Ground perspective building facade image stitching method, device, equipment and medium

Abstract

The application provides a ground-view building facade image splicing method and device, equipment and medium, and relates to the technical field of image processing. Based on the target region image features extracted from the target region marked in the selected base image, the effective region image features extracted from other shooting images, and the image pairs corresponding to each shooting image, the transformation matrix between each shooting image can be calculated by sequentially transmitting the image adjacency relationship, so as to match the target region image features from other shooting images. Thus, only the region of interest of a single image needs to be known, and the regions of interest of multiple images can be matched, reducing the number of manually marked regions of interest and reducing labor consumption. Moreover, the method can splice multiple ground-view building facade images with short baselines and small overlaps, solving the weak applicability of traditional methods under this condition and improving the automation degree of image splicing.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and more specifically, to a method, apparatus, equipment, and medium for stitching architectural facade images from a ground perspective. Background Technology

[0002] Image stitching combines multiple overlapping images into a single panoramic image. Ground-view images of building facades, including clearly visible shop signs, are of significant importance in applications such as the enhancement of realistic 3D models. However, due to limitations in sensor size and image acquisition conditions, the coverage of a single image is finite, often necessitating the stitching of multiple ground-view images.

[0003] Currently, image stitching methods are mainly divided into two types. One method obtains the interior and exterior orientation elements of images through feature matching and aerial triangulation adjustment, and then fuses them using models such as collinearity equations. This method requires at least 50% overlap between images. The other method calculates the transformation relationship between images through feature matching and then performs stitching and fusion. This method is more applicable when the shooting scene is approximately planar. To improve image acquisition efficiency, when acquiring ground-view architectural images, the overlap of the captured images is often less than 50%, and the shooting distance is relatively close with varying scene depths. Therefore, traditional image stitching methods cannot produce correct results in stitching ground-view architectural facade images.

[0004] To address the above issues, existing methods primarily involve manually cropping the region of interest, then using image processing software for position adjustment and image fusion, or acquiring highly overlapping images, restoring their pose, and then stitching them together. These methods are extremely labor-intensive and time-consuming. Therefore, there is a need for an efficient ground-view image stitching method for building facades, optimizing the stitching results and improving image stitching efficiency. Summary of the Invention

[0005] The purpose of this application is to address the shortcomings of the prior art by providing a method, apparatus, device, and medium for stitching building facade images from a ground perspective, so as to improve the efficiency of image stitching.

[0006] To achieve the above objectives, the technical solutions adopted in the embodiments of this application are as follows:

[0007] In a first aspect, embodiments of this application provide a method for stitching building facade images from a ground perspective, including:

[0008] Based on the image information of each captured image, an image pair corresponding to each captured image is generated. The image pair includes at least one adjacent image corresponding to the captured image, and each adjacent image has an overlapping area with the captured image.

[0009] Based on the texture information of each captured image, a base image is determined and the target region in the base image is marked. Image features of the target region in the base image and image features of the effective region in other captured images are extracted.

[0010] Based on the target region image features of the base image, the effective region image features of other captured images, and the image pairs corresponding to each captured image, the transformation matrix between the captured image and each adjacent image in each image pair is determined respectively.

[0011] Based on the transformation matrix between the captured image and each adjacent image in each image pair, and the initial projection plane, the projection position of each captured image is determined, and the captured images are stitched together according to the projection position of each captured image to generate the target stitched image.

[0012] Optionally, generating image pairs corresponding to each captured image based on the image information of each captured image includes:

[0013] Based on the position information of each captured image, the distance between the captured image and other captured images is determined;

[0014] Based on the distance between the captured image and other captured images, a preset number of adjacent images corresponding to the captured image are determined from the other captured images;

[0015] Based on the captured image and each adjacent image, an image pair corresponding to the captured image is generated.

[0016] Optionally, based on the target region image features of the base image, the effective region image features of other captured images, and the image pairs corresponding to each captured image, the transformation matrix between the captured image and each adjacent image in each image pair is determined, including:

[0017] Based on the target region image features of the base image and the effective region image features of each adjacent image in the image pair corresponding to the base image, the transformation matrix between the base image and each adjacent image corresponding to the base image is determined respectively.

[0018] Based on the transformation matrix between the base image and the adjacent images, the target region image features of the base image are transformed into the adjacent images to obtain the target region image features in the adjacent images;

[0019] Based on the target region image features of the adjacent images and the effective region image features of each adjacent image in the image pair corresponding to the adjacent images, the transformation matrix between the adjacent images and each adjacent image corresponding to the adjacent images is determined, and so on, until the transformation matrix between the captured image and each adjacent image in each image pair is determined.

[0020] Optionally, determining the transformation matrix between the base image and each adjacent image corresponding to the base image based on the target region image features of the base image and the effective region image features of each adjacent image in the image pair corresponding to the base image includes:

[0021] Based on the target region image features of the base image and the effective region image features of each adjacent image in the image pair corresponding to the base image, feature point matching is performed between the base image and each adjacent image corresponding to the base image to obtain multiple sets of matching feature points between the base image and each adjacent image.

[0022] The transformation parameters are calculated based on multiple sets of matching feature points between the base image and the adjacent images, as well as the transformation matrix function relationship.

[0023] Based on the transformation parameters, a transformation matrix is ​​generated between the base image and the adjacent images.

[0024] Optionally, after transforming the target region image features of the base image into the adjacent images based on the transformation matrix between the base image and the adjacent images to obtain the target region image features in the adjacent images, the process includes:

[0025] Based on the image capture size and image stitching direction of the adjacent images, the target region image features in the obtained adjacent images are corrected, and based on the corrected target region image features, the transformation matrix between the base image and the adjacent images is optimized to obtain the optimized transformation matrix between the base image and the adjacent images.

