Unmanned aerial vehicle remote sensing image rapid splicing system and method for surveying and mapping

Abstract

The invention discloses an unmanned aerial vehicle remote sensing image rapid splicing system and method for surveying and mapping, and belongs to the technical field of image processing. In order to solve the problem of inaccurate matching result or low efficiency caused by easy mismatching and remote sensing image interference, the accuracy and efficiency of remote sensing image splicing are enhanced through accurate feature extraction and efficient feature matching, during feature extraction, descriptors containing local information are generated, feature point selection is optimized, mismatching is reduced, and the accuracy and efficiency of remote sensing image splicing are improved. In the matching process, nearest neighbor search and descriptor similarity ranking are adopted, inner points are screened in combination with iteration, mismatching is effectively eliminated, matching precision and stability are improved, based on a matching result, a geometric transformation model is accurately solved, image alignment precision is ensured, a complex transformation relation is adapted, and through resampling, transformation and smooth transition, the matching accuracy is improved. The high-quality panoramic image is merged, key information is reserved, and the problem of inaccurate matching result or low efficiency caused by remote sensing image interference is integrally solved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly relates to a rapid stitching system and method for unmanned aerial vehicle remote sensing images for surveying and mapping. Background Art

[0002] As a new type of aerial remote sensing platform, an unmanned aerial vehicle has the characteristics of high efficiency, flexibility, speed, and low cost. The digital cameras and digital video cameras carried on the vehicle can obtain high-resolution images. Due to the limitations of flight altitude and camera focal length during the aerial photography process of the unmanned aerial vehicle remote sensing platform, the obtained images have the characteristics of low flight altitude and small image format, and cannot reflect the overall situation of the shooting area, often unable to meet the ground information requirements and applications, so stitching is required.

[0003] And there are still the following problems in actual operations:

[0004] In the prior art, when extracting features, it is difficult to fully express the local image information of feature points, resulting in insufficient uniqueness and distinguishability of feature points, and prone to false matching. Unmanned aerial vehicle remote sensing images are often interfered by various factors such as illumination change, angle change, and scale change. At the same time, when the matching algorithm processes remote sensing images, it is often interfered by factors such as illumination, rotation, and scaling, resulting in inaccurate or inefficient matching results. Summary of the Invention

[0005] The purpose of the present invention is to provide a rapid stitching system and method for unmanned aerial vehicle remote sensing images for surveying and mapping to solve the problems raised in the above background art.

[0006] To achieve the above purpose, the present invention provides the following technical solution: A rapid stitching system for unmanned aerial vehicle remote sensing images for surveying and mapping, including:

[0007] A data acquisition module, used for:

[0008] Controlling the unmanned aerial vehicle to fly autonomously along a predetermined route, collecting high-resolution remote sensing images through the high-precision remote sensing equipment carried on the unmanned aerial vehicle, and receiving the original image data in real time, and performing preliminary storage, preprocessing, and classification on the original image data;

[0009] A feature extraction module, used for:

[0010] Extracting corner points, edge feature points, and feature regions from the preprocessed remote sensing image data, and describing the feature points;

[0011] A feature matching module, used for:

[0012] Based on the nearest neighbor search, matching the feature points and feature regions in different remote sensing image data, removing false matching point pairs, and establishing the corresponding relationship between remote sensing image data;

[0013] An image stitching module, for:

[0014] According to the feature matching result, solve the geometric transformation model between images, splice multiple remote sensing image data according to the corresponding relationship, perform smooth transition on the splicing area and generate a panoramic image;

[0015] A user interaction module, for:

[0016] Provide a user interface, the user interface includes function buttons for data import, preprocessing, feature extraction, matching, stitching and result display, provide parameter setting options, and display the stitching progress and results in real time;

[0017] A data management module, for:

[0018] Classify, store, manage and back up the collected original data and the spliced results, and provide data query and retrieval functions.

[0019] Furthermore, the feature extraction module includes:

[0020] A feature point detection unit, for:

[0021] According to the characteristics of the remote sensing image data and the splicing requirements, set the parameters of the algorithm, the parameters include the number of layers of the scale space pyramid and the threshold of feature point detection, traverse the remote sensing image data based on the scale-invariant feature transform algorithm, detect and extract all feature points, and generate a feature point set;

[0022] Generate a descriptor for each feature point, the descriptor contains the local image information around the feature point;

[0023] A feature region extraction unit, for:

[0024] Segment the remote sensing image into multiple regions, evaluate the uniqueness of each region, and select the regions with significant texture structure features as feature regions, the texture structure includes buildings and roads;

[0025] Generate a descriptor for each feature region, the descriptor contains the texture and shape feature information of the region;

[0026] A feature point optimization unit, for:

[0027] Screen and optimize the distribution of the detected feature points, evaluate the quality of the feature points according to the stability and distinguishability of the feature points, and eliminate the feature points with low contrast and weak edge response.

