Method and System for Evaluating the Mosaic Quality of Low-Altitude Remote Sensing Images of Unmanned Aerial Vehicles and Reducing Redundancy

Abstract

The present invention relates to a method and system for evaluating the quality of and reducing redundancy in the stitching of low-altitude remote sensing images of unmanned aerial vehicles, belonging to the technical field of image processing. First, ground images are collected by an unmanned aerial vehicle under a preset overlap degree setting to obtain a set of low-altitude remote sensing images of the unmanned aerial vehicle under the preset overlap degree. Then, the set of low-altitude remote sensing images of the unmanned aerial vehicle is stitched to obtain a stitched image. Finally, the improved BRISQUE algorithm is used to evaluate the quality of the stitched image to obtain an image quality score. By improving the BRISQUE algorithm, it is made applicable to the quality evaluation of both visible light images and multispectral images. In addition, the present invention also proposes an image redundancy reduction method based on the improved BRISQUE algorithm, which can improve the image stitching efficiency and stitching quality.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method and system for stitching quality evaluation and redundancy reduction of low-altitude remote sensing images suitable for unmanned aerial vehicles (UAVs). Background Art

[0002] Low-altitude remote sensing by UAVs is an important means for monitoring the growth and physiological conditions of farmland crops. The overlap setting is one of the indispensable parameters during the flight of UAVs. In order to explore the influence of the overlap on the quality of stitched UAV images and obtain high-quality low-altitude remote sensing image data of UAVs, it is very necessary to use an octocopter UAV flight platform equipped with multi-spectral and RGB cameras to collect crop images with different overlap settings under the same flight speed, flight altitude, and flight path. However, most of the existing image quality evaluation methods are only applicable to visible light images and not to multi-spectral images.

[0003] Based on this, there is an urgent need for a method and system that can simultaneously evaluate the quality of visible light images and multi-spectral images. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for stitching quality evaluation and redundancy reduction of low-altitude remote sensing images of UAVs. The improved BRISQUE (Blind / Referenceless Image Spatial Quality Evaluator) algorithm is used to evaluate the quality of low-altitude remote sensing images of UAVs, which is applicable to both visible light images and multi-spectral images. At the same time, an image redundancy reduction method is proposed based on the improved BRISQUE algorithm, which can improve the image stitching efficiency and stitching quality.

[0005] To achieve the above purpose, the present invention provides the following solutions:

[0006] In a first aspect, the present invention provides a method for evaluating the stitching quality of low-altitude remote sensing images of UAVs, and the evaluation method includes:

[0007] Using a UAV to collect ground images under a preset overlap setting to obtain a set of low-altitude remote sensing images of UAVs under the preset overlap; the set of low-altitude remote sensing images of UAVs includes multiple remote sensing images, and the remote sensing images are visible light images or multi-spectral images;

[0008] Stitching the set of low-altitude remote sensing images of UAVs to obtain a stitched image;

[0009] Using the improved BRISQUE algorithm to evaluate the quality of the stitched image to obtain an image quality score.

[0010] The present invention also provides a system for evaluating the stitching quality of low-altitude remote sensing images of UAVs, and the evaluation system includes:

[0011] An image acquisition module, configured to use a drone to acquire ground images under a preset overlap degree setting, so as to obtain a drone low-altitude remote sensing image set under the preset overlap degree; the drone low-altitude remote sensing image set includes multiple remote sensing images, and the remote sensing images are visible light images or multispectral images;

[0012] An image stitching module, configured to stitch the drone low-altitude remote sensing image set to obtain a stitched image;

[0013] A quality assessment module, configured to use an improved BRISQUE algorithm to assess the quality of the stitched image, so as to obtain an image quality score.