[0026] Optionally, before determining the projection position of each captured image based on the transformation matrix between the captured image and each adjacent image in each image pair, and the initial projection plane, the method further includes:

[0027] Based on the spatial distribution of each captured image or the number of image matches, an initial seed image is determined from each captured image, and the plane containing the initial seed image is determined as the initial projection plane;

[0028] The step of determining the projection position of each captured image based on the transformation matrix between the captured image and each adjacent image in each image pair, and the initial projection plane, includes:

[0029] Based on the initial seed image and the image pairs corresponding to the initial seed image, determine the adjacent images corresponding to the initial seed image;

[0030] Based on the optimized transformation matrix between the initial seed image and its adjacent images, the projection position of the adjacent images corresponding to the initial seed image in the initial projection plane is determined.

[0031] The adjacent image is used as a new seed image. The adjacent image corresponding to the new seed image is determined. Based on the optimized transformation matrix between the new seed image and the adjacent image corresponding to the new seed image, the projection position of the adjacent image corresponding to the new seed image in the initial projection plane is determined. This process is repeated until the projection position of each captured image in the initial projection plane is determined.

[0032] Optionally, determining the adjacent image corresponding to the initial seed image based on the initial seed image and the image pair corresponding to the initial seed image includes:

[0033] The number of feature point matches between the initial seed image and each of its adjacent images is determined respectively.

[0034] The neighboring image with the largest number of feature point matches with the initial seed image is determined as the adjacent image corresponding to the initial seed image.

[0035] Secondly, this application also provides a device for stitching building facade images from a ground perspective, including: a generation module, an extraction module, a determination module, and a stitching module;

[0036] The generation module is used to generate image pairs corresponding to each captured image based on the image information of each captured image. The image pairs include at least one adjacent image corresponding to the captured image, and each adjacent image has an overlapping area with the captured image.

[0037] The extraction module is used to determine a base image based on the texture information of each captured image, mark the target region in the base image, and extract the image features of the target region in the base image and the image features of the effective region in other captured images.

[0038] The determining module is used to determine the transformation matrix between the captured image and each adjacent image in each image pair based on the target region image features of the base image, the effective region image features in other captured images, and the image pairs corresponding to each captured image.

[0039] The stitching module is used to determine the projection position of each captured image based on the transformation matrix between the captured image and each adjacent image in each image pair and the initial projection plane, and to stitch the captured images together based on the projection position of each captured image to generate a target stitched image.

[0040] Optionally, the generation module is specifically used to determine the distance between the captured image and other captured images based on the position information of each captured image;

[0041] Based on the distance between the captured image and other captured images, a preset number of adjacent images corresponding to the captured image are determined from the other captured images;

[0042] Based on the captured image and each adjacent image, an image pair corresponding to the captured image is generated.

[0043] Optionally, the determining module is specifically used to determine the transformation matrix between the base image and each adjacent image corresponding to the base image based on the target region image features of the base image and the effective region image features of each adjacent image in the image pair corresponding to the base image.

[0044] Based on the transformation matrix between the base image and the adjacent images, the target region image features of the base image are transformed into the adjacent images to obtain the target region image features in the adjacent images;

[0045] Based on the target region image features of the adjacent images and the effective region image features of each adjacent image in the image pair corresponding to the adjacent images, the transformation matrix between the adjacent images and each adjacent image corresponding to the adjacent images is determined, and so on, until the transformation matrix between the captured image and each adjacent image in each image pair is determined.

[0046] Optionally, the determining module is specifically used to perform feature point matching between the base image and each of the adjacent images corresponding to the base image based on the target region image features of the base image and the effective region image features of each of the adjacent images in the image pair corresponding to the base image, so as to obtain multiple sets of matching feature points between the base image and each of the adjacent images.

[0047] The transformation parameters are calculated based on multiple sets of matching feature points between the base image and the adjacent images, as well as the transformation matrix function relationship.

[0048] Based on the transformation parameters, a transformation matrix is ​​generated between the base image and the adjacent images.

[0049] Optionally, the device further includes: a correction module;

[0050] The correction module is used to correct the target region image features in the adjacent images according to the image capture size and image stitching direction of the adjacent images, and to optimize the transformation matrix between the base image and the adjacent images according to the corrected target region image features, so as to obtain the optimized transformation matrix between the base image and the adjacent images.

[0051] Optionally, the determining module is further configured to determine an initial seed image from each captured image based on the spatial distribution of each captured image or the number of image matches, and to determine the plane where the initial seed image is located as the initial projection plane;

[0052] The determining module is specifically used to determine the adjacent image corresponding to the initial seed image based on the initial seed image and the image pair corresponding to the initial seed image;

[0053] Based on the optimized transformation matrix between the initial seed image and its adjacent images, the projection position of the adjacent images corresponding to the initial seed image in the initial projection plane is determined.

[0054] The adjacent image is used as a new seed image. The adjacent image corresponding to the new seed image is determined. Based on the optimized transformation matrix between the new seed image and the adjacent image corresponding to the new seed image, the projection position of the adjacent image corresponding to the new seed image in the initial projection plane is determined. This process is repeated until the projection position of each captured image in the initial projection plane is determined.

[0055] Optionally, the determining module is specifically used to determine the number of feature point matches between the initial seed image and each adjacent image corresponding to the initial seed image;

[0056] The neighboring image with the largest number of feature point matches with the initial seed image is determined as the adjacent image corresponding to the initial seed image.

[0057] Thirdly, embodiments of this application provide an electronic device, including: a processor, a storage medium, and a bus. The storage medium stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the method provided in the first aspect.

[0058] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the method provided in the first aspect.