[0028] Furthermore, the feature extraction module further includes:

[0029] A descriptor generation unit, for:

[0030] Cooperate with the feature point detection unit and the feature region extraction unit to generate descriptors for the detected feature points and feature regions;

[0031] Calculate the magnitude and direction of the image gradient around the feature points, assign one or more main directions to the feature points according to the gradient direction, select a window around the feature points, divide the pixel points in the window according to polar coordinates, and count the magnitude and direction information of the gradient in each divided region, and finally generate a descriptor vector.

[0032] Furthermore, the feature matching module includes:

[0033] The feature point matching unit is used for:

[0034] Read the feature point sets and corresponding descriptors of two remotely sensed images to be matched, and based on the nearest neighbor search, find the nearest feature point in the feature point set of one image as a matching candidate for each feature point in the other image;

[0035] Sort the matching candidates according to the similarity between the descriptors, select the one with the highest similarity as the matching point, and repeat until all feature points find matching points or no more suitable matching points can be found;

[0036] Generate a set of feature point pairs.

[0037] Furthermore, the feature matching module further includes:

[0038] The false matching point elimination unit is used for:

[0039] Read the set of feature point pairs obtained by preliminary matching;

[0040] Screen the matching point pairs. Randomly select a set of samples from the matching point pairs in an iterative manner, calculate the best transformation model, and count the number of inliers that satisfy the model;

[0041] Select the model with the largest number of inliers as the final model, and use all the point pairs that satisfy the model as the final matching point pairs, and the remaining point pairs are eliminated as false matching point pairs, and output the final set of matching point pairs.

[0042] Furthermore, the image stitching module includes:

[0043] The model solution unit is used for:

[0044] According to the feature point correspondence provided by the feature matching module, solve the geometric transformation model between multiple remotely sensed images;

[0045] Obtain the matched feature point pairs from the feature matching module, select a geometric transformation model according to the application scenario and data characteristics, estimate the parameters of the transformation model, and verify the model by calculating the proportion of the matching point pairs that conform to the transformation model;

[0046] The resampling transformation unit is used for:

[0047] According to the calculated geometric transformation model, convert the parameters of the geometric transformation model into the form of a transformation matrix, resample the original image, and obtain the transformed image;

[0048] Apply the transformation matrix to the resampled image to perform translation, rotation, and scaling transformations on the image.

[0049] Furthermore, the image stitching module further includes:

[0050] The stitching smoothing unit is used for:

[0051] Detect the seam position by comparing the pixel value differences of adjacent images in the stitching area, perform smoothing processing on the seam position area, and adjust the brightness of the images in the stitching area;

[0052] The image generation unit is used for:

[0053] Merge multiple remote sensing images that have undergone smooth transition processing into a panoramic image according to the corresponding relationship, crop the panoramic image, and remove unnecessary edge areas;

[0054] Perform compression processing, save the finally generated panoramic image, and notify the user interaction module to update the display result.

[0055] Furthermore, the user interaction module includes:

[0056] The interface display control unit is used for:

[0057] Generate a user interaction interface, where the user interaction interface includes a menu bar, a toolbar, a status bar, a data import area, a processing progress display area, and a result display area, and design corresponding function buttons for each processing step;

[0058] The result display interaction unit is used for:

[0059] Display the processing result to the user, and at the same time mark the stitching information on the displayed image, where the stitching information includes feature points, matching line segments, and seam positions, and provide a save button or menu item to allow the user to save the processing result to a local file.

[0060] Furthermore, the data management module includes:

[0061] The data storage and classification unit is used for:

[0062] Classify and store the collected original image data and the stitched panoramic image data. During the data storage process, classify the data according to the preset classification rules, where the classification rules include date, UAV number, and task type, and create corresponding folders and database entries;

[0063] The data backup and recovery unit is used for:

[0064] Formulate a backup strategy based on the importance and change frequency of the data, regularly back up important data, and in the case of data loss and damage, quickly find the corresponding backup data according to the backup record and restore it to the original location.