[0014] In a second aspect, the present invention is used to provide a method for reducing redundancy of drone low-altitude remote sensing images. The redundancy reduction method includes:

[0015] Using a drone to acquire ground images under a preset overlap degree setting, so as to obtain a drone low-altitude remote sensing image set under the preset overlap degree; the drone low-altitude remote sensing image set includes multiple remote sensing images, and the remote sensing images are visible light images or multispectral images;

[0016] Using an improved BRISQUE algorithm to assess the quality of each remote sensing image, so as to obtain an image quality score;

[0017] Taking the lower limit of the preset overlap degree as a fixed redundancy interval;

[0018] According to the fixed redundancy interval and the image quality score of each remote sensing image, performing redundancy reduction processing on the drone low-altitude remote sensing image set to obtain a redundancy-reduced image set;

[0019] Judging whether the redundancy-reduced image set can be stitched;

[0020] If so, adding 1 to the fixed redundancy interval to obtain a new redundancy interval, and taking the new redundancy interval as the fixed redundancy interval for the next cycle, and returning to the step of "performing redundancy reduction processing on the drone low-altitude remote sensing image set according to the fixed redundancy interval and the image quality score of each remote sensing image";

[0021] If not, stitching the redundancy-reduced image set obtained in the previous cycle to obtain a stitched image.

[0022] The present invention is also used to provide a system for reducing redundancy of drone low-altitude remote sensing images. The redundancy reduction system includes:

[0023] An image acquisition module, configured to use a drone to acquire ground images under a preset overlap degree setting, so as to obtain a low-altitude remote sensing image set of the drone under the preset overlap degree; the low-altitude remote sensing image set of the drone includes multiple remote sensing images, and the remote sensing images are visible light images or multi-spectral images;

[0024] A quality assessment module, configured to use the improved BRISQUE algorithm to perform quality assessment on each of the remote sensing images, so as to obtain an image quality score;

[0025] A redundancy reduction processing module, configured to use the lower limit of the preset overlap degree as a fixed redundancy interval; according to the fixed redundancy interval and the image quality score of each of the remote sensing images, perform redundancy reduction processing on the low-altitude remote sensing image set of the drone, so as to obtain a redundancy-reduced image set;

[0026] A judgment module, configured to judge whether the redundancy-reduced image set can be stitched;

[0027] A return module, configured to if so, increment the fixed redundancy interval by 1 to obtain a new redundancy interval, and use the new redundancy interval as the fixed redundancy interval for the next cycle, and return to the step of "performing redundancy reduction processing on the low-altitude remote sensing image set of the drone according to the fixed redundancy interval and the image quality score of each of the remote sensing images";

[0028] A stitching module, configured to if not, stitch the redundancy-reduced image set obtained in the previous cycle, so as to obtain a stitched image.

[0029] According to the specific embodiments provided by the present invention, the following technical effects are disclosed by the present invention:

[0030] The present invention is used to provide a method and system for evaluating the stitching quality and reducing redundancy of low-altitude remote sensing images of a drone. First, a drone is used to acquire ground images under a preset overlap degree setting to obtain a low-altitude remote sensing image set of the drone under the preset overlap degree. Then, the low-altitude remote sensing image set of the drone is stitched to obtain a stitched image. Finally, the improved BRISQUE algorithm is used to perform quality assessment on the stitched image to obtain an image quality score. By improving the BRISQUE algorithm, it is made applicable to the quality assessment of both visible light images and multi-spectral images. In addition, the present invention also proposes an image redundancy reduction method based on the improved BRISQUE algorithm, which can improve the image stitching efficiency and stitching quality. Description of the Drawings

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0032] Figure 1 It is the flowchart of the evaluation method provided in Embodiment 1 of the present invention;

[0033] Figure 2 It is the schematic diagram for calculating the overlap degree provided in Embodiment 1 of the present invention;

[0034] Figure 3 It is the flowchart of the improved BRISQUE algorithm provided in Embodiment 1 of the present invention;

[0035] Figure 4 It is the system block diagram of the evaluation system provided in Embodiment 2 of the present invention;

[0036] Figure 5 It is the flowchart of the redundancy reduction method provided in Embodiment 3 of the present invention;

[0037] Figure 6 It is the overall process flowchart of the redundancy reduction method provided in Embodiment 3 of the present invention;

[0038] Figure 7 It is the system block diagram of the redundancy reduction system provided in Embodiment 4 of the present invention. Specific embodiments

[0039] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a 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 those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0040] The purpose of the present invention is to provide a method and system for evaluating the stitching quality and reducing redundancy of low-altitude remote sensing images of unmanned aerial vehicles. The improved BRISQUE algorithm is used to evaluate the quality of low-altitude remote sensing images of unmanned aerial vehicles, which is applicable to both visible light images and multispectral images. At the same time, an image redundancy reduction method is proposed based on the improved BRISQUE algorithm to solve the problem that a large amount of time cost is required for image stitching after image acquisition, improve the image stitching efficiency, and improve the image stitching quality at the same time.