[0059] The beneficial effects of this application are:

[0060] This application provides a method, apparatus, device, and medium for stitching architectural facade images from a ground perspective. The method generates image pairs of each captured image. Based on the target region image features extracted from the marked target region in a selected base image, the effective region image features extracted from other captured images, and the corresponding image pairs, the transformation matrix between each captured image is calculated through the sequential transfer of image adjacency relationships. Therefore, based on the determined initial projection plane and the transformation matrices between the captured images, the projection position information of each captured image can be obtained. Thus, the captured images can be stitched together according to the projection position information to obtain the target stitched image. By generating image pairs, the adjacency relationships between each captured image can be accurately obtained. Therefore, when calculating the transformation matrix between the captured images, the adjacency relationships between the captured images can be sequentially transferred for calculation. This allows for matching regions of interest (ROIs) of multiple images only requiring knowledge of the ROI of a single image, reducing the amount of manual marking of ROIs and lowering labor costs. Furthermore, this method can stitch together multiple ground-view building facade images with short baselines and small overlaps, solving the problem of weak applicability of traditional methods under such conditions and improving the automation level of image stitching. Attached Figure Description

[0061] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0062] Figure 1 A schematic flowchart illustrating a method for stitching building facade images from a ground perspective, provided in an embodiment of this application;

[0063] Figure 2 A flowchart illustrating another method for stitching building facade images from a ground perspective, provided in an embodiment of this application;

[0064] Figure 3 A flowchart illustrating another method for stitching building facade images from a ground perspective, provided in an embodiment of this application;

[0065] Figure 4 A schematic flowchart illustrating another method for stitching building facade images from a ground perspective, provided in an embodiment of this application;

[0066] Figure 5A flowchart illustrating another method for stitching building facade images from a ground perspective, provided in an embodiment of this application;

[0067] Figure 6 A schematic flowchart illustrating another method for stitching building facade images from a ground perspective, provided in an embodiment of this application;

[0068] Figure 7 A schematic diagram of a building facade image stitching device from a ground perspective provided in an embodiment of this application;

[0069] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0070] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the accompanying drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0071] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0072] It should be noted that the term "comprising" will be used in the embodiments of this application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.

[0073] Figure 1 This is a flowchart illustrating a method for stitching building facade images from a ground perspective, provided in an embodiment of this application. The execution entity of this method can be a computing device with data processing capabilities, such as a terminal or server. Figure 1 As shown, the method may include:

[0074] S101. Based on the image information of each captured image, generate an image pair corresponding to each captured image. In the image pair, there is at least one adjacent image corresponding to the captured image, and there is an overlapping area between each adjacent image and the captured image.

[0075] Each captured image can refer to all images to be stitched together. Based on the image information of each captured image, image pairs corresponding to each captured image can be generated. For example, if there are 3 captured images to be stitched together, captured image 1 will generate an image pair. The image pair of captured image 1 may include at least one other captured image adjacent to captured image 1. Similarly, captured image 2 and captured image 3 will also generate an image pair. The at least one other captured image adjacent to each captured image is the captured image other than the captured image itself.

[0076] In each image pair of captured images, the adjacent images corresponding to the captured image are all images with overlapping areas with the captured image. That is, each adjacent image may be an image with a part that needs to be stitched together with the captured image.

[0077] S102. Based on the texture information of each captured image, determine the base image and mark the target area in the base image, extract the image features of the target area in the base image and the image features of the effective area in other captured images.

[0078] In some embodiments, a captured image with richer texture can be selected as the base image, that is, a captured image containing more image features can be selected as the base image.

[0079] Optionally, regions of interest can be marked in the base image. These regions of interest can also refer to the areas to be stitched together, i.e., the target areas. For example, a ground-view image of a building facade contains clear shop signs and other content, which is of great significance in applications such as the modification of realistic 3D models. Therefore, when stitching together shop signs from various images, the region of interest here can refer to the shop photos in the base image.

[0080] Based on the marking of the region of interest in the base image, the intersection of the base image and the region of interest can be found and image features can be extracted from the intersection. In this case, the image features of the target region (region of interest image features) can be extracted from the base image by using an image feature extraction algorithm. That is, the image features of the target region in the base image can be extracted by extracting the marked target region.

[0081] For example, image feature extraction algorithms such as SIFT (Scale-invariant feature transform) can be used to obtain the image features of the target region in the base image.

[0082] For images other than the base image, features of all valid regions in the other images can be extracted.

[0083] It is worth noting that the image features of the target region can be composed of the location and attribute information of the feature points in the target region.

[0084] In different application scenarios, the regions of interest are different. By marking and extracting the image features of the target region from the base image, it is possible to process only the image features of the target region of interest in the subsequent feature matching and calculation process, without the need for global feature calculation, thereby greatly reducing the amount of computation.

[0085] S103. Based on the target region image features of the base image, the effective region image features of other captured images, and the image pairs corresponding to each captured image, determine the transformation matrix between the captured image and each adjacent image in each image pair.

[0086] In some embodiments, based on the target region image features of the base image extracted above, the effective region image features in other captured images, and the image pairs corresponding to each captured image generated, the transformation matrix between the captured image and each adjacent image in each image pair can be calculated.

[0087] For a given image pair, a transformation matrix is ​​calculated between the captured image and each adjacent image in the image pair.

[0088] For example, if the image pair containing image 1 includes adjacent image 1 and adjacent image 2, then the transformation matrix between image 1 and adjacent image 1, and the transformation matrix between image 1 and adjacent image 2 can be calculated respectively.

[0089] For adjacent image 1, there is also a corresponding image pair. Assuming that the image pair of adjacent image 1 includes: adjacent image 3 and adjacent image 4 corresponding to adjacent image 1; then, the transformation matrix between adjacent image 1 and adjacent image 3, and the transformation matrix between adjacent image 1 and adjacent image 4 can be calculated respectively.

[0090] By analogy, through the continuous transfer of adjacent images between images, the transformation matrix between the captured image and each adjacent image in each image pair can be calculated.

[0091] S104. Based on the transformation matrix between the captured images and each adjacent image in each image pair, and the initial projection plane, determine the projection position of each captured image, and stitch the captured images together according to the projection position of each captured image to generate the target stitched image.