[0065] Furthermore, a method for rapid stitching of remote sensing images of a mapping UAV is applied in the above-mentioned rapid stitching system for remote sensing images of a mapping UAV, and includes the following steps:

[0066] Step 1: Control the UAV to fly autonomously along a predetermined route, use the high-precision remote sensing equipment carried to collect high-resolution remote sensing images, receive the original image data in real time, and perform preliminary storage and preprocess the original image data;

[0067] Step 2: Traverse the preprocessed images, detect and extract corner points and edge feature points, divide the images into multiple regions, evaluate and select feature regions, generate descriptors for each feature point and feature region, screen and optimize the distribution of feature points, and eliminate feature points with low contrast and weak edge response;

[0068] Step 3: Based on the nearest neighbor search, find the best matching points for the feature points in each image to form a set of feature point pairs, calculate the best transformation model through an iterative method, count the number of inliers, select the model with the most inliers as the final model, and eliminate the matching point pairs that do not satisfy the model;

[0069] Step 4: According to the matched feature point pairs, solve the geometric transformation model between multiple remote sensing images, use the transformation model to resample and transform the original images to align the images in the same coordinate system, splice the transformed images, and smooth the seam positions;

[0070] Step 5: Display the stitched panoramic image on the user interface, mark the feature points, matching line segments and seam positions, and classify, manage and back up the original data and the stitched results.

[0071] Compared with the prior art, the beneficial effects of the present invention are:

[0072] 1. The present invention generates descriptors containing surrounding local image information for each feature point, further enhancing the uniqueness and distinguishability of the feature points, improving the accuracy and efficiency of subsequent feature matching. By segmenting the remote sensing image and evaluating the uniqueness of each region, regions with significant texture structure features are selected as feature regions, feature points with low contrast and weak edge response are removed, and feature points with high quality and good stability are retained, which helps to reduce the false matching rate in the subsequent feature matching process, improve the accuracy and reliability of the stitching result, and can better adapt to the changes of remote sensing images under different illuminations, angles and scales. Through the precise operation of the above-mentioned feature extraction module, the stitching process becomes more efficient and accurate, and the stitching result is also more natural and accurate.

[0073] 2. The present invention reads the feature point sets and corresponding descriptors of two remote sensing images, uses the nearest neighbor search algorithm to find matching candidates for each feature point, and sorts them according to the similarity between the descriptors, and selects the most similar one as the matching point. It can effectively resist interference factors such as illumination changes, rotation, and scaling between images, thereby improving the accuracy and precision of matching. By iteratively screening inliers that satisfy the optimal transformation model from the set of feature point pairs obtained from the preliminary matching, and removing point pairs that do not satisfy this model as false matching point pairs, it can effectively remove false matches caused by noise, occlusion, repeated textures, etc., and further improve the reliability and stability of the matching result. By quickly locating matching points through nearest neighbor search and descriptor similarity sorting, and then removing false matching points by iteratively calculating the optimal transformation model and screening inliers, the computational amount and processing time can be significantly reduced, and the overall efficiency of the system can be improved.

[0074] 3. According to the precise feature point correspondence provided by the feature matching module, the present invention can accurately calculate the geometric transformation model between multiple remote sensing images, ensuring the alignment accuracy between images during the stitching process and avoiding stitching errors caused by position deviation. Obtain the matched feature point pairs from the feature matching module, and select a suitable geometric transformation model according to the application scenario and data characteristics, so that the system can handle different types of remote sensing images and complex transformation relationships, improving the application scope and practicability of the system. Resample and transform the original images using the transformation matrix to ensure the geometric consistency and continuity of the transformed images. Merge multiple remote sensing images that have undergone smooth transition processing into a panoramic image, and remove unnecessary edge regions through cropping, not only retaining the important information of the images, but also making the panoramic image neater and more beautiful. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 is a schematic diagram of the modules of the remote sensing image fast stitching system of the present invention;

[0076] Figure 2Schematic flow diagram of the method for rapid stitching of remote sensing images of the present invention. Detailed implementation manners

[0077] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to 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 the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0078] In order to solve the technical problems that in the prior art, it is difficult to fully express the local image information of feature points during feature extraction, resulting in insufficient uniqueness and distinguishability of feature points and prone to false matching, and that unmanned aerial vehicle (UAV) remote sensing images are often interfered by various factors such as illumination changes, angle changes, and scale changes, and at the same time, the matching algorithm is often interfered by factors such as illumination, rotation, and scaling when processing remote sensing images, resulting in inaccurate or inefficient matching results, please refer to Figure 1-2 , the present invention provides the following technical solutions:

[0079] A rapid stitching system for UAV remote sensing images for surveying and mapping, comprising:

[0080] A data acquisition module, configured to:

[0081] Control the UAV to autonomously fly along a predetermined route, collect high-resolution remote sensing images through the high-precision remote sensing equipment carried by the UAV, and receive the original image data in real time, and perform preliminary storage, preprocessing, and classification on the original image data;

[0082] A feature extraction module, configured to:

[0083] Extract corner points, edge feature points, and feature regions from the preprocessed remote sensing image data, and describe the feature points;

[0084] A feature matching module, configured to:

[0085] Match the feature points and feature regions in different remote sensing image data based on nearest neighbor search, eliminate false matching point pairs, and establish the corresponding relationship between the remote sensing image data;

[0086] An image stitching module, configured to:

[0087] Solve the geometric transformation model between the images according to the feature matching result, stitch multiple remote sensing image data according to the corresponding relationship, perform smooth transition on the stitching area, and generate a panoramic image;

[0088] A user interaction module, configured to:

[0089] Provide a user interface which includes function buttons for data import, preprocessing, feature extraction, matching, stitching, and result display, provide parameter setting options, and display the stitching progress and results in real time;

[0090] A data management module, for:

[0091] Classify, store, manage, and backup the collected raw data and the stitched results, and provide data query and retrieval functions.

[0092] In the above embodiment, through the automatic control of the data acquisition module, the drone can fly autonomously along a predetermined route and complete the acquisition of high-resolution remote sensing images, greatly improving the efficiency and accuracy of data acquisition. At the same time, the real-time reception, preliminary storage, preprocessing, and classification of data are also completed in the automatic process, reducing manual intervention and improving work efficiency.

[0093] In the above embodiment, through the feature extraction module, the corner points, edge feature points, and feature regions in the remote sensing images are accurately extracted and described, and then the feature matching module is used for efficient and accurate matching to eliminate the false matching point pairs, thereby establishing an accurate correspondence relationship between the remote sensing images, providing a solid foundation for the subsequent image stitching module, and ensuring the high precision and reliability of the stitching results.

[0094] In the above embodiment, the image stitching module calculates the geometric transformation model between the images according to the matching results, stitches multiple remote sensing images according to the corresponding relationship, and performs smooth transition processing on the stitching area, and finally generates a panoramic image, which not only retains the original details of the images, but also makes the stitching area transition naturally, improving the visual quality of the panoramic image.

[0095] In the above embodiment, the system can be widely applied to multiple fields such as surveying and mapping, geological exploration, environmental monitoring, and urban planning, providing an efficient and accurate remote sensing image stitching solution for these fields, and helping to promote the rapid development of related industries.

[0096] A feature extraction module, including:

[0097] A feature point detection unit, for:

[0098] According to the characteristics of the remote sensing image data and the stitching requirements, set the parameters of the algorithm, where the parameters include the number of layers of the scale space pyramid and the threshold of feature point detection, traverse the remote sensing image data based on the scale-invariant feature transform algorithm, detect and extract all feature points, and generate a feature point set;

[0099] Generate a descriptor for each feature point, where the descriptor contains the local image information around the feature point;

[0100] A feature region extraction unit, configured to:

[0101] Segment the remote sensing image into multiple regions, evaluate the uniqueness of each region, and select the regions with significant texture structure features as feature regions, where the texture structure includes buildings and roads;

[0102] Generate a descriptor for each feature region, where the descriptor contains texture and shape feature information of the region;

[0103] A feature point optimization unit, configured to:

[0104] Screen and optimize the distribution of the detected feature points, evaluate the quality of the feature points according to their stability and distinguishability, and eliminate the feature points with low contrast and weak edge response;

[0105] A descriptor generation unit, configured to:

[0106] Cooperate with the feature point detection unit and the feature region extraction unit to generate descriptors for the detected feature points and feature regions;

[0107] Calculate the magnitude and direction of the image gradient around the feature points, assign one or more main directions to the feature points according to the gradient direction, select a window around the feature points, divide the pixel points in the window according to polar coordinates, and count the magnitude and direction information of the gradient in each divided region, and finally generate a descriptor vector.

