Visual positioning method and system for low-texture high-repetition environment

Abstract

The invention discloses a visual positioning method and system for a low-texture high-repetition environment, and the method comprises the steps: data collection: collecting an environment image in a preset range, and recording the corresponding coordinate information; performing image preprocessing and image cutting on the acquired environment image, and performing wall view angle correction and coordinate conversion based on transverse and vertical line features; image compression is carried out based on the wall cutting blocks, and an environment wall image database in a preset range is constructed in combination with the coordinate information; space positioning: acquiring an environment image of a target position, performing wall view angle correction based on transverse and vertical line features, and performing wall dicing and image preprocessing; retrieving the most similar wall surface image from an environment wall surface image database based on the wall surface image after image preprocessing, and obtaining corresponding coordinate information; and fusing the positioning results of the plurality of wall surface images according to a visual angle, and outputting a final positioning result. According to the invention, visual positioning in a low-texture high-repetition environment can be realized.

Description

Technical Field

[0001] The present invention relates to the field of visual positioning technology, and in particular to a visual positioning method and system for a low-texture and high-repetitive environment. Background Art

[0002] Visual localization algorithms are an important research direction in the field of computer vision and robotics, aiming to determine the position and posture of a camera or object in an environment by analyzing image or video data. These algorithms include feature point matching methods (such as SIFT, SURF, ORB), optical flow methods (such as Lucas-Kanade method), direct methods (such as DSO), deep learning-based methods (such as PoseNet), visual odometry, visual SLAM (such as ORB-SLAM), marker-based localization (such as ArUco), model-based localization (such as PnP algorithm), image retrieval-based localization (such as DBoW), and hybrid methods (such as SfM, VIO). Each algorithm has its own unique advantages and applicable scenarios. In practical applications, it is often necessary to select the appropriate algorithm based on factors such as accuracy requirements, real-time requirements, environmental complexity, and hardware limitations, or combine multiple algorithms and sensor data to achieve a more robust and accurate localization effect.

[0003] However, the above algorithms have difficulty in dealing with low-texture and high-repetitive environments and are prone to positioning errors. For example, low-texture and high-repetitive environments lead to low positioning accuracy, inaccurate image feature extraction leads to large positioning errors, single and repetitive positioning information cannot fully utilize multi-source data, target positioning is unstable in complex scenes, and the ability to recognize different types of walls is limited. Summary of the invention

[0004] In order to solve the problem that it is difficult to achieve position positioning through traditional features such as traditional corners in low-texture and high-repetitive environments (such as tunnel environments), the present invention proposes a visual positioning method and system for low-texture and high-repetitive environments, which collects and cuts wall images and searches in the collected wall image set to obtain the wall image position coordinates and the shooting position coordinates, thereby achieving visual navigation in low-texture and high-repetitive environments.

[0005] The technical solution adopted by the present invention is as follows:

[0006] A visual localization method for low-texture and high-repetitive environments, comprising:

[0007] Data collection: Collect environmental images within a preset range and record the corresponding coordinate information; perform image preprocessing and image segmentation on the collected environmental images, and perform wall perspective correction and coordinate conversion based on horizontal and vertical line features; perform image compression based on wall segmentation, and build an environmental wall image database within a preset range in combination with coordinate information;

[0008] Spatial positioning: Collect environmental images of the target location, perform wall perspective correction based on horizontal and vertical line features, and perform wall segmentation and image preprocessing; based on the wall image after image preprocessing, retrieve the most similar wall image from the environmental wall image database and obtain the corresponding coordinate information; fuse the positioning results of multiple wall images according to the perspective and output the final positioning result.

[0009] Furthermore, in the data collection, wall perspective correction and coordinate conversion are performed based on horizontal and vertical line features, including: using the detected horizontal and vertical line information to perform perspective correction on the wall image, and obtaining corresponding coordinate information according to the corrected wall image transformation.

[0010] Furthermore, in the data collection, image compression is performed based on wall segmentation, including: identifying wall segmentation lines and cutting out image blocks of the wall, and compressing the image after segmentation.

[0011] Furthermore, in the spatial positioning, the retrieval methods for retrieving the most similar wall image from the environmental wall image database include: a traditional method based on feature descriptors, a traditional method based on global features, a deep learning method based on convolutional neural networks, an approximate nearest neighbor search method based on hashing, a representation learning method based on image embedding, and a feature aggregation method.

[0012] Furthermore, in the data collection and spatial positioning, image preprocessing includes denoising processing, contrast enhancement processing, filtering processing and flatness processing.