[0041] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0042] Embodiment 1:

[0043] This embodiment is used to provide a method for evaluating the stitching quality of low-altitude remote sensing images of unmanned aerial vehicles. As Figure 1 shown, the evaluation method includes:

[0044] S1: Use a drone to collect ground images under a preset overlap degree setting to obtain a set of low-altitude remote sensing images of the drone under the preset overlap degree; the set of low-altitude remote sensing images of the drone includes multiple remote sensing images, and the remote sensing images are visible light images or multispectral images;

[0045] In this embodiment, an octocopter drone flight platform can be used to carry a multispectral and RGB camera to collect ground images under a preset overlap degree setting. During the flight, the remote sensing images collected by the multispectral camera are multispectral images, and all the multispectral images form a set of low-altitude remote sensing multispectral images. The remote sensing images collected by the RGB camera are visible light images, and all the visible light images form a set of low-altitude remote sensing visible light images.

[0046] As Figure 2 shown, since the later remote sensing images need to be stitched to obtain a complete orthophoto image, there are certain requirements for the overlap degree setting when shooting remote sensing images. The selection principle of the preset overlap degree in this embodiment is: taking the condition that the obtained set of low-altitude remote sensing images of the drone can be completed for stitching, determine the lower limit of the preset overlap degree; the camera exposure time setting has a certain influence on the upper limit of the overlap degree. In this way, clear and effective images can be ensured. Therefore, taking the condition that the obtained set of low-altitude remote sensing images of the drone is not distorted (ghosting), determine the upper limit of the preset overlap degree.

[0047] Specifically, the set of low-altitude remote sensing images of the drone is generally stitched in software. Therefore, when determining the lower limit of the preset overlap degree, the lowest overlap degree setting is that a single remote sensing image in the later stage can be completed for stitching in the selected software. For example, if the selected software is Agisoft Photoscan software, then the lowest overlap degree setting is that a single remote sensing image in the later stage can be completed for stitching in Agisoft Photoscan software.

[0048] When determining the upper limit of the preset overlap degree, the highest overlap degree setting is that the captured images are not ghosted (distorted). The basic requirement for the images not to be distorted during the movement of the camera is: the camera exposure time setting should be less than or equal to one-third of the time for the actual image to collect the length of one pixel. Meeting this basic requirement, the distortion caused by image ghosting can be ignored in the image quality assessment. Therefore, the upper limit value of the preset overlap degree setting is related to the exposure time (t(ms)), the camera sensor size (m(mm)*n(mm)), the pixel size (p(pixel)*q(pixel)), and the drone flight speed (v(mm / s)). The specific calculation formula for the exposure time setting is:

[0049]

[0050] During this process, the camera aperture, ISO, etc. settings are kept consistent.

[0051] Based on the condition that the obtained UAV low-altitude remote sensing image set is not distorted, determining the upper limit of the preset overlap may include: determining the upper limit of the preset overlap according to the camera exposure time, camera sensor size, pixel size and UAV flight speed.

[0052] It should be noted that the overlap in this embodiment may refer to the heading overlap or the lateral overlap. The heading overlap refers to the image overlap between adjacent photos when the aircraft is photographing along the route, and the lateral overlap refers to the image overlap between adjacent routes when the aircraft is photographing along the route. When collecting remote sensing images, the overlap can be adjusted by obtaining the camera's field of view in advance, and the imaging range is also fixed after the height is fixed. At this time, according to GPS positioning, the camera trigger signal is emitted at a fixed point to determine the overlap size. The camera trigger signal refers to the camera taking a picture once it is emitted.