[0108] In the above embodiment, the feature point detection unit can more effectively detect stable and representative feature points in the remote sensing image by precisely setting the parameters of the algorithm. At the same time, it generates a descriptor containing the surrounding local image information for each feature point, further enhancing the uniqueness and distinguishability of the feature points, improving the accuracy and efficiency of subsequent feature matching. By segmenting the remote sensing image and evaluating the uniqueness of each region, selecting the regions with significant texture structure features as feature regions, and eliminating the feature points with low contrast and weak edge response, the feature points with high quality and good stability are retained, which helps to reduce the false matching rate in the subsequent feature matching process and improve the accuracy and reliability of the stitching result.

[0109] In the above embodiment, the descriptor generation unit works in cooperation with the feature point detection unit and the feature region extraction unit. It not only generates descriptors for the detected feature points, but also generates descriptors containing texture and shape feature information for the feature regions, and can better adapt to the changes of remote sensing images under different illuminations, angles and scales. Through the precise work of the above feature extraction module, the system can quickly extract high-quality feature points and feature regions and generate accurate descriptors, making the stitching process more efficient and precise. At the same time, due to the high representativeness and stability of the feature points and feature regions, the stitching result is also more natural and accurate.

[0110] A feature matching module, comprising:

[0111] A feature point matching unit, configured to:

[0112] Read the feature point sets and corresponding descriptors of two remote sensing images to be matched, and based on nearest neighbor search, find the nearest feature point in the feature point set of one image as a matching candidate for each feature point in the other image;

[0113] Sort the matching candidates according to the similarity between the descriptors, select the one with the highest similarity as the matching point, and repeat until all feature points find matching points or no more suitable matching points can be found;

[0114] Generate a set of feature point pairs;

[0115] A false matching point elimination unit, configured to:

[0116] Read the set of feature point pairs obtained by preliminary matching;

[0117] Screen the matching point pairs, randomly select a set of samples from the matching point pairs in an iterative manner, calculate the best transformation model, and count the number of inliers that satisfy the model;

[0118] Select the model with the largest number of inliers as the final model, and use all the point pairs that satisfy the model as the final matching point pairs, and the remaining point pairs are eliminated as false matching point pairs, and output the final set of matching point pairs.

[0119] In the above embodiment, by reading the feature point sets and corresponding descriptors of two remote sensing images, using the nearest neighbor search algorithm to find matching candidates for each feature point, and sorting according to the similarity between the descriptors, selecting the most similar one as the matching point, it can effectively resist interference factors such as illumination changes, rotation, and scaling between images, thereby improving the accuracy and precision of matching. By iteratively screening out the inliers that satisfy the best transformation model from the set of feature point pairs obtained by preliminary matching, and eliminating the point pairs that do not satisfy the model as false matching point pairs, it can effectively eliminate false matches caused by noise, occlusion, repeated textures, etc., and further improve the reliability and stability of the matching result.

[0120] In the above embodiments, by performing nearest neighbor search and descriptor similarity sorting to quickly locate matching points, and then iteratively calculating the optimal transformation model and screening inliers to eliminate false matching points, the computational amount and processing time can be significantly reduced, and the overall efficiency of the system can be improved. Through the processing of the above feature matching module, a set of high-quality matching point pairs can be obtained, which provides accurate reference information for subsequent image stitching. Based on these matching point pairs, the geometric transformation model between images can be accurately calculated, and then seamless stitching of the images can be achieved, enhancing the visual effect and practical value of the stitching result.

[0121] The image stitching module includes:

[0122] The model calculation unit is used for:

[0123] According to the corresponding relationship of feature points provided by the feature matching module, calculate the geometric transformation model between multiple remote sensing images;

[0124] Obtain the matched feature point pairs from the feature matching module, select a geometric transformation model according to the application scenario and data characteristics, estimate the parameters of the transformation model, and verify the model by calculating the proportion of matching point pairs that conform to the transformation model;

[0125] The resampling transformation unit is used for:

[0126] According to the calculated geometric transformation model, convert the parameters of the geometric transformation model into the form of a transformation matrix, perform resampling on the original image to obtain the transformed image;

[0127] Apply the transformation matrix to the resampled image to perform translation, rotation, and scaling transformations on the image;

[0128] The stitching smoothing unit is used for:

[0129] Detect the seam position by comparing the pixel value differences of adjacent images in the stitching area, perform smoothing processing on the seam position area, and adjust the brightness of the images in the stitching area;

[0130] The image generation unit is used for:

[0131] Merge multiple remote sensing images that have undergone smooth transition processing into a panoramic image according to the corresponding relationship, and crop the panoramic image to remove unnecessary edge areas;

[0132] Perform compression processing, save the finally generated panoramic image, and notify the user interaction module to update the display result.