[0092] Optionally, based on the determined initial projection plane and the transformation matrix between each captured image calculated above, the projection position of each captured image onto the initial projection plane can be determined. Then, the seam line can be calculated based on the projection position of each captured image, specifically using methods such as dynamic programming and graph cut. The captured images are then stitched together based on the calculation results. Furthermore, the stitched target image can be uniformly blended using methods such as feathering and Laplacian blending to make the visual effect of the resulting target stitched image more realistic.

[0093] In summary, the ground-view building facade image stitching method provided in this embodiment generates image pairs of each captured image. Based on the target region image features extracted from the marked target region in the selected base image, the effective region image features extracted from other captured images, and the corresponding image pairs of each captured image, the transformation matrix between each captured image is calculated by sequentially passing the image adjacency relationship. Therefore, based on the determined initial projection plane and the transformation matrix between each captured image, the projection position information of each captured image can be obtained. Thus, the captured images can be stitched together according to the projection position information of each captured image to obtain the target stitched image. Specifically, by generating image pairs of each captured image, the adjacency relationship between each captured image can be accurately obtained. Therefore, when calculating the transformation matrix between each captured image, the transformation matrix between each captured image can be calculated sequentially based on the adjacency relationship. This allows for matching regions of interest (ROIs) of multiple images only requiring knowledge of the ROI of a single image, reducing the amount of manual marking of ROIs and lowering manpower costs. Furthermore, this method can stitch together multiple ground-view building facade images with short baselines and small overlaps, solving the problem of weak applicability of traditional methods under such conditions and improving the automation level of image stitching.

[0094] Figure 2 A schematic flowchart illustrating another method for stitching building facade images from a ground perspective, as provided in this application embodiment; Figure 2 As shown, in step S101, generating image pairs corresponding to each captured image based on the image information of each captured image may include:

[0095] S201. Determine the distance between each captured image and other captured images based on the position information of each captured image.

[0096] This embodiment uses the generation method of an image pair from a captured image as an example.

[0097] In one feasible approach, the distance between each captured image and all other captured images can be calculated based on the positional information of each captured image; that is, the distance between each captured image and all other captured images can be calculated one by one.

[0098] The location information of the captured image can refer to the camera location corresponding to the captured image.

[0099] S202. Based on the distance between the captured image and other captured images, determine a preset number of adjacent images corresponding to the captured image from the other captured images.

[0100] Optionally, the distances between captured images and other captured images can be sorted sequentially, and a preset number of neighboring images with the closest distance can be selected from the sorting results.

[0101] The preset quantity can be set according to the characteristics of the splicing scene. When the splicing scene is more complex, the preset quantity can be set to be relatively large.

[0102] Assuming the preset number is 3, then the 3 images closest to the captured image can be selected as the adjacent images of the captured image based on the distance between the captured image and other captured images.

[0103] The distance between images here can refer to absolute distance. Assuming the location information of the captured images is: P i ={x i y i , z i The location information of other captured images is P. j Therefore, the distance between the captured image and other captured images is: L ij =|P i -P j |

[0104] S203. Generate image pairs corresponding to the captured image based on the captured image and each adjacent image.

[0105] Therefore, the image pair corresponding to the captured image includes: the captured image, and each adjacent image corresponding to the captured image as determined above.

[0106] The calculation of any image pair of captured images can be performed in the manner described above.

[0107] In another possible approach, when the location information of the captured image is not obtained, other captured images can be traversed sequentially, the image overlap rate between the captured image and other captured images can be calculated, and a preset number of captured images whose overlap rate meets the threshold can be selected as the adjacent images of the captured image.

[0108] In some embodiments, since the signs on the ground-level building facades are rectangular areas and relatively long horizontally, outdoor images are easily affected by the shooting angle, and the actual signs are not horizontal in the images. Knowing the position and orientation of the image, the image can be deformed using the pose to obtain corrected images, and the above calculations can all be performed based on the corrected images.

[0109] Optionally, distortion or deformation correction can be performed on the captured image based on its interior or exterior orientation elements. The interior orientation elements may include information such as distortion parameters and focal length; the exterior orientation elements may include the image's position and orientation. The image's position may refer to the position of the camera's projection center when the image was captured, and the image's orientation may refer to the camera's orientation.

[0110] Figure 3 This is a flowchart illustrating another method for stitching building facade images from a ground perspective, provided in an embodiment of this application. Optionally, in step S103, determining the transformation matrix between the captured image and each adjacent image in each image pair based on the target area image features of the base image, the effective area image features of other captured images, and the image pairs corresponding to each captured image may include:

[0111] S301. Based on the target region image features of the base image and the effective region image features of each adjacent image in the image pair corresponding to the base image, determine the transformation matrix between the base image and each adjacent image corresponding to the base image.

[0112] Optionally, feature matching can be performed on each adjacent image in the image pair corresponding to the base image and the base image, and the transformation matrix between the base image and each adjacent image can be calculated based on the set of matched feature points.

[0113] S302. Based on the transformation matrix between the base image and the adjacent images, the target region image features of the base image are transformed into the adjacent images to obtain the target region image features in the adjacent images.

[0114] Optionally, based on the transformation matrix between the base image and the adjacent images, the target region image features in the base image can be transformed onto the adjacent images to obtain the target region image features in the adjacent images. That is, based on the target region image features of the base image and the transformation matrix between the base image and the adjacent images, the regions that match the region of interest in the base image are matched in the adjacent images, and the regions that need to be stitched are selected from all the feature regions of the adjacent images.

[0115] S303. Based on the target region image features of adjacent images and the effective region image features of each adjacent image in the image pair corresponding to the adjacent images, determine the transformation matrix between each adjacent image and each adjacent image corresponding to the adjacent images, and so on, until the transformation matrix between the captured image and each adjacent image in each image pair is determined.

[0116] Since adjacent images also have corresponding image pairs, the transformation matrix between an adjacent image and its corresponding adjacent image can be calculated based on the target region image features in the adjacent images and the set of feature points matched between the adjacent images. By recursively applying this formula, the transformation matrix between adjacent images can be calculated based on the adjacency relationships between each captured image.