[0053] S2: stitching the UAV low-altitude remote sensing image set to obtain a stitched image;

[0054] Specifically, S2 may include:

[0055] (1) Perform image preprocessing on the UAV low-altitude remote sensing image set, remove damaged images caused by equipment problems in the UAV low-altitude remote sensing image set, and obtain the preprocessed image set;

[0056] This step is the data preprocessing process, which involves performing image preprocessing on all acquired multi-band spectral low-altitude remote sensing images (i.e., multispectral images) or visible light low-altitude remote sensing images (i.e., visible light images), removing damaged images caused by equipment problems, and combining the drone multi-band spectral low-altitude remote sensing images or visible light low-altitude remote sensing images with good quality into a preprocessed image set.

[0057] (2) Stitching the preprocessed image set to obtain a stitched image.

[0058] Agisoft Photoscan software was used to stitch the preprocessed image set. The specific steps were image alignment, grid establishment and image stitching, and finally a complete large-scale stitched image was obtained.

[0059] S3: Using the improved BRISQUE algorithm to evaluate the quality of the stitched image to obtain an image quality score.

[0060] Specifically, the BRISQUE algorithm is a quality assessment method for spatial statistical image features proposed based on the change in the normalized features of the distorted image. It is applicable to natural scene images. The improved BRISQUE algorithm used in this embodiment can perform image quality assessment on both the acquired visible light image and the multispectral image.Figure 3 As shown in Figure 3 , S3 may include:

[0061] (1) Convert the stitched image into a single - band grayscale image;

[0062] Specifically, use MATLAB to convert the stitched image into a single - band grayscale image.

[0063] (2) Normalize the pixels of the single - band grayscale image to obtain a normalized image;

[0064] (3) Calculate the Pearson linear correlation coefficients (PLCC) of the MSCN adjacent coefficients of the normalized image in four directions: horizontal, vertical, main diagonal, and secondary diagonal;

[0065] (4) Fit the Pearson correlation coefficients into a non - zero - mean asymmetric generalized Gaussian distribution model, extract the features of the non - zero - mean asymmetric generalized Gaussian distribution model to obtain the image spatial domain features;

[0066] (5) Input the image spatial domain features into the support vector machine SVM regression model for regression to obtain an image quality score; the lower the image quality score, the higher the quality of the stitched image.

[0067] In this embodiment, by modifying the type of the input image and then modifying the relevant parameters in the BRISQUE algorithm, the method can be applicable to the quality assessment of both visible - light images and multi - spectral images.

[0068] The evaluation method provided in this embodiment uses the improved BRISQUE algorithm to evaluate the quality of the stitched image, which can be applicable to the image quality assessment of both unmanned aerial vehicle (UAV) remote - sensing multi - spectral images and visible - light images. Compared with traditional research that is only applicable to natural - scene images, it provides a new idea for the quality assessment of multi - spectral images.

[0069] Of course, the evaluation method used in this embodiment can also be used to explore the influence of the overlap degree on the stitching quality. Specifically, use an octocopter UAV flight platform equipped with multi - spectral and RGB cameras to collect crop images with different overlap - degree settings under the same flight speed (5 m / s), flight altitude (50 m), and flight path. The overlap degree is between the lower limit and the upper limit of the preset overlap degree. Use S2 to perform image stitching using Agisoft Photoscan software according to the flight sorties (i.e., the overlap degree) to obtain the stitched images of the complete test area at each overlap degree, and then use S3 to evaluate the quality of all stitched images to determine the influence of the overlap degree on the stitching quality. At this time, it can be applicable to the quality assessment of both multi - spectral images and visible - light images.

[0070] Example 2:

[0071] This example is used to provide a system for evaluating the stitching quality of low-altitude remote sensing images of unmanned aerial vehicles (UAVs). As Figure 4 shown, the evaluation system includes:

[0072] An image acquisition module M1, which is used to collect ground images by using a UAV under a preset overlap degree setting, so as to obtain a set of low-altitude remote sensing images of the UAV under the preset overlap degree; the set of low-altitude remote sensing images of the UAV includes multiple remote sensing images, and the remote sensing images are visible light images or multispectral images;

[0073] An image stitching module M2, which is used to stitch the set of low-altitude remote sensing images of the UAV to obtain a stitched image;

[0074] A quality evaluation module M3, which is used to evaluate the quality of the stitched image by using an improved BRISQUE algorithm to obtain an image quality score.