[0133] In the above embodiments, according to the exact feature point correspondence provided by the feature matching module, the geometric transformation model between multiple remote sensing images can be accurately calculated, ensuring the alignment accuracy between images during the stitching process, avoiding stitching errors caused by position deviations, obtaining the matched feature point pairs from the feature matching module, and selecting an appropriate geometric transformation model according to the application scenario and data characteristics, enabling the system to handle different types of remote sensing images and complex transformation relationships, and improving the applicability and practicality of the system.

[0134] In the above embodiments, the original image is resampled and transformed using the transformation matrix, ensuring the geometric consistency and continuity of the transformed image. At the same time, the stitching smoothing unit effectively eliminates the seam lines or brightness inconsistencies generated due to stitching by detecting the seam position and performing smoothing processing, improving the visual quality of the stitched image. Multiple remote sensing images after smooth transition processing are merged into a panoramic image, and unnecessary edge regions are removed by cropping, not only retaining the important information of the images but also making the panoramic image neater and more beautiful. At the same time, the panoramic image is compressed and saved, reducing the occupancy of storage space and facilitating subsequent processing and sharing.

[0135] The user interaction module includes:

[0136] The interface display control unit is used for:

[0137] Generating a user interaction interface, where the user interaction interface includes a menu bar, a toolbar, a status bar, a data import area, a processing progress display area, and a result display area, and designing corresponding function buttons for each processing step;

[0138] The result display interaction unit is used for:

[0139] Showing the processing result to the user, and at the same time marking the stitching information on the displayed image, where the stitching information includes feature points, matching line segments, and seam positions, providing a save button or menu item, and allowing the user to save the processing result to a local file.

[0140] In the above embodiments, the interface display control unit simplifies the complex processing process into a series of user-friendly operations by generating an intuitive and easy-to-use user interface. The design of the menu bar, toolbar, and status bar enables users to conveniently access system functions and understand the current status. At the same time, function buttons corresponding to each processing step are designed, allowing users to easily control the processing flow, improving the operation convenience and user satisfaction. The result display interaction unit not only displays the processing results but also marks the stitching information. This information not only helps users better understand the stitching process and results but also increases the transparency of the processing process, giving users more confidence in the system's output results. In addition, providing a save button or menu item allows users to save the processing results to a local file, facilitating subsequent analysis and application by users.

[0141] The data management module includes:

[0142] The data storage classification unit is used for:

[0143] Classifying and storing the collected original image data and the stitched panoramic image data. During the data storage process, classify the data according to the preset classification rules, where the classification rules include date, drone number, and task type, and create corresponding folders and database entries;

[0144] The data backup and recovery unit is used for:

[0145] According to the importance and change frequency of the data, formulate a backup strategy, regularly back up important data, and in the case of data loss and damage, quickly find the corresponding backup data according to the backup record and restore it to the original location.

[0146] In the above embodiments, through the preset classification rules, the data storage classification unit can automatically allocate the image data to the corresponding folders and create corresponding database entries. This classification method makes data searching and retrieval more convenient and efficient, avoiding the difficulties and time waste caused by data chaos, improving work efficiency. The classified and stored data is easier to back up and recover. Once the data is lost or damaged, it can be quickly restored from the backup, reducing the risk of data loss. Combining with the data security control mechanism, different access permissions can be set for different classified data to ensure data security. When specific data is needed for analysis or processing, it can quickly locate the corresponding folder and database entry, saving time. The orderly stored data provides a solid foundation for subsequent data analysis and mining, helping to discover valuable information and insights from the vast amount of data, supporting decision-making and business optimization.