[0013] A visual positioning system for low-texture and high-repetitive environments, comprising:

[0014] The data acquisition module is configured to acquire environmental images within a preset range and record corresponding coordinate information; perform image preprocessing and image segmentation on the acquired environmental images, and perform wall perspective correction and coordinate conversion based on horizontal and vertical line features; perform image compression based on wall segmentation, and construct an environmental wall image database within a preset range in combination with the coordinate information;

[0015] The spatial positioning module is configured to collect environmental images of the target location, perform wall perspective correction based on horizontal and vertical line features, and perform wall segmentation and image preprocessing; based on the wall image after image preprocessing, retrieve the most similar wall image from the environmental wall image database to obtain the corresponding coordinate information; fuse the positioning results of multiple wall images according to the perspective, and output the final positioning result.

[0016] Furthermore, in the data acquisition module, wall perspective correction and coordinate conversion are performed based on horizontal and vertical line features, including: using the detected horizontal and vertical line information to perform perspective correction on the wall image, and obtaining corresponding coordinate information according to the corrected wall image transformation.

[0017] Furthermore, in the data acquisition module, image compression is performed based on wall segmentation, including: identifying wall segmentation lines and cutting out image blocks of the wall, and compressing the image after segmentation.

[0018] Furthermore, in the spatial positioning module, the retrieval methods for retrieving the most similar wall image from the environmental wall image database include: a traditional method based on feature descriptors, a traditional method based on global features, a deep learning method based on convolutional neural networks, an approximate nearest neighbor search method based on hashing, a representation learning method based on image embedding, and a feature aggregation method.

[0019] Furthermore, in the data acquisition module and the spatial positioning module, image preprocessing includes denoising processing, contrast enhancement processing, filtering processing and flatness processing.

[0020] The beneficial effects of the present invention are:

[0021] The present invention corrects the wall perspective by detecting the horizontal and vertical line features in the image, adopts the wall slicing technology and combines it with the preprocessing step to improve the accuracy of feature extraction and reduce positioning errors. Furthermore, the positioning information fusion technology is used to integrate multiple positioning results to enhance the reliability of data and the accuracy of positioning. Especially in complex scenes, the regional processing and positioning improve the stability and robustness of the system, while optimizing the recognition ability of various wall types. In addition, the preprocessing step effectively reduces the impact of image noise and distortion on positioning, thereby improving the overall positioning accuracy and adaptability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a flow chart of a visual positioning method for a low-texture and high-repetitive environment according to Example 1 of the present invention.

[0023] Figure 2 This is a data collection flow chart of Example 1 of the present invention.

[0024] Figure 3 This is a flow chart of spatial positioning according to Embodiment 1 of the present invention. DETAILED DESCRIPTION

[0025] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, the specific implementation methods of the present invention are now described. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention, that is, the embodiments described are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present invention.

[0026] Example 1

[0027] like Figure 1 As shown, this embodiment provides a visual positioning method for a low-texture and high-repetitive environment, including:

[0028] Data collection: Collect environmental images within a preset range and record the corresponding coordinate information; perform image preprocessing and image segmentation on the collected environmental images, and perform wall perspective correction and coordinate conversion based on horizontal and vertical line features; perform image compression based on wall segmentation, and build an environmental wall image database within a preset range in combination with coordinate information;

[0029] Spatial positioning: Collect environmental images of the target location, perform wall perspective correction based on horizontal and vertical line features, and perform wall segmentation and image preprocessing; based on the wall image after image preprocessing, retrieve the most similar wall image from the environmental wall image database and obtain the corresponding coordinate information; fuse the positioning results of multiple wall images according to the perspective and output the final positioning result.

[0030] Preferably, if Figure 2 As shown, the data collection process of this embodiment can be implemented by the following steps:

[0031] 1. Start: Start the data collection process.

[0032] 2. Take photos and record coordinates: Use the camera to capture the target image and record the camera's location information during shooting.

[0033] 3. Block image preprocessing: Preprocess the captured image and divide the image into blocks to reduce noise and improve processing efficiency.

[0034] 4. Detect horizontal and vertical lines: Detect horizontal and vertical line features in the cut image for subsequent image correction and coordinate transformation.

[0035] 5. Wall perspective correction / coordinate conversion: Use the detected horizontal and vertical line information to correct the image perspective, and obtain the corresponding coordinate information based on the corrected image transformation.

[0036] 6. Wall cutting: Cut the wall into blocks.

[0037] 7. Block image compression: compress the block images to reduce storage space and improve subsequent processing efficiency.

[0038] 8. Image + coordinate storage: The processed image and the corresponding coordinate information are stored together as the basic data for subsequent retrieval and positioning.