[0117] For example: based on the target region image features in base image 1 and the effective region image features of the adjacent image 2 corresponding to base image 1, calculate the transformation matrix between base image 1 and the adjacent image 2 corresponding to base image 1; based on the target region image features in adjacent image 2 and the effective region image features of adjacent image 3, calculate the transformation matrix between adjacent image 2 and the adjacent image 3 corresponding to adjacent image 2, and so on, calculate the transformation matrix between any two adjacent images.

[0118] Figure 4 This is a flowchart illustrating another method for stitching building facade images from a ground perspective, provided in an embodiment of this application. Optionally, in step S301, determining the transformation matrix between the base image and each adjacent image corresponding to the base image based on the target area image features of the base image and the effective area image features of each adjacent image in the image pair corresponding to the base image may include:

[0119] S401. Based on the target region image features of the base image and the effective region image features of each adjacent image in the image pair corresponding to the base image, feature point matching is performed between the base image and each adjacent image corresponding to the base image to obtain multiple sets of matching feature points between the base image and each adjacent image.

[0120] In this embodiment, the calculation of the transformation matrix between the base image and any adjacent image is taken as an example.

[0121] First, feature point matching is performed between the base image and its adjacent images. This involves matching the target region image features in the base image with the effective region image features in the adjacent images to obtain multiple sets of matching feature points between the base image and the adjacent images.

[0122] S402. Based on the multiple sets of matching feature points between the base image and adjacent images, and the relationship between the transformation matrix function, the transformation parameters are calculated.

[0123] Assuming feature points (x) in the base image 11 ,y 11 ) and feature points (x) in adjacent images 21 ,y 21 ) are matched with each other, and the feature points (x) in the base image are matched with each other. 12 ,y 12 ) and feature points (x) in adjacent images 22 ,y 22 ) are matched with each other, and the feature points (x) in the base image are matched with each other. 13 ,y 13 0 and feature points (x) in neighboring images 23 ,y 23 If the features are mutually matched, then substitute the above sets of mutually matched feature points into the formula. In the given information, (x1, y1) takes values ​​of (x... 11 ,y 11 ), (x 12 ,y 12 ), (x 13 ,y 13 ), (x2, y2) take values ​​of (x 21 ,y 21 ), (x 22 ,y 22 ), (x 23 ,y 23 ).

[0124] Thus, by solving the system of multivariate equations, the transformation parameters in the formula can be calculated. 11 , 12 , 13 , 21 , 22 , 23 , 31 , 32 .

[0125] S403. Based on the transformation parameters, generate the transformation matrix between the base image and the adjacent images.

[0126] Then, substitute the transformation parameters obtained from the above calculations into the formula. Then the transformation matrix between the base image and the adjacent images can be obtained.

[0127] Since the matching feature points between the base image and different neighboring images are different, the resulting transformation parameters also change accordingly, thus the base image and any neighboring image have a unique transformation matrix.

[0128] Optionally, in step S302, after transforming the target region image features of the base image into the adjacent images based on the transformation matrix between the base image and the adjacent images, and obtaining the target region image features in the adjacent images, the method further includes: correcting the target region image features in the adjacent images based on the image capture size and image stitching direction of the adjacent images, and optimizing the transformation matrix between the base image and the adjacent images based on the corrected target region image features, to obtain the optimized transformation matrix between the base image and the adjacent images.

[0129] In some embodiments, based on the transformation matrix between the base image and neighboring images, the base image i is transformed... f Image features R of the target region f Transform to adjacent image i k After that, the adjacent image i can be obtained. k Target region image features R k ={Rpt i}, Rpt i =(x m ,y m Therefore, based on the characteristics of the signage area on the building facade, Rpt can be calculated. i Maximum and minimum values ​​x in the horizontal and vertical directions min ,x max ,y min ,y max .

[0130] When the image stitching direction is horizontal, R can be expanded. k The horizontal range is the adjacent image i k The image capture width, with a vertical range of R. k Vertical range y min ,y max If the image stitching direction is vertical, the expansion R k The vertical range is the adjacent image i k The image capture height, with a horizontal range of R. k Horizontal range x min ,x max This ensures that adjacent images i k Image features R of the target region k Complete extraction.

[0131] Then update the R of the adjacent images. k Features within the range, using R k Feature reconstruction and underlying image i within the range f The target region image features are matched to eliminate those that do not match the target region image features R. k Mismatches other than those in the image i are used to obtain neighboring images. k Effective features f k Then, based on the target region image features in the base image and the effective features in the adjacent images, the transformation matrix between the base image and the adjacent images is optimized to obtain the optimized transformation matrix between the base image and the adjacent images.

[0132] By executing the above method sequentially, the optimized transformation matrix between any captured image and its neighboring images can be calculated.

[0133] Figure 5 This is a flowchart illustrating another method for stitching building facade images from a ground perspective, provided in an embodiment of this application. Optionally, in step S104, before determining the projection position of each captured image based on the transformation matrix between the captured image and each adjacent image, and the initial projection plane, the method may further include:

[0134] S501. Based on the spatial distribution of each captured image or the number of image matches, determine the initial seed image from each captured image, and determine the plane where the initial seed image is located as the initial projection plane.

[0135] Optionally, the image closest to the center can be selected as the initial seed image based on the spatial distribution of the captured images, or the image with the most matching results can be selected as the initial seed image based on the number of matching results for each captured image. The image with the most matching results refers to the image that can be matched with the most images.

[0136] Alternatively, the plane containing the initial seed image can be defined as the initial projection plane.

[0137] In step S104, determining the projection position of each captured image based on the transformation matrix between the captured image and each adjacent image in each image pair, and the initial projection plane, may include:

[0138] S502. Based on the initial seed image and the image pairs corresponding to the initial seed image, determine the adjacent images corresponding to the initial seed image.

[0139] Optionally, feature point matching can be performed between the initial seed image and each adjacent image in the image pair of the initial seed image, and the adjacent image corresponding to the initial seed image can be determined from each adjacent image based on the feature point matching result.