[0147] To better demonstrate the rapid stitching system for UAV remote sensing images in surveying and mapping, this embodiment now proposes a rapid stitching method for UAV remote sensing images in surveying and mapping, including the following steps:

[0148] Step 1: Control the UAV to fly autonomously along a predetermined route, use the high-precision remote sensing equipment carried to collect high-resolution remote sensing images, receive the original image data in real time, and perform preliminary storage, and preprocess the original image data;

[0149] Step 2: Traverse the preprocessed images, detect and extract corner points and edge feature points, divide the images into multiple regions, evaluate and select feature regions, generate descriptors for each feature point and feature region, screen and optimize the distribution of feature points, and eliminate feature points with low contrast and weak edge response;

[0150] Step 3: Based on the nearest neighbor search, find the best matching points for the feature points in each image to form a set of feature point pairs, calculate the best transformation model through an iterative method, count the number of inliers, select the model with the most inliers as the final model, and eliminate the matching point pairs that do not satisfy the model;

[0151] Step 4: According to the matching feature point pairs, calculate the geometric transformation model between multiple remote sensing images, use the transformation model to resample and transform the original images to align the images in the same coordinate system, splice the transformed images and smooth the seam positions;

[0152] Step 5: Display the stitched panoramic image on the user interface, mark the feature points, matching line segments and seam positions, and classify, manage and back up the original data and the stitched results.

[0153] As mentioned above, it is only the preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. A system for rapid stitching of UAV remote sensing images for surveying and mapping, characterized in that: include: Data acquisition module for: Control the UAV to fly autonomously along the predetermined route, collect high-resolution remote sensing images through the high-precision remote sensing equipment carried by the UAV, receive the original image data in real time, and perform preliminary storage, preprocessing and classification of the original image data; Feature extraction module for: Extract corner points, edge feature points and feature areas from preprocessed remote sensing image data, and describe the feature points; Feature matching module for: Based on the nearest neighbor search, the feature points and feature areas in different remote sensing image data are matched, the mismatched point pairs are eliminated, and the corresponding relationship between remote sensing image data is established; Image stitching module for: According to the feature matching results, the geometric transformation model between images is solved, multiple remote sensing image data are spliced ​​according to the corresponding relationship, the splicing area is smoothly transitioned and a panoramic image is generated; User interaction module, used to: Providing a user interface, the user interface includes function buttons for data import, preprocessing, feature extraction, matching, splicing and result display, providing parameter setting options, and displaying splicing progress and results in real time; Data management module for: The collected original data and spliced ​​results are classified, stored, managed and backed up, and data query and retrieval functions are provided.

2. The system for rapid stitching of remote sensing images from an unmanned aerial vehicle for surveying and mapping as claimed in claim 1, characterized in that: The feature extraction module comprises: Feature point detection unit, used for: According to the characteristics of remote sensing image data and stitching requirements, the parameters of the algorithm are set, including the number of layers of the scale space pyramid and the threshold of feature point detection. The remote sensing image data is traversed based on the scale-invariant feature transformation algorithm to detect and extract all feature points and generate a feature point set. Generate a descriptor for each feature point, wherein the descriptor contains local image information around the feature point; Feature region extraction unit, used for: Segment the remote sensing image into multiple regions, evaluate the uniqueness of each region, and select regions with significant texture structure features as feature regions, wherein the texture structures include buildings and roads; Generate a descriptor for each feature region, wherein the descriptor contains texture and shape feature information of the region; Feature point optimization unit, used for: The detected feature points are screened and optimized for distribution, the quality of the feature points is evaluated based on their stability and distinguishability, and feature points with low contrast and weak edge response are eliminated.

3. A system for rapid stitching of remote sensing images from an unmanned aerial vehicle for surveying and mapping as claimed in claim 2, characterized in that: The feature extraction module further includes: Descriptor generation unit, used to: Cooperate with the feature point detection unit and the feature region extraction unit to generate descriptors for the detected feature points and feature regions; The size and direction of the image gradient are calculated around the feature point, one or more main directions are assigned to the feature point according to the gradient direction, a window is selected around the feature point, the pixels in the window are divided according to polar coordinates, and the size and direction information of the gradient in each divided area are counted, and finally a descriptor vector is generated.

4. The system for rapid stitching of remote sensing images from an unmanned aerial vehicle for surveying and mapping as claimed in claim 1, characterized in that: The feature matching module comprises: Feature point matching unit, used for: Read the feature point sets and corresponding descriptors of the two remote sensing images to be matched, and find the nearest feature point in the feature point set of one image for each feature point in the other image as a matching candidate based on the nearest neighbor search; Sort the matching candidates according to the similarity between the descriptors, select the one with the highest similarity as the matching point, and repeat until all feature points have found matching points or no more suitable matching points can be found; Generate a set of feature point pairs.

