value of the two-dimensional map construction time;

[0028] For the multi-source information fusion step, the grid resolution of the three-dimensional wide-area map generated by multi-source information fusion is calculated:

[0029]

[0030] Among them, dis represents the actual distance, and pixel represents the number of pixels corresponding to the actual distance in the generated map. After multiple tests, the average value is taken as the quantitative value of the grid resolution.

[0031] Furthermore, when the quantitative values ​​corresponding to all assessment factors are within the quantitative ranges corresponding to the assessment indicators, it indicates whether the three-dimensional map construction method meets the assessment requirements, otherwise it does not meet the assessment requirements.

[0032] A terminal device includes a processor, a memory, and a computer program stored in the memory; when the processor is executed by a computer, the wide-area three-dimensional map construction technology test and evaluation method is implemented.

[0033] A computer-readable storage medium stores a computer program; when the computer program is executed by a processor, the wide-area three-dimensional map construction technology test and evaluation method is implemented.

[0034] Compared with the prior art, the present invention has the following technical features:

[0035] The present invention constructs a standard test evaluation index system, designs a complete test process, breaks through the technical bottleneck that wide-area three-dimensional map construction technology cannot be accurately evaluated, improves the performance evaluation system of three-dimensional map construction technology, and can effectively realize the quantitative evaluation of wide-area three-dimensional map construction methods, thereby providing a reference for algorithm selection when constructing wide-area three-dimensional maps. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 Schematic diagram of the process of this method.

[0037] Figure 2 The three-dimensional map constructed by the three-dimensional map construction method in the embodiment of the present invention is displayed, wherein (a) is the overall map, and (b) is a partial enlarged view of a certain area. DETAILED DESCRIPTION

[0038] Wide-area three-dimensional map construction technology refers to a method that integrates multiple image processing methods to construct the three-dimensional spatial structure of land combat scenes, improve the light and dark contrast, spatial hierarchy and surface details of the scene, and generate high-precision color three-dimensional land combat scene maps. Traditional wide-area three-dimensional map construction methods generally include multiple key image steps such as image feature point detection, image stitching, two-dimensional map construction, and multi-source information fusion, which have an important impact on the performance of the three-dimensional map construction method and need to be paid special attention to; in this field, most three-dimensional map construction methods have the above key steps. The difference is that different three-dimensional map construction methods use different image processing algorithms in the above steps, resulting in different effects of the final constructed three-dimensional maps.

[0039] What the present invention is to do is to evaluate the performance of key links such as feature point detection, image registration, image splicing and fusion in the three-dimensional map construction technology to meet the actual use requirements. The core idea of ​​the present invention is to establish a standard evaluation index system, design a set of standardized test processes, calculate the relevant technical parameters in the key steps of the three-dimensional map construction method, and characterize the overall performance of the method through the quantitative evaluation results of these key steps.

[0040] Referring to the accompanying drawings, a wide-area three-dimensional map construction technology testing and evaluation method provided by the present invention comprises the following steps:

[0041] Step 1: setting evaluation indicators for the three-dimensional map construction method according to actual application requirements; the evaluation indicators include quantitative ranges of multiple evaluation factors, and the evaluation factors correspond to key image processing steps in the three-dimensional map construction method.

[0042] Among them, the key image processing steps of the three-dimensional map construction method include: image feature point detection, image registration, two-dimensional map construction, and multi-source information fusion; in this embodiment, the assessment factors of the set assessment indicators include: feature point repetition rate, feature point registration accuracy, registration time, two-dimensional map construction time, and wide-area information grid resolution; the assessment indicators are set according to actual application requirements. For example, in one embodiment of the present invention, the quantification range of the assessment indicators is:

[0043] Feature point repetition rate ≥ 70%, feature point registration accuracy ≥ 80%, registration time ≤ 1 second, two-dimensional map construction time ≤ 3s, wide area information grid resolution ≤ 1m.

[0044] The feature point repetition rate refers to the feature point repetition rate between the two images to be matched during the image feature point detection process, which is used to judge the feature point detection effect; the feature point registration accuracy represents whether the feature point registration in the image to be registered is accurate during the image registration process; the registration time refers to the average registration time of the images to be registered during the image registration process; the two-dimensional map construction time refers to the time for generating a two-dimensional map, which is used to judge the overall performance of the two-dimensional map construction algorithm; the wide-area information grid resolution refers to the grid resolution of the three-dimensional wide-area map generated by the multi-source information fusion algorithm.

[0045] Step 2, construct a test data set; the images in the test data set are composed of real aerial scene images of the flying platform. The flying platform needs to keep the camera angle unchanged during the flight process of acquiring the image, the image resolution reaches above the preset resolution, and the height of the flying platform remains unchanged or is within a preset range; there should be a certain degree of overlap between adjacent images in the test data set.

