AR-SLAM intelligent navigation system and method for underground station building
Through the AR-SLAM intelligent navigation system, a virtual three-dimensional model of the underground garage is established and the image lighting intensity is adjusted, which solves the problem of low efficiency in searching underground station buildings and realizes fast and accurate navigation path planning.
Patent Information
- Application Number
- CN202510450915.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-11
- Publication Date
- 2025-07-01
AI Technical Summary
In the prior art, when going to an underground station building for maintenance, operation and maintenance personnel need to hike to find the target station building, resulting in inefficient search and unable to meet the power supply service requirements of State Grid Corporation.
Using the AR-SLAM intelligent navigation system, a virtual three-dimensional model is established by obtaining the two-dimensional image sequence of the underground garage, marking the station building location, and adjusting the image lighting intensity under different lighting conditions, comparing the current image in real time to determine the user's location, and providing optimal path navigation.
It improves the accuracy and real-time navigation of underground station buildings, helping users quickly find the target station location, save time and energy, and adapt to different lighting conditions.
Smart Images

Figure CN120232427A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of intelligent navigation, and particularly to an AR-SLAM intelligent navigation system and method for underground station houses. Background Art
[0002] AR-SLAM is an augmented reality technology, which means simultaneous localization and mapping. It refers to the process in which a moving object calculates its own position while constructing an environmental map based on sensor information; it integrates a variety of technical means, can achieve accurate positioning and environmental perception in different scenarios, provides strong technical support for AR applications, and has broad application prospects and important research value.
[0003] Intelligent navigation can autonomously identify user needs and guide users to achieve efficient information retrieval and acquisition, be used to analyze a large amount of traffic data, predict traffic condition changes, optimize path planning, and implement decision-making and control in autonomous driving.
[0004] According to the relevant requirements of State Grid power supply services, when repairing faults in the underground station houses of urban communities, the average time for urban grid operation and maintenance and emergency repair personnel to reach the scene is generally 45 minutes; however, in the prior art, during the process of going to the underground station house for repair, after the operation and maintenance and repair personnel receive the work task and reach the garage, the emergency repair personnel often need to search for the target station house on foot and finally reach the scene; this way of searching for the target station house on foot has no accurate position information of the target station house and can only rely on the way of artificial memory to search on foot in a complex underground garage; during the search process, there are often situations such as taking a detour, turning back, and making multiple phone calls to seek external assistance, resulting in low search efficiency, time-consuming and laborious, and unable to meet the power supply service requirements of the State Grid Corporation. Summary of the Invention
[0005] The purpose of the present invention is to provide an AR-SLAM intelligent navigation system and method for underground station houses to solve the above technical problems.
[0006] The purpose of the present invention can be achieved by the following technical solutions: An AR-SLAM intelligent navigation method for underground station houses includes the following steps: Step S1: Obtain all the underground garages in the urban area, select several shooting points in the underground garage, obtain the shooting images at the shooting points to obtain a two-dimensional image sequence; based on the two-dimensional image sequence, establish a virtual three-dimensional model of the underground garage by means of augmented reality technology; and in the virtual three-dimensional model of the underground garage, obtain all the walking routes in the underground garage, mark the positions of the station houses on the walking routes to obtain an offline three-dimensional model; the station house position is the position where the station house is located in the underground garage; Step S2: Convert the captured image into a grayscale image, obtain the grayscale histogram of the grayscale image, where the abscissa of the grayscale histogram is the grayscale value and the ordinate is the number of pixels; obtain the average value of the number of pixels for each grayscale value in the grayscale histogram, and record the grayscale value with the number of pixels higher than the average value of the number of pixels as the peak grayscale value; Set a high grayscale interval and a low grayscale interval, obtain the number of peak grayscale values with grayscale values in the high grayscale interval, denoted as N high , and obtain the number of peak grayscale values with grayscale values in the low grayscale interval, denoted as N down ; then obtain the illumination intensity value of the captured image , where λ is a preset correction coefficient and λ > 0; Step S3: When the user enters the underground warehouse and starts navigation, obtain the image directly in front of the current user, denoted as the current image; set a standard illumination intensity value, adjust the current image and all the captured images in the two-dimensional image sequence so that the illumination intensity values of the current image and the captured images reach the standard illumination intensity value, and denote the adjusted current image as the current corrected image and the adjusted captured images as the corrected captured images; Compare the current corrected image with each corrected captured image, and obtain the image similarity between the current corrected image and each corrected captured image; select the corrected captured image with the highest image similarity to the current corrected image and with the image similarity exceeding the preset similarity threshold, denoted as the position image; according to the position image, determine the shooting point where the user is currently located in the underground warehouse in the offline three-dimensional model, denoted as the current shooting point; Step S4: Obtain the target station location, where the target station is the station location that the user needs to reach; in the offline three-dimensional model, obtain the optimal path according to the current shooting point and the target station location, and prompt it to the user.