[0039] Preferably, if Figure 3 As shown, the spatial positioning process of this embodiment can be implemented by the following steps:

[0040] 1. Start: The algorithm starts and prepares to enter the image processing and positioning process.

[0041] 2. Take photos: The system captures images of the environment for subsequent image processing and positioning operations.

[0042] 3. Detect horizontal and vertical lines: In the captured image, detect horizontal and vertical lines. These lines may correspond to the boundaries of wall corners, pipes, wiring ducts, or other geometric structures.

[0043] 4. Wall perspective correction: Based on the detected horizontal and vertical lines, the wall image is corrected through affine transformation to ensure that the wall image is facing the right direction.

[0044] 5. Wall Slicing: The system identifies the wall segmentation line and cuts out the image block of the wall, which can analyze the specific area more finely. The cutting line is shown in the figure below.

[0045] 6. Sliced ​​image preprocessing: Preprocess the sliced ​​wall image, including but not limited to denoising, contrast enhancement, filtering, smoothing, etc., in order to improve the accuracy of subsequent feature extraction.

[0046] 7. Sliced ​​image retrieval: Retrieve specific features or patterns from the preprocessed image, which can be compared with the reference image in the pre-collected wall image database. The retrieval algorithm can include traditional methods based on feature descriptors (such as SIFT, SURF, ORB, AKAZE), traditional methods based on global features (such as color histograms, texture and shape features), deep learning methods based on convolutional neural networks (such as pre-trained CNN features, Siamese Network, TripletLoss, contrastive learning), hash-based approximate nearest neighbor search (such as LSH, Deep Hashing), and representation learning based on image embedding (such as deep metric learning and visual semantic embedding). In addition, there are feature aggregation methods such as BoVW, Fisher Vectors, and VLAD.

[0047] 8. Sliced ​​image positioning: Based on the search results, determine the image that is most similar to the sliced ​​image and obtain the position coordinate information of the image.

[0048] 9. Positioning information fusion: The positioning results of multiple images are fused according to the viewing angle to improve the accuracy and reliability of positioning.

[0049] 10. Final positioning output: The algorithm outputs the final positioning result, including the location of the target in the image or in the actual physical space.

[0050] It should be noted that the visual positioning method of this embodiment can be applied to fields such as building inspection and navigation, and accurate spatial positioning can be achieved through analysis and processing of visual images.

[0051] In addition, the visual positioning method of this embodiment can be implemented on a mobile terminal by writing H5 or APP; the environment image is captured by the mobile phone camera, and the above-mentioned process is processed and then matched with the image data set captured in advance, so as to achieve visual positioning.

[0052] Example 2

[0053] This embodiment is based on embodiment 1:

[0054] This embodiment provides a visual positioning system for a low-texture and high-repetitive environment, including:

[0055] The data acquisition module is configured to acquire environmental images within a preset range and record corresponding coordinate information; perform image preprocessing and image segmentation on the acquired environmental images, and perform wall perspective correction and coordinate conversion based on horizontal and vertical line features; perform image compression based on wall segmentation, and construct an environmental wall image database within a preset range in combination with the coordinate information;

[0056] The spatial positioning module is configured to collect environmental images of the target location, perform wall perspective correction based on horizontal and vertical line features, and perform wall segmentation and image preprocessing; based on the wall image after image preprocessing, retrieve the most similar wall image from the environmental wall image database to obtain corresponding coordinate information; fuse the positioning results of multiple wall images according to the perspective, and output the final positioning result.

[0057] Preferably, in the data acquisition module, wall perspective correction and coordinate conversion are performed based on horizontal and vertical line features, including: using the detected horizontal and vertical line information to perform perspective correction on the wall image, and obtaining corresponding coordinate information according to the corrected wall image transformation.

[0058] Preferably, in the data acquisition module, image compression is performed based on wall segmentation, including: identifying wall segmentation lines and cutting out image blocks of the wall, and compressing the image after segmentation.

[0059] Preferably, in the spatial positioning module, the retrieval method for retrieving the most similar wall image from the environmental wall image database includes: a traditional method based on feature descriptors, a traditional method based on global features, a deep learning method based on convolutional neural networks, an approximate nearest neighbor search method based on hashing, a representation learning method based on image embedding, and a feature aggregation method.

[0060] Preferably, in the data acquisition module and the spatial positioning module, image preprocessing includes denoising processing, contrast enhancement processing, filtering processing and flatness processing.