[0075] Example 3:

[0076] Example 1 is a method for evaluating the stitching quality of low-altitude remote sensing images of UAVs based on an improved BRISQUE algorithm. This example is based on this method for evaluating the stitching quality of low-altitude remote sensing images of UAVs, and proposes a method for reducing redundancy in images obtained by UAV flight to improve the image stitching efficiency and stitching quality. As Figure 5 and Figure 6 shown, this example is used to provide a method for reducing redundancy in low-altitude remote sensing images of UAVs, and the method for reducing redundancy includes:

[0077] T1: Use a UAV to collect ground images under a preset overlap degree setting to obtain a set of low-altitude remote sensing images of the UAV under the preset overlap degree; the set of low-altitude remote sensing images of the UAV includes multiple remote sensing images, and the remote sensing images are visible light images or multispectral images;

[0078] The steps of T1 are the same as those of S1 in Example 1, and will not be elaborated here. The preset overlap degree also needs to meet the lower and upper limits of the preset overlap degree proposed in Example 1.

[0079] T2: Use an improved BRISQUE algorithm to evaluate the quality of each remote sensing image to obtain an image quality score;

[0080] The method for evaluating the quality of remote sensing images is the same as the method for evaluating the quality of stitched images in Example 1. T2 may include:

[0081] (1) Convert the remote sensing image into a single-band grayscale image;

[0082] (2) Normalize the pixels of the single-band grayscale image to obtain a normalized image;

[0083] (3) Calculate the Pearson correlation coefficients of the MSCN adjacent coefficients in the four directions of horizontal, vertical, main diagonal, and secondary diagonal of the normalized image;

[0084] (4) Fit the Pearson correlation coefficients into a non-zero mean asymmetric generalized Gaussian distribution model, extract the features of the non-zero mean asymmetric generalized Gaussian distribution model, and obtain the image spatial domain features;

[0085] (5) Input the image spatial domain features into the support vector machine SVM regression model for regression to obtain an image quality score; the lower the image quality score, the higher the quality of the remote sensing image.

[0086] T3: Use the lower limit of the preset overlap degree as the fixed redundancy interval;

[0087] In this embodiment, the selection of the fixed redundancy interval is based on the overlap degree, not lower than the minimum overlap degree requirement, and the minimum overlap degree refers to the lower limit of the preset overlap degree.

[0088] T4: According to the fixed redundancy interval and the image quality score of each remote sensing image, perform redundancy reduction processing on the UAV low-altitude remote sensing image set to obtain a redundancy-reduced image set;

[0089] In this embodiment, the improved BRISQUE algorithm is used to perform image quality scoring on all remote sensing images, and the image with the best quality is selected as the stitching image within the fixed redundancy interval according to the high and low of the scoring results to form a redundancy-reduced image set. Specifically, T4 may include:

[0090] (1) Arrange the remote sensing images in the UAV low-altitude remote sensing image set in the order of shooting time;

[0091] (2) Store the first remote sensing image in the redundancy-reduced image set and use the first remote sensing image as the initial image;

[0092] (3) Select the remote sensing image with the lowest image quality score among the consecutive N remote sensing images after the initial image as the selected image, and store the selected image in the redundancy-reduced image set; the value of N is determined according to the fixed redundancy interval;

[0093] (4) Determine whether all remote sensing images have been screened;

[0094] (5) If not, use the selected image as the initial image in the next loop, and return to the step of "selecting the remote sensing image with the lowest image quality score among the consecutive N remote sensing images after the initial image as the selected image".

[0095] (6) If so, the image set after redundancy reduction is obtained.