[0007] As a further solution of the present invention: In step S1, the process of selecting the shooting point includes: Obtain the planar area of the underground warehouse, set an interval threshold, and the setting range of the interval threshold is [0.5, 1]; divide the planar area into grids according to the interval threshold to obtain a number of squares, and obtain the center point of each square, denoted as the shooting point.
[0008] As a further solution of the present invention: In step S1, the process of establishing the virtual three-dimensional model includes: Extract the feature points of the captured image based on the feature point extraction algorithm, where the feature points include corner points, edge points, and texture feature points; obtain the camera parameters, and generate 3D point cloud data according to the feature points and camera parameters; perform meshing processing on the generated 3D point cloud data, convert the 3D point cloud data into a 3D model, and obtain a virtual 3D model.
[0009] As a further solution of the present invention: in step S2, the process of obtaining the gray histogram includes: Divide the captured image into several pixels, and obtain the gray values of each pixel according to the gray image; obtain all the gray values that appear in the gray image, and obtain the number of occurrences of each gray value, that is, the number of pixels; establish a rectangular coordinate system with the gray value as the abscissa and the number of pixels as the ordinate; establish a histogram in the rectangular coordinate system according to each gray value and its corresponding number of pixels to obtain the gray histogram.
[0010] As a further solution of the present invention: in step S2, the process of obtaining the average value of the number of pixels includes: Obtain the number of pixels {N1, N2,..., N n} of each gray value in the gray histogram, where N n represents the number of pixels of the nth gray value, and n is the number of types of gray values; then obtain the average value of the number of pixels , where i ∈ [1, n] and i is a positive integer.
[0011] As a further solution of the present invention: in step S2, the process of setting the high gray interval and the low gray interval includes: The gray value setting range of the high gray interval is [240, 255], and the gray value setting range of the low gray interval is [0, 10].
[0012] As a further solution of the present invention: in step S3, the process of obtaining the image similarity includes: Extract the features of the current corrected image to obtain the image features of the current corrected image, and convert the image features into feature vectors, denoted as corrected feature vectors; and extract the features of the captured image to obtain the feature vectors of the captured image, denoted as captured feature vectors; obtain the cosine value of the included angle between the corrected feature vector and the captured feature vector, denoted as the image similarity.