[0140] S503. Based on the optimized transformation matrix between the initial seed image and its adjacent images, determine the projection position of the adjacent images corresponding to the initial seed image in the initial projection plane.

[0141] Optionally, a coordinate origin can be set in the initial projection plane, and then the adjacent images can be projected onto the initial projection plane according to the optimized transformation matrix between the initial seed image and the adjacent images corresponding to the initial seed image, as well as the set coordinate origin, to obtain the projection position of the adjacent images of the initial seed image in the initial projection plane.

[0142] S504. Use the adjacent image as the new seed image, determine the adjacent image corresponding to the new seed image, and determine the projection position of the adjacent image corresponding to the new seed image in the initial projection plane based on the optimized transformation matrix between the new seed image and the adjacent image corresponding to the new seed image. Repeat this process until the projection position of each captured image in the initial projection plane is determined.

[0143] Next, the adjacent images are used as new seed images. Then, based on the image pairs corresponding to the new seed images, the adjacent images corresponding to the new seed images are determined. Based on the optimized transformation matrix between the new seed images and their corresponding adjacent images, the adjacent images corresponding to the new seed images are projected onto the initial projection plane to obtain the projection position of the adjacent images corresponding to the new seed images in the initial projection plane.

[0144] By recursively calculating in sequence, the projection position of each projected image on the initial projection plane can be calculated.

[0145] Next, based on the projection positions of each captured image in the initial projection plane, the seam line can be calculated, specifically using methods such as dynamic programming and graph cut. Finally, the captured images are uniformly blended, using methods such as feathering and Laplacian blending, to obtain the target stitched image.

[0146] Figure 6 This is a flowchart illustrating another method for stitching building facade images from a ground perspective, provided in an embodiment of this application. Optionally, in step S502, determining the adjacent images corresponding to the initial seed image based on the initial seed image and the image pairs corresponding to the initial seed image may include:

[0147] S601. Determine the number of feature point matches between the initial seed image and each of the adjacent images corresponding to the initial seed image.

[0148] Optionally, feature matching can be performed on the initial seed image and each of its corresponding neighboring images to obtain the number of feature point matches between the initial seed image and each of its neighboring images.

[0149] S602. The neighboring image with the largest number of feature point matches with the initial seed image is determined as the adjacent image corresponding to the initial seed image.

[0150] Therefore, the neighboring image with the highest number of feature point matches with the initial seed image can be determined as the adjacent image corresponding to the initial seed image. The higher the number of feature point matches, the higher the matching degree between the images, i.e., the higher the correlation, and the higher the overlap rate of the regions to be stitched together.

[0151] In summary, this application provides a method for stitching building facade images from a ground perspective. This method generates image pairs of each captured image. Based on the target region image features extracted from the marked target region in the selected base image, the effective region image features extracted from other captured images, and the corresponding image pairs, the transformation matrix between each captured image is calculated through the sequential transfer of image adjacency relationships. Therefore, based on the determined initial projection plane and the transformation matrix between each captured image, the projection position information of each captured image can be obtained. Thus, the captured images can be stitched together according to the projection position information to obtain the target stitched image. By generating image pairs of each captured image, the adjacency relationship between each captured image can be accurately obtained. Therefore, when calculating the transformation matrix between each captured image, the transformation matrix between each captured image can be calculated sequentially based on the adjacency relationship. This allows for matching regions of interest (ROIs) of multiple images only requiring knowledge of the ROI of a single image, reducing the amount of manual marking of ROIs and lowering manpower costs. Furthermore, this method can stitch together multiple ground-view building facade images with short baselines and small overlaps, solving the problem of weak applicability of traditional methods under such conditions and improving the automation level of image stitching.

[0152] The following describes the apparatus, equipment, and storage medium used to perform the ground-view building facade image stitching method provided in this application. The specific implementation process and technical effects are described above and will not be repeated below.

[0153] Figure 7This is a schematic diagram of a ground-view building facade image stitching device provided in an embodiment of this application. The functions implemented by this ground-view building facade image stitching device correspond to the steps performed by the method described above. This device can be understood as the aforementioned server, or the server's processor, or it can be understood as a component independent of the aforementioned server or processor that implements the functions of this application under the control of the server, such as... Figure 7 As shown, the device may include: a generation module 710, an extraction module 720, a determination module 730, and a splicing module 740;

[0154] The generation module 710 is used to generate image pairs corresponding to each captured image based on the image information of each captured image. The image pairs include at least one adjacent image corresponding to the captured image, and each adjacent image has an overlapping area with the captured image.

[0155] The extraction module 720 is used to determine the base image based on the texture information of each captured image, mark the target region in the base image, extract the image features of the target region in the base image and the image features of the effective region in other captured images;

[0156] The determination module 730 is used to determine the transformation matrix between the captured image and each adjacent image in each image pair based on the target region image features of the base image, the effective region image features in other captured images, and the image pairs corresponding to each captured image.

[0157] The stitching module 740 is used to determine the projection position of each captured image based on the transformation matrix between the captured image and each adjacent image in each image pair, as well as the initial projection plane, and to stitch the captured images together based on the projection position of each captured image to generate a target stitched image.

[0158] Optionally, the generation module 710 is specifically used to determine the distance between the captured image and other captured images based on the position information of each captured image;

[0159] Based on the distance between the captured image and other captured images, determine a preset number of adjacent images corresponding to the captured image from the other captured images;

[0160] Based on the captured image and its adjacent images, generate image pairs corresponding to the captured image.

[0161] Optionally, the determining module 730 is specifically used to determine the transformation matrix between the base image and each adjacent image corresponding to the base image based on the target region image features of the base image and the effective region image features of each adjacent image in the image pair corresponding to the base image.