[0096] More specifically, taking the fixed redundancy interval as 2 as an example, the redundancy reduction steps are introduced as follows: (1) Starting from the first image of the consecutive aerial photos, select the first image, and screen out the one with the lowest image score, that is, the best-quality one, from the consecutive 2 images. Then, starting from the selected image, screen out the one with the lowest image score, that is, the best-quality one, from the consecutive 2 images. Repeat the above steps until the photos of this flight strip are screened out, and then continue the screening of the next flight strip. The images with the best quality selected in total are combined into the image set after redundancy reduction. The images in the image set after redundancy reduction are subjected to stitching processing to obtain a high-quality orthophoto.

[0097] T5: Determine whether the image set after redundancy reduction can be stitched;

[0098] T6: If so, increase the fixed redundancy interval by 1 to obtain a new redundancy interval, and use the new redundancy interval as the fixed redundancy interval for the next cycle, and return to the step of "performing redundancy reduction on the UAV low-altitude remote sensing image set according to the fixed redundancy interval and the image quality score of each remote sensing image";

[0099] T7: If not, stitch the image set after redundancy reduction obtained in the previous cycle to obtain the stitched image.

[0100] Under the same computer environment, stitching high-resolution UAV remote sensing images is a very laborious task. For example, when doing experiments, stitching 400 RGB images with 40 million pixels requires 40 hours, and 200 images only require more than ten hours. The more the number of images, the working time during stitching increases exponentially. After redundancy reduction, it means fewer image numbers, and reducing the number of images can greatly improve the stitching efficiency. In this embodiment, it is first confirmed whether the remote sensing images obtained under this overlap degree can meet the redundancy reduction conditions, so an image stitching is performed first. If it can be stitched, it is checked whether there is image redundancy in the images under this overlap degree. Starting from taking the minimum value of the fixed redundancy interval, image redundancy reduction is carried out. After the redundancy reduction is completed, it is checked again whether image stitching can be performed. If it can still be done, increase the fixed redundancy interval by 1 and start image redundancy reduction, and so on, until the images cannot complete the overall stitching, indicating that the images left in the previous cycle are the fewest image numbers that can obtain a complete orthophoto. At this time, the image stitching efficiency is the highest and the quality is the best. This embodiment can perform redundancy reduction to the greatest extent, greatly reduce the number of images, improve work efficiency, and during the redundancy reduction process, the images with the best quality will be selected based on the image quality assessment method to form the image set after redundancy reduction, which can improve the accuracy while improving work efficiency.

[0101] As an alternative implementation, the redundancy reduction method of this embodiment may further include: using the improved BRISQUE algorithm to evaluate the quality of the spliced image, and obtaining the image quality score of the spliced image. In this embodiment, redundancy reduction processing is performed by evaluating the quality of remote sensing images. The purpose of evaluating the quality of the spliced image is to compare the impact on the image splicing quality before and after redundancy reduction, which can effectively improve the image splicing quality.

[0102] Use an octocopter UAV flight platform to carry a multispectral and RGB camera to collect crop images with different overlap degree settings under the same flight speed (5 m / s), flight altitude (50 m), and flight path. The overlap degree is between the lower limit and the upper limit. Use the redundancy reduction method of this embodiment to screen each remote sensing image at each overlap degree, and then splice them in sequence to obtain an orthophoto map after screening, record the splicing time, and at the same time evaluate the quality of the spliced image. It can be seen that the redundancy reduction method of this embodiment can greatly improve the splicing efficiency and the splicing quality.

[0103] Since generally, during the flight process, the higher the overlap degree, the shorter the shooting interval, the more images of the target area are obtained, the more data volume, and the longer the subsequent image splicing time. This parameter setting is limited by the hardware parameters of the imaging device and the requirements of image splicing, and there are upper and lower thresholds. The quality of the spliced image obtained by the image splicing technology directly determines the correctness of information acquisition. This step can obtain higher-quality spliced images while improving the image splicing efficiency. Compared with the traditional UAV flight and data processing methods, it can improve the operation efficiency, reduce the operation cost, and better guide agricultural practice.