[0013] An AR-SLAM intelligent navigation system for underground station houses, including: Offline model building module: Obtain all underground garages in the urban area, select several shooting points in the underground garage, obtain the shooting images at the shooting points to get a two-dimensional image sequence; based on the two-dimensional image sequence, establish a virtual three-dimensional model of the underground garage using augmented reality technology; and in the virtual three-dimensional model of the underground garage, obtain all walking routes in the underground garage, mark the position of the station building on the walking route to obtain an offline three-dimensional model; the position of the station building is the position where the station building is located in the underground garage. Light intensity determination module: Convert the shooting image into a grayscale image, obtain the grayscale histogram of the grayscale image, the abscissa of the grayscale histogram is the grayscale value, and the ordinate is the number of pixels; obtain the average value of the number of pixels of each grayscale value in the grayscale histogram, and record the grayscale value with the number of pixels higher than the average value of the number of pixels as the peak grayscale value. Set a high grayscale interval and a low grayscale interval, obtain the number of peak grayscale values with grayscale values in the high grayscale interval, denoted as N high , and obtain the number of peak grayscale values with grayscale values in the low grayscale interval, denoted as N down ; then obtain the light intensity value of the shooting image , where λ is a preset correction coefficient and λ > 0; Shooting point determination module: When the user enters the underground garage and starts navigation, obtain the image directly in front of the current user, denoted as the current image; set a standard light intensity value, adjust the current image and all shooting images in the two-dimensional image sequence so that the light intensity values of the current image and the shooting images reach the standard light intensity value, and denote the adjusted current image as the current corrected image, and denote the adjusted shooting images as corrected shooting images; Compare the current corrected image with each corrected shooting image, and obtain the image similarity between the current corrected image and each corrected shooting image; select the corrected shooting image with the highest image similarity to the current corrected image and the image similarity exceeding the preset similarity threshold, denoted as the position image; based on the position image, determine the shooting point where the user is currently located in the underground garage in the offline three-dimensional model, denoted as the current shooting point. Navigation module: Obtain the target station building position, where the target station building is the station building position that the user needs to reach; in the offline three-dimensional model, obtain the optimal path based on the current shooting point and the target station building position, and prompt it to the user.
[0014] Advantages of the present invention: The present invention can more accurately reflect the actual situation of the underground garage by acquiring a two-dimensional image sequence in the underground garage and establishing a virtual three-dimensional model based on augmented reality technology. Marking the position of the station building in the virtual three-dimensional model can help users more accurately find the target station building position. By converting the captured image into a grayscale image and calculating the light intensity value according to the grayscale histogram, the image can be adjusted under different lighting conditions so that the light intensity values of the current image and the captured image reach the standard light intensity value. This can improve the quality and similarity of the image, thereby improving the accuracy of navigation to adapt to different lighting conditions. When the user enters the underground garage and starts navigation, the system will acquire the image directly in front of the current user in real time and compare it with the corrected captured image taken before to determine the shooting point where the user is currently located. This can update and adjust the navigation information in real time according to the actual situation, improving the accuracy and real-time performance of navigation. In the offline three-dimensional model, the optimal path is obtained based on the current shooting point and the target station building position and prompted to the user. This can help users find the target station building position faster, saving time and effort. The present invention provides a more convenient and accurate underground station building navigation service for users through beneficial effects such as improving navigation accuracy, adapting to different lighting conditions, real-time updating and adjustment, and providing optimal path planning. Brief Description of the Drawings
[0015] The present invention will be further described below with reference to the accompanying drawings.
[0016] Figure 1 It is a schematic flowchart of the AR-SLAM intelligent navigation method for underground station buildings of the present invention. Detailed Embodiments
[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.
[0018] Please refer to Figure 1 As shown, the present invention is an AR-SLAM intelligent navigation method for underground station buildings, including the following steps: Step S1: Obtain all the underground garages in the urban area. Select several shooting points in the underground garage, obtain the shooting images at the shooting points, and obtain a two-dimensional image sequence. Based on the two-dimensional image sequence, establish a virtual three-dimensional model of the underground garage using augmented reality technology. And in the virtual three-dimensional model of the underground garage, obtain all the walking routes in the underground garage, and mark the positions of the station buildings on the walking routes to obtain an offline three-dimensional model. The position of the station building is the position where the station building is located in the underground garage. It can be understood that obtaining the shooting images of all the underground garages in the urban area, that is, obtaining the information of the underground garage, which usually includes the layout of the underground garage, the