[0046] In one embodiment of the present invention, a drone is used to maintain a certain height and shooting angle to shoot a certain area, and 1125 scene images are obtained with a resolution of 7360×4912. The image data source type is a visible light image. The shooting interval is adjusted to allow a certain degree of overlap between adjacent images, and the latitude, longitude and altitude information when the image is shot is recorded for three-dimensional mapping.

[0047] Step 3: Obtain the 3D map construction method to be evaluated and extract key image processing steps in the 3D map construction method.

[0048] In one embodiment of the present invention, the three-dimensional map construction method to be evaluated uses the Deep-SIFT algorithm to perform image feature point detection and image registration, uses a multi-band fusion algorithm to perform image stitching and two-dimensional map construction, and finally fuses the three-dimensional elevation information of GIS data on the basis of the two-dimensional map to construct a three-dimensional map of the battlefield environment.

[0049] Step 4: Based on the constructed test data set, the three-dimensional map is constructed using the three-dimensional map construction method; during the map construction process, the key image processing steps in the three-dimensional map construction method are monitored in real time, and the corresponding quantitative value calculation steps are executed to obtain the quantitative value corresponding to each assessment factor. Specifically:

[0050] Step 4.1, for the image feature point detection and image registration steps, calculate the feature point repetition rate, feature point registration accuracy and registration time as quantitative values:

[0051] The feature point repetition rate is:

[0052]

[0053] Among them, Det represents the number of feature points detected jointly in the overlapping area of ​​the two images to be registered in the test data set, and K1 and K2 represent the number of feature points detected in the overlapping area of ​​the two images respectively.

[0054] Feature point registration accuracy:

[0055]

[0056] Among them, True_Match represents the number of pairs of feature points correctly matched by the two images to be registered using the image registration algorithm, and Match represents the total number of pairs of feature points matched.

[0057] Record the time required to complete the registration of the two images to be registered.

[0058] The feature point repetition rate, feature point registration accuracy and registration time need to be tested continuously for multiple times, and the average value is taken as the final quantitative value; that is, the registration process of multiple two images to be registered is continuously monitored to obtain a series of feature point repetition rates, feature point registration accuracy and registration time, and the average value of all feature point repetition rates is taken as the final feature point repetition rate quantitative value; the same method is also used for feature point registration accuracy and registration time.

[0059] Step 4.2, for the two-dimensional map construction step, record the time it takes for the three-dimensional map construction method to generate a two-dimensional map from a registered image to the use of the image; and use the average value of the time recorded through multiple tests as the quantitative value of the two-dimensional map construction time.

[0060] Step 4.3, for the multi-source information fusion step, calculate the grid resolution of the three-dimensional wide-area map generated by multi-source information fusion:

[0061]

[0062] Among them, dis represents the actual distance, and pixel represents the number of pixels corresponding to the actual distance in the generated map, which is dimensionless data here.

[0063] Similarly, through multiple tests, the average value is taken as the quantitative value of the grid resolution.

[0064] Step 5: Based on the quantitative value corresponding to each assessment factor of the three-dimensional map construction method and the set assessment indicators, determine whether the three-dimensional map construction method meets the assessment requirements.

[0065] In one embodiment of the present invention, a three-dimensional map construction method is tested using the test data set constructed as above, and the key image processing steps in the method include: Deep-SIFT algorithm for image feature point detection and image registration, multi-band fusion algorithm for image stitching and two-dimensional map construction, and three-dimensional elevation information of GIS data is integrated on the basis of the two-dimensional map to construct a three-dimensional map of the battlefield environment.

[0066] The test environment is a high-performance graphics card, Windows 10 system, and Python 3.9 programming language. Each assessment factor is tested more than 10 times, and the final quantitative value of each assessment factor is:

[0067] The feature point repetition rate is 72.1%, the registration accuracy is 89.8%, the registration time is 0.49s, the two-dimensional map construction time is 0.58s, and in the spliced ​​wide-area map, 1 meter resolution corresponds to 10 pixels, and the resolution is 0.1 meter.

[0068] The above test results are compared with the quantitative range of the assessment indicators given in the embodiment of step 1, and they are all within the corresponding quantitative range. Therefore, all assessment factors of the three-dimensional map construction methods participating in the test meet the assessment indicators and the assessment results are qualified; the same method is used to test different three-dimensional map construction methods, and the three-dimensional map construction method for the final application is selected from the qualified methods based on the quantitative values.