[0162] Based on the transformation matrix between the base image and the adjacent images, the target region image features of the base image are transformed into the adjacent images to obtain the target region image features in the adjacent images;

[0163] Based on the target region image features of adjacent images and the effective region image features of each adjacent image in the corresponding image pair, the transformation matrix between adjacent images and each adjacent image is determined, and so on, until the transformation matrix between the captured image and each adjacent image in each image pair is determined.

[0164] Optionally, the determining module 730 is specifically used to perform feature point matching between the base image and each of the adjacent images corresponding to the base image based on the target region image features of the base image and the effective region image features of each of the adjacent images in the image pair corresponding to the base image, so as to obtain multiple sets of matching feature points between the base image and each of the adjacent images.

[0165] The transformation parameters are calculated based on multiple sets of matching feature points between the base image and adjacent images, as well as the relationship between the transformation matrix function.

[0166] Based on the transformation parameters, a transformation matrix is ​​generated between the base image and its neighboring images.

[0167] Optionally, the device further includes: a correction module;

[0168] The correction module is used to correct the target region image features in the obtained adjacent images based on the image capture size and image stitching direction of the adjacent images, and to optimize the transformation matrix between the base image and the adjacent images based on the corrected target region image features, so as to obtain the optimized transformation matrix between the base image and the adjacent images.

[0169] Optionally, the determining module 730 is further configured to determine an initial seed image from each captured image based on the spatial distribution of each captured image or the number of image matches, and to determine the plane where the initial seed image is located as the initial projection plane;

[0170] The determination module 730 is specifically used to determine the adjacent images corresponding to the initial seed image based on the initial seed image and the image pairs corresponding to the initial seed image.

[0171] Based on the optimized transformation matrix between the initial seed image and its adjacent images, the projection position of the adjacent images corresponding to the initial seed image in the initial projection plane is determined.

[0172] The adjacent image is used as the new seed image. The adjacent image corresponding to the new seed image is determined. Based on the optimized transformation matrix between the new seed image and the adjacent image corresponding to the new seed image, the projection position of the adjacent image corresponding to the new seed image in the initial projection plane is determined. This process is repeated until the projection position of each captured image in the initial projection plane is determined.

[0173] Optionally, the determining module 730 is specifically used to determine the number of feature point matches between the initial seed image and each adjacent image corresponding to the initial seed image;

[0174] The neighboring image with the largest number of feature point matches with the initial seed image is determined as the adjacent image corresponding to the initial seed image.

[0175] The above-described device is used to execute the method provided in the foregoing embodiments, and its implementation principle and technical effect are similar, so they will not be described again here.

[0176] These modules can be one or more integrated circuits configured to implement the above methods, such as one or more Application Specific Integrated Circuits (ASICs), one or more digital signal processors (DSPs), or one or more Field Programmable Gate Arrays (FPGAs). Alternatively, when a module is implemented using processing element scheduler code, the processing element can be a general-purpose processor, such as a Central Processing Unit (CPU) or other processor capable of calling program code. Furthermore, these modules can be integrated together as a system-on-a-chip (SOC).

[0177] The modules described above can be connected or communicate with each other via wired or wireless connections. Wired connections can include metal cables, optical fibers, hybrid cables, or any combination thereof. Wireless connections can include connections via LAN, WAN, Bluetooth, ZigBee, or NFC, or any combination thereof. Two or more modules can be combined into a single module, and any module can be divided into two or more units. Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the method embodiments, and will not be repeated here.

[0178] Figure 8This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The device may be a computing device with data processing capabilities.

[0179] The device may include: a processor 801 and a storage medium 802.

[0180] Storage medium 802 is used to store programs, and processor 801 calls the programs stored in storage medium 802 to execute the above method embodiments. The specific implementation and technical effects are similar, and will not be described in detail here.

[0181] The storage medium 802 stores program code, which, when executed by the processor 801, causes the processor 801 to perform various steps in the methods according to various exemplary embodiments of this application described in the "Exemplary Methods" section above.

[0182] The processor 801 can be a general-purpose processor, such as a central processing unit (CPU), digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0183] Storage medium 802, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. The storage medium can include at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type storage medium, random access memory (RAM), static random access memory (SRAM), programmable read-only memory (PROM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), magnetic storage medium, magnetic disk, optical disk, etc. The storage medium is any other medium capable of carrying or storing desired program code in the form of instructions or data structures that can be accessed by a computer, but is not limited thereto. In the embodiments of this application, storage medium 802 can also be a circuit or any other device capable of implementing storage functions for storing program instructions and / or data.

[0184] Optionally, this application also provides a program product, such as a computer-readable storage medium, including a program that, when executed by a processor, performs the above-described method embodiments.

[0185] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0186] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0187] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in a combination of hardware and software functional units.

[0188] The integrated units implemented as software functional units described above can be stored in a computer-readable storage medium. These software functional units, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for stitching together building facade images from a ground perspective, characterized in that, include: Based on the image information of each captured image, an image pair corresponding to each captured image is generated. The image pair includes at least one adjacent image corresponding to the captured image, and each adjacent image has an overlapping area with the captured image. Based on the texture information of each captured image, a base image is determined and the target region in the base image is marked. Image features of the target region in the base image and image features of the effective region in other captured images are extracted. Based on the target region image features of the base image, the effective region image features of other captured images, and the image pairs corresponding to each captured image, the transformation matrix between the captured image and each adjacent image in each image pair is determined respectively. Based on the transformation matrix between the captured image and each adjacent image in each image pair, and the initial projection plane, the projection position of each captured image is determined, and the captured images are stitched together according to the projection position of each captured image to generate the target stitched image. The step of determining the transformation matrix between the captured image and each adjacent image in each image pair based on the target region image features of the base image, the effective region image features of other captured images, and the image pairs corresponding to each captured image includes: Based on the target region image features of the base image and the effective region image features of each adjacent image in the image pair corresponding to the base image, the transformation matrix between the base image and each adjacent image corresponding to the base image is determined respectively. Based on the transformation matrix between the base image and the adjacent images, the target region image features of the base image are transformed into the adjacent images to obtain the target region image features in the adjacent images; Based on the target region image features of the adjacent images and the effective region image features of each adjacent image in the image pair corresponding to the adjacent images, the transformation matrix between the adjacent images and each adjacent image corresponding to the adjacent images is determined respectively. This process is repeated until the transformation matrix between the captured image and each adjacent image in each image pair is determined. Before determining the projection position of each captured image based on the transformation matrix between the captured image and each adjacent image in each image pair, and the initial projection plane, the method further includes: Based on the spatial distribution of each captured image or the number of image matches, an initial seed image is determined from each captured image, and the plane containing the initial seed image is determined as the initial projection plane; The step of determining the projection position of each captured image based on the transformation matrix between the captured image and each adjacent image in each image pair, and the initial projection plane, includes: Based on the initial seed image and the image pairs corresponding to the initial seed image, determine the adjacent images corresponding to the initial seed image; Based on the optimized transformation matrix between the initial seed image and its adjacent images, the projection position of the adjacent images corresponding to the initial seed image in the initial projection plane is determined. The adjacent image is used as a new seed image. The adjacent image corresponding to the new seed image is determined. Based on the optimized transformation matrix between the new seed image and the adjacent image corresponding to the new seed image, the projection position of the adjacent image corresponding to the new seed image in the initial projection plane is determined. This process is repeated until the projection position of each captured image in the initial projection plane is determined.