[0104] Embodiment 4:

[0105] This embodiment is used to provide a redundancy reduction system for UAV low-altitude remote sensing images, as Figure 7 shown. The redundancy reduction system includes:

[0106] An image acquisition module M4, configured to use a UAV to collect ground images under a preset overlap degree setting, and obtain a set of UAV low-altitude remote sensing images at the preset overlap degree; the set of UAV low-altitude remote sensing images includes multiple remote sensing images, and the remote sensing images are visible light images or multispectral images;

[0107] A quality evaluation module M5, configured to use the improved BRISQUE algorithm to evaluate the quality of each remote sensing image, and obtain an image quality score;

[0108] A redundancy reduction processing module M6, configured to use the lower limit of the preset overlap degree as a fixed redundancy interval; according to the fixed redundancy interval and the image quality score of each remote sensing image, perform redundancy reduction processing on the set of UAV low-altitude remote sensing images to obtain a set of images after redundancy reduction;

[0109] A judgment module M7, configured to judge whether the image set after redundancy reduction can be stitched;

[0110] A return module M8, configured to, if so, increment the fixed redundancy interval by 1 to obtain a new redundancy interval, and use the new redundancy interval as the fixed redundancy interval for the next cycle, and return to the step of "performing redundancy reduction processing on the UAV low-altitude remote sensing image set according to the fixed redundancy interval and the image quality score of each remote sensing image";

[0111] A stitching module M9, configured to, if not, stitch the image set after redundancy reduction obtained in the previous cycle to obtain a stitched image.

[0112] The various embodiments in this specification are described in a progressive manner. The key point of each embodiment is the difference from other embodiments. The same or similar parts among the various embodiments can be referred to each other. For the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description in the method part.

[0113] Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A method for evaluating the stitching quality of low-altitude remote sensing images of unmanned aerial vehicles, characterized in that, The evaluation method includes: Using a drone to collect ground images under a preset overlap degree setting to obtain a drone low-altitude remote sensing image set under the preset overlap degree; the drone low-altitude remote sensing image set includes multiple remote sensing images, and the remote sensing images are visible light images or multispectral images; Stitching the drone low-altitude remote sensing image set to obtain a stitched image; Using the improved BRISQUE algorithm to evaluate the quality of the stitched image to obtain an image quality score, specifically including: converting the stitched image into a single-band grayscale image, and using the BRISQUE algorithm to evaluate the quality of the single-band grayscale image to obtain an image quality score.

2. The evaluation method according to claim 1, wherein The selection principle of the preset overlap degree is: taking the condition that the obtained drone low-altitude remote sensing image set can be stitched as the condition, determining the lower limit of the preset overlap degree; taking the condition that the obtained drone low-altitude remote sensing image set is not distorted as the condition, determining the upper limit of the preset overlap degree.

3. The evaluation method according to claim 2, wherein The specific process of determining the upper limit of the preset overlap degree with the condition that the obtained drone low-altitude remote sensing image set is not distorted includes: Determining the upper limit of the preset overlap degree according to the camera exposure time, camera sensor size, pixel size, and drone flight speed.

4. The evaluation method according to claim 1, characterized in that The specific process of stitching the drone low-altitude remote sensing image set to obtain a stitched image includes: Performing image preprocessing on the drone low-altitude remote sensing image set to remove damaged images in the drone low-altitude remote sensing image set to obtain a preprocessed image set; Stitching the preprocessed image set to obtain a stitched image.

5. The evaluation method according to claim 1, wherein The specific process of using the improved BRISQUE algorithm to evaluate the quality of the stitched image to obtain an image quality score includes: Converting the stitched image into a single-band grayscale image; Normalizing the pixels of the single-band grayscale image to obtain a normalized image; Calculating the Pearson correlation coefficients of the MSCN adjacent coefficients of the normalized image in four directions: horizontal, vertical, main diagonal, and secondary diagonal; Fitting the Pearson correlation coefficients into a non-zero mean asymmetric generalized Gaussian distribution model to obtain image spatial domain features; Inputting the image spatial domain features into a support vector machine to obtain an image quality score; the lower the image quality score, the higher the quality of the stitched image.