[0069] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A wide-area three-dimensional map construction technology testing and evaluation method, characterized in that: The following steps are involved: The evaluation index of the three-dimensional map construction method is set according to the actual application requirements; the evaluation index includes the quantitative range of multiple evaluation factors, and the evaluation factors correspond to the key image processing steps in the three-dimensional map construction method; Construct a test data set; the images in the test data set are composed of real aerial scene images of the flight platform, the flight platform needs to keep the camera angle unchanged during the flight process of acquiring the image, the image resolution reaches above the preset resolution, and the height of the flight platform remains unchanged; there is overlap between adjacent images in the test data set; Obtaining a 3D map construction method to be evaluated, and extracting key image processing steps in the 3D map construction method; Based on the constructed test data set, a three-dimensional map is constructed using the three-dimensional map construction method; during the map construction process, key image processing steps in the three-dimensional map construction method are monitored in real time, corresponding quantitative value calculations are performed, and the quantitative value corresponding to each assessment factor is obtained; Based on the quantitative value corresponding to each assessment factor of the three-dimensional map construction method and the set assessment indicators, it is determined whether the three-dimensional map construction method meets the assessment requirements.

2. The wide-area three-dimensional map construction technology testing and evaluation method according to claim 1, characterized in that: The key image processing steps of the three-dimensional map construction method include: image feature point detection, image registration, two-dimensional map construction, and multi-source information fusion.

3. The wide-area three-dimensional map construction technology testing and evaluation method according to claim 1, characterized in that: The assessment factors of the set assessment indicators include: feature point repetition rate, feature point registration accuracy, registration time, two-dimensional map construction time, and wide-area information grid resolution.

4. The wide-area three-dimensional map construction technology testing and evaluation method according to claim 1, characterized in that: The quantitative range of the assessment indicators is: Feature point repetition rate ≥ 65%, feature point registration accuracy ≥ 80%, registration time ≤ 1.5 seconds, two-dimensional map construction time ≤ 3s, wide area information grid resolution ≤ 1m.

5. The wide-area three-dimensional map construction technology testing and evaluation method according to claim 1, characterized in that: The feature point repetition rate refers to the feature point repetition rate between the two images to be matched during the image feature point detection process. The feature point registration accuracy represents whether the feature point registration in the image to be registered is accurate during the image registration process. The registration time refers to the average registration time of the image to be registered during the image registration process. The two-dimensional map construction time refers to the time it takes to generate the two-dimensional map; the wide-area information grid resolution refers to the grid resolution of the three-dimensional wide-area map generated by the multi-source information fusion algorithm.

6. The wide-area three-dimensional map construction technology testing and evaluation method according to claim 1, characterized in that: The key image processing step in the real-time monitoring three-dimensional map construction method is to perform corresponding quantitative value calculations to obtain the quantitative values ​​corresponding to each assessment factor, including: For the image feature point detection and image registration steps, the feature point repetition rate, feature point registration accuracy and registration time are calculated as quantitative values: The feature point repetition rate is: Among them, Det represents the number of feature points detected jointly in the overlapping area of ​​the two images to be registered in the test data set, and K1 and K2 represent the number of feature points detected in the overlapping area of ​​the two images respectively; Feature point registration accuracy: Among them, True_Match represents the number of pairs of feature points correctly matched by the two images to be registered using the image registration algorithm, and Match represents the total number of pairs of feature points matched; The time required for the two images to be registered to complete the registration is recorded as the registration time; The feature point repetition rate, feature point registration accuracy and registration time need to be tested multiple times in succession, and the average value is taken as the final quantitative value; For the two-dimensional map construction step, the time from a registered image to the generation of a two-dimensional map using the image is recorded; the average value of the time recorded through multiple tests is used as the quantitative value of the two-dimensional map construction time; For the multi-source information fusion step, the grid resolution of the three-dimensional wide-area map generated by multi-source information fusion is calculated: Among them, dis represents the actual distance, and pixel represents the number of pixels corresponding to the actual distance in the generated map. After multiple tests, the average value is taken as the quantitative value of the grid resolution.

7. The wide-area three-dimensional map construction technology testing and evaluation method according to claim 1, characterized in that: When the quantitative values ​​corresponding to all assessment factors are within the quantitative range corresponding to the assessment indicators, it indicates whether the three-dimensional map construction method meets the assessment requirements, otherwise it does not meet the assessment requirements.

8. A terminal device, comprising a processor, a memory and a computer program stored in the memory; when the processor is executed by a computer, it implements the wide-area three-dimensional map construction technology testing and evaluation method according to any one of claims 1-7.

9. A computer-readable storage medium, wherein a computer program is stored in the medium; characterized in that: When the computer program is executed by a processor, the wide-area three-dimensional map construction technology testing and evaluation method according to any one of claims 1 to 7 is implemented.

Images