2. The method according to claim 1, characterized in that, The step of generating image pairs corresponding to each captured image based on the image information of each captured image includes: Based on the position information of each captured image, the distance between the captured image and other captured images is determined; Based on the distance between the captured image and other captured images, a preset number of adjacent images corresponding to the captured image are determined from the other captured images; Based on the captured image and each adjacent image, an image pair corresponding to the captured image is generated.

3. The method according to claim 1, characterized in that, The step of determining the transformation matrix between the base image and each adjacent image corresponding to the base image based on the target region image features of the base image and the effective region image features of each adjacent image in the image pair corresponding to the base image includes: Based on the target region image features of the base image and the effective region image features of each adjacent image in the image pair corresponding to the base image, feature point matching is performed between the base image and each adjacent image corresponding to the base image to obtain multiple sets of matching feature points between the base image and each adjacent image. The transformation parameters are calculated based on multiple sets of matching feature points between the base image and the adjacent images, as well as the transformation matrix function relationship. Based on the transformation parameters, a transformation matrix is ​​generated between the base image and the adjacent images.

4. The method according to claim 1, characterized in that, After transforming the target region image features of the base image into the adjacent images based on the transformation matrix between the base image and the adjacent images, and obtaining the target region image features in the adjacent images, the process includes: Based on the image capture size and image stitching direction of the adjacent images, the target region image features in the obtained adjacent images are corrected, and based on the corrected target region image features, the transformation matrix between the base image and the adjacent images is optimized to obtain the optimized transformation matrix between the base image and the adjacent images.

5. The method according to claim 1, characterized in that, The step of determining the adjacent image corresponding to the initial seed image based on the initial seed image and the image pair corresponding to the initial seed image includes: The number of feature point matches between the initial seed image and each of its adjacent images is determined respectively. The neighboring image with the largest number of feature point matches with the initial seed image is determined as the adjacent image corresponding to the initial seed image.

6. A device for stitching together architectural facade images from a ground perspective, characterized in that, include: Generation module, extraction module, determination module, and splicing module; The generation module is used to generate image pairs corresponding to each captured image based on the image information of each captured image. The image pairs include at least one adjacent image corresponding to the captured image, and each adjacent image has an overlapping area with the captured image. The extraction module is used to determine a base image based on the texture information of each captured image, mark the target region in the base image, and extract the image features of the target region in the base image and the image features of the effective region in other captured images. The determining module is used to determine the transformation matrix between the captured image and each adjacent image in each image pair based on the target region image features of the base image, the effective region image features in other captured images, and the image pairs corresponding to each captured image. The stitching module is used to determine the projection position of each captured image based on the transformation matrix between the captured image and each adjacent image in each image pair and the initial projection plane, and to stitch the captured images together based on the projection position of each captured image to generate a target stitched image. The determining module is specifically used to determine the transformation matrix between the base image and each of the adjacent images corresponding to the base image based on the target region image features of the base image and the effective region image features of each adjacent image in the image pair corresponding to the base image. Based on the transformation matrix between the base image and the adjacent images, the target region image features of the base image are transformed into the adjacent images to obtain the target region image features in the adjacent images; based on the target region image features of the adjacent images and the effective region image features of each adjacent image in the image pair corresponding to the adjacent images, the transformation matrix between the adjacent images and each adjacent image corresponding to the adjacent images is determined respectively, and so on, until the transformation matrix between the captured image and each adjacent image in each image pair is determined; The determining module is further configured to determine an initial seed image from each captured image based on the spatial distribution of each captured image or the number of image matches, and to determine the plane where the initial seed image is located as the initial projection plane; The determining module is specifically configured to: determine the adjacent image corresponding to the initial seed image based on the initial seed image and the image pair corresponding to the initial seed image; determine the projection position of the adjacent image corresponding to the initial seed image in the initial projection plane based on the optimized transformation matrix between the initial seed image and the adjacent image corresponding to the initial seed image; use the adjacent image as a new seed image, determine the adjacent image corresponding to the new seed image, and determine the projection position of the adjacent image corresponding to the new seed image in the initial projection plane based on the optimized transformation matrix between the new seed image and the adjacent image corresponding to the new seed image; repeat this process until the projection position of each captured image in the initial projection plane is determined.

7. An electronic device, characterized in that, include: The device includes a processor, a storage medium, and a bus. The storage medium stores program instructions executable by the processor. When the electronic device is running, the processor communicates with the storage medium via the bus. The processor executes the program instructions to perform the steps of the ground-view building facade image stitching method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, performs the steps of the building facade image stitching method from a ground perspective as described in any one of claims 1 to 5.

Images