5. The system for rapid stitching of remote sensing images from an unmanned aerial vehicle for surveying and mapping as claimed in claim 4, characterized in that: The feature matching module further includes: Mismatched point elimination unit, used for: Read the feature point pair set obtained by preliminary matching; Screen the matching point pairs, randomly select a group of samples from the matching point pairs in an iterative manner, calculate the best transformation model, and count the number of internal points that meet the transformation model; The model with the largest number of inliers is selected as the final model, and all point pairs that satisfy the final model are used as the final matching point pairs. The remaining point pairs are eliminated as mismatched point pairs, and the final set of matching point pairs is output.

6. The system for rapid stitching of remote sensing images from an unmanned aerial vehicle for surveying and mapping as claimed in claim 1, characterized in that: The image stitching module comprises: Model solving unit for: According to the corresponding relationship of feature points provided by the feature matching module, the geometric transformation model between multiple remote sensing images is solved; Obtain matched feature point pairs from the feature matching module, select a geometric transformation model based on the application scenario and data characteristics, estimate the parameters of the transformation model, and verify the model by calculating the proportion of matching point pairs that conform to the transformation model; Resampling transform unit, used to: According to the solved geometric transformation model, the parameters of the geometric transformation model are converted into a transformation matrix form, and the original image is resampled to obtain a transformed image; Apply the transformation matrix to the resampled image to perform translation, rotation, and scaling transformations on the image.

7. The system for rapid stitching of remote sensing images from an unmanned aerial vehicle for surveying and mapping as claimed in claim 6, characterized in that: The image stitching module further includes: Splicing smoothing unit for: The seam position is detected by comparing the pixel value differences of adjacent images in the stitching area, the seam position area is smoothed, and the brightness of the image in the stitching area is adjusted; An image generation unit, for: The multiple remote sensing images after smooth transition processing are merged into a panoramic image according to the corresponding relationship, and the panoramic image is cropped to remove unnecessary edge areas; Perform compression processing, save the final panoramic image, and notify the user interaction module to update the display results.

8. The system for rapid stitching of remote sensing images from an unmanned aerial vehicle for surveying and mapping as claimed in claim 1, characterized in that: The user interaction module comprises: Interface display control unit, used for: Generate a user interaction interface, which includes a menu bar, a tool bar, a status bar, a data import area, a processing progress display area, and a result display area, and design corresponding function buttons for each processing step; Results display interactive unit, used to: The processing result is displayed to the user, and the stitching information is marked on the displayed image, where the stitching information includes feature points, matching line segments, and seam positions. A save button or menu item is provided to allow the user to save the processing result to a local file.

9. The system for rapid stitching of remote sensing images from an unmanned aerial vehicle for surveying and mapping as claimed in claim 1, characterized in that: The data management module comprises: Data storage classification unit for: The collected original image data and the stitched panoramic image data are classified and stored. During the data storage process, the data is classified according to preset classification rules, which include date, drone number, and mission type, and corresponding folders and database entries are created; Data backup and recovery unit, used for: Develop a backup strategy based on the importance and frequency of data changes, regularly back up important data, and in the event of data loss or damage, quickly find the corresponding backup data based on the backup records and restore it to its original location.

10. A method for rapid stitching of remote sensing images of unmanned aerial vehicles for surveying and mapping, applied in a system for rapid stitching of remote sensing images of unmanned aerial vehicles for surveying and mapping as claimed in any one of claims 1 to 9, characterized in that: The steps include: Step 1: Control the drone to fly autonomously along the predetermined route, use the high-precision remote sensing equipment on board to collect high-resolution remote sensing images, receive the original image data in real time, perform preliminary storage, and pre-process the original image data; Step 2: Traverse the preprocessed image, detect and extract corner points and edge feature points, segment the image into multiple regions, evaluate and select feature regions, generate descriptors for each feature point and feature region, screen and optimize the distribution of feature points, and remove feature points with low contrast and weak edge response; Step 3: Based on the nearest neighbor search, find the best matching point for the feature point in each image to form a set of feature point pairs. Calculate the best transformation model in an iterative manner, count the number of inliers, select the model with the most inliers as the final model, and remove the matching point pairs that do not meet the model. Step 4: According to the matched feature point pairs, the geometric transformation model between multiple remote sensing images is solved, and the original images are resampled and transformed using the transformation model so that each image is aligned in the same coordinate system. The transformed images are spliced ​​and the seam positions are smoothed; Step 5: Display the stitched panoramic image on the user interface, mark the feature points, matching line segments and seam positions, and classify, store, manage and back up the original data and stitched results.

Images