[0061] Example 3

[0062] This embodiment is based on embodiment 1:

[0063] This embodiment provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, a visual positioning method for a low-texture and high-repetitive environment of embodiment 1 is implemented. The computer program may be in source code form, object code form, executable file, or some intermediate form.

[0064] Example 4

[0065] This embodiment is based on embodiment 1:

[0066] The present embodiment provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements a visual positioning method for a low-texture, high-repetitive environment of embodiment 1. The computer program may be in source code form, object code form, executable file, or some intermediate form, etc. The storage medium includes: any entity or device capable of carrying computer program code, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content contained in the storage medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the storage medium does not include electric carrier signals and telecommunication signals.

[0067] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity of description, they are expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the order of the actions described, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

Claims

1. A visual positioning method for low-texture and high-repetitive environments, characterized in that: include: Data collection: Collect environmental images within a preset range and record the corresponding coordinate information; Perform image preprocessing and image segmentation on the collected environmental images, and perform wall perspective correction and coordinate conversion based on horizontal and vertical line features; perform image compression based on wall segmentation, and build an environmental wall image database within a preset range based on coordinate information; Spatial positioning: Collect the environmental image of the target location, perform wall perspective correction based on horizontal and vertical line features, and perform wall segmentation and image preprocessing. Based on the preprocessed wall image, retrieve the most similar wall image from the environmental wall image database to obtain the corresponding coordinate information. The positioning results of multiple wall images are fused according to the viewing angle and the final positioning result is output.

2. The visual positioning method for low-texture and high-repetitive environments according to claim 1, characterized in that: In the data collection, wall perspective correction and coordinate conversion are performed based on horizontal and vertical line features, including: using the detected horizontal and vertical line information to perform perspective correction on the wall image, and obtaining corresponding coordinate information according to the corrected wall image transformation.

3. The visual positioning method for low-texture and high-repetitive environments according to claim 1, characterized in that: In the data collection, image compression is performed based on wall segmentation, including: identifying wall segmentation lines and cutting out image blocks of the wall, and compressing the image after segmentation.

4. The visual positioning method for low-texture and high-repetitive environments according to claim 1, characterized in that: In the spatial positioning, the retrieval methods for retrieving the most similar wall image from the environmental wall image database include: a traditional method based on feature descriptors, a traditional method based on global features, a deep learning method based on convolutional neural networks, an approximate nearest neighbor search method based on hashing, a representation learning method based on image embedding, and a feature aggregation method.

5. The visual positioning method for low-texture and high-repetitive environments according to claim 1, characterized in that: In the data collection and spatial positioning, image preprocessing includes denoising, contrast enhancement, filtering and flatness processing.

6. A visual positioning system for low-texture and high-repetitive environments, characterized in that: include: A data acquisition module is configured to acquire an environmental image within a preset range and record corresponding coordinate information; Perform image preprocessing and image segmentation on the collected environmental images, and perform wall perspective correction and coordinate conversion based on horizontal and vertical line features; perform image compression based on wall segmentation, and build an environmental wall image database within a preset range based on coordinate information; The spatial positioning module is configured to collect the environment image of the target location, perform wall perspective correction based on horizontal and vertical line features, and perform wall segmentation and image preprocessing; Based on the wall image after image preprocessing, the most similar wall image is retrieved from the environmental wall image database to obtain the corresponding coordinate information; The positioning results of multiple wall images are fused according to the viewing angle and the final positioning result is output.

7. The visual positioning system for low-texture and high-repetitive environments according to claim 6, characterized in that: In the data acquisition module, wall viewing angle correction and coordinate conversion are performed based on horizontal and vertical line features, including: using the detected horizontal and vertical line information to perform viewing angle correction on the wall image, and obtaining corresponding coordinate information according to the corrected wall image transformation.

8. The visual positioning system for low-texture and high-repetitive environments according to claim 6, characterized in that: In the data acquisition module, image compression is performed based on wall segmentation, including: identifying wall segmentation lines and cutting out image blocks of the wall, and compressing the image after segmentation.

9. The visual positioning system for low-texture and high-repetitive environments according to claim 6, characterized in that: In the spatial positioning module, the retrieval methods for retrieving the most similar wall image from the environmental wall image database include: a traditional method based on feature descriptors, a traditional method based on global features, a deep learning method based on convolutional neural networks, an approximate nearest neighbor search method based on hashing, a representation learning method based on image embedding, and a feature aggregation method.

10. The visual positioning system for low-texture and high-repetitive environments according to claim 6, characterized in that: In the data acquisition module and the space positioning module, image preprocessing includes denoising processing, contrast enhancement processing, filtering processing and flatness processing.

Images