6. A low-altitude remote sensing image stitching quality evaluation system for unmanned aerial vehicles, characterized in that, The evaluation system includes: An image acquisition module for using a drone to collect ground images under a preset overlap degree setting to obtain a drone low-altitude remote sensing image set under the preset overlap degree; the drone low-altitude remote sensing image set includes multiple remote sensing images, and the remote sensing images are visible light images or multispectral images; An image stitching module for stitching the drone low-altitude remote sensing image set to obtain a stitched image; A quality evaluation module for using the improved BRISQUE algorithm to evaluate the quality of the stitched image to obtain an image quality score, specifically including: converting the stitched image into a single-band grayscale image, and using the BRISQUE algorithm to evaluate the quality of the single-band grayscale image to obtain an image quality score.

7. A method for reducing redundancy in low-altitude remote sensing images of unmanned aerial vehicles, characterized in that, The redundancy reduction method includes: Collect ground images using a drone under a preset overlap degree setting to obtain a low-altitude remote sensing image set of the drone under the preset overlap degree; the low-altitude remote sensing image set of the drone includes multiple remote sensing images, and the remote sensing images are visible light images or multispectral images; Use the improved BRISQUE algorithm to evaluate the quality of each remote sensing image to obtain an image quality score; Take the lower limit of the preset overlap degree as a fixed redundancy interval; According to the fixed redundancy interval and the image quality score of each remote sensing image, perform redundancy reduction processing on the low-altitude remote sensing image set of the drone to obtain a redundancy-reduced image set; Judge whether the redundancy-reduced image set can be stitched; If so, increase the fixed redundancy interval by 1 to obtain a new redundancy interval, and use the new redundancy interval as the fixed redundancy interval for the next cycle, and return to the step of "performing redundancy reduction processing on the low-altitude remote sensing image set of the drone according to the fixed redundancy interval and the image quality score of each remote sensing image"; If not, stitch the redundancy-reduced image set obtained in the previous cycle to obtain a stitched image.

8. The redundancy reduction method according to claim 7, characterized in that, Performing redundancy reduction processing on the low-altitude remote sensing image set of the drone according to the fixed redundancy interval and the image quality score of each remote sensing image to obtain a redundancy-reduced image set specifically includes: Arrange the remote sensing images in the low-altitude remote sensing image set of the drone in the order of shooting time; Store the first remote sensing image in the redundancy-reduced image set, and use the first remote sensing image as the initial image; Select the remote sensing image with the lowest image quality score among the consecutive N remote sensing images after the initial image as the selected image, and store the selected image in the redundancy-reduced image set; the value of N is determined according to the fixed redundancy interval; Judge whether all the remote sensing images have been screened; If not, use the selected image as the initial image in the next cycle, and return to the step of "selecting the remote sensing image with the lowest image quality score among the consecutive N remote sensing images after the initial image as the selected image".

9. The redundancy reduction method according to claim 7, wherein After obtaining the stitched image, the redundancy reduction method further includes: using the improved BRISQUE algorithm to evaluate the quality of the stitched image to obtain the image quality score of the stitched image.

10. A redundancy reduction system for low-altitude remote sensing images of unmanned aerial vehicles, characterized in that, The redundancy reduction system includes: An image acquisition module for collecting ground images using a drone under a preset overlap degree setting to obtain a low-altitude remote sensing image set of the drone under the preset overlap degree; the low-altitude remote sensing image set of the drone includes multiple remote sensing images, and the remote sensing images are visible light images or multispectral images; A quality evaluation module for using the improved BRISQUE algorithm to evaluate the quality of each remote sensing image to obtain an image quality score; A redundancy reduction processing module for taking the lower limit of the preset overlap degree as a fixed redundancy interval; according to the fixed redundancy interval and the image quality score of each remote sensing image, performing redundancy reduction processing on the low-altitude remote sensing image set of the drone to obtain a redundancy-reduced image set; A judgment module for judging whether the redundancy-reduced image set can be stitched; A return module, which is used to, if so, increment the fixed redundancy interval by 1 to obtain a new redundancy interval, and use the new redundancy interval as the fixed redundancy interval for the next cycle, and return to the step of "performing redundancy reduction processing on the UAV low-altitude remote sensing image set according to the fixed redundancy interval and the image quality score of each remote sensing image"; A splicing module, which is used to, if not, splice the redundancy-reduced image set obtained in the previous cycle to obtain a spliced image.

Images