positions of the entrances and exits, etc., for subsequent three-dimensional modeling. Based on the obtained two-dimensional image sequence, establish a virtual three-dimensional model of the underground garage using augmented reality technology. Augmented reality technology can convert two-dimensional images into three-dimensional models through computer vision algorithms, so as to more realistically reflect the actual situation of the underground garage. In the virtual three-dimensional model, the system will obtain all the walking routes in the underground garage and mark the positions of the station buildings on the walking routes. This can help users better understand the layout of the underground garage and facilitate users to find the target station building position. By establishing an offline three-dimensional model, users can still use the navigation system without a network connection. This is very useful in some places with weak network signals or inaccessible Internet, saves the time of network connection, and avoids the problem of poor communication signals in the underground garage. As a preferred embodiment of the present invention, the process of selecting the shooting points includes: Obtain the planar area of the underground garage, set an interval threshold, and the setting range of the interval threshold is [0.5, 1]. Divide the planar area into grids according to the interval threshold to obtain several squares, and obtain the center points of each square, which are recorded as shooting points. It can be understood that in order to determine the distribution density of the shooting points, an interval threshold needs to be set, and the setting range of the interval threshold is [0.5, 1], which can be adjusted according to actual needs. By obtaining the planar area of the underground garage, setting the interval threshold, performing grid division, and selecting shooting points and other steps, basic data for subsequent image acquisition and three-dimensional modeling are provided. As a preferred embodiment of the present invention, the process of establishing the virtual three-dimensional model includes: Extract the feature points of the shooting images based on the feature point extraction algorithm. The feature points include corner points, edge points, and texture feature points. Obtain the camera parameters, and generate three-dimensional point cloud data according to the feature points and camera parameters. Perform grid processing on the generated three-dimensional point cloud data to convert the three-dimensional point cloud data into a three-dimensional model to obtain a virtual three-dimensional model. It should be noted that the process of establishing a three-dimensional model based on the shooting images usually involves the following steps: 1. Data collection: Use a camera to capture an object or scene from different angles to obtain multiple two-dimensional images. These images should cover as many sides of the object as possible and have a certain degree of overlap to facilitate subsequent feature matching and 3D reconstruction. During the shooting process, accurately record the internal parameters of the camera, such as focal length, aperture, shutter speed, etc., and the external parameters, such as the position and orientation of the camera. These parameters are crucial for subsequent camera calibration and 3D modeling. 2. Feature extraction and matching: Extract representative feature points in each two-dimensional image, such as corner points, edge points, texture feature points, etc. Commonly used feature point extraction algorithms include SIFT algorithm, SURF algorithm, ORB algorithm, etc. According to the extracted feature points, perform feature matching between different images to find the corresponding points of the same spatial point in different images. This step can be achieved by calculating the similarity of feature descriptors, such as using metric methods like Euclidean distance and Hamming distance. 3. Camera calibration: Print a black-and-white checkerboard calibration board and fix it on a plane as a reference object with known size and shape. Take several images of the calibration board from different angles and import the calibration images into the program to detect the feature points in the calibration images. According to the camera imaging model, solve the internal and external parameters of the camera based on the corresponding relationship between the coordinates of the matching points in the image and the coordinates in the scene. 4. 3D reconstruction: Based on the matching feature points and camera parameters, use methods such as triangulation to calculate the position of each matching point in 3D space, thereby generating 3D point cloud data. The 3D point cloud data is a model composed of a series of 3D coordinate points, which can roughly reflect the shape and position information of the object. Perform meshing on the generated 3D point cloud to convert it into a polygon mesh model composed of triangular patches, that is, the commonly referred to 3D model. This step can be achieved by methods such as the Poisson surface reconstruction algorithm, or manually edited and optimized using professional 3D modeling software. 5. Post-processing: Optimize the generated 3D model, remove noise points, fill holes, smooth the surface, etc. to improve the quality and accuracy of the model. Map the texture information in the captured two-dimensional images to the surface of the 3D model to make the model more realistic and vivid. This can be achieved by calculating texture coordinates and normal vectors and fitting the image texture to the corresponding positions on the model. Step S2: Convert the captured image into a grayscale image, obtain the grayscale histogram of the grayscale image. The abscissa of the grayscale histogram is the grayscale value, and the ordinate is the number of pixels. Obtain the average value of the number of pixels for each grayscale value in the grayscale histogram, and record the grayscale value with the number of pixels higher than the average value of the number of pixels as the peak grayscale value. Set a high grayscale interval and a low grayscale interval, and obtain the number of peak grayscale values with grayscale values in the high grayscale interval, denoted as N high, and obtain the number of peak gray values in the low gray level range of the gray values, denoted as N down ; then obtain the light intensity value of the captured image , where λ is a preset correction coefficient and λ > 0; It should be noted that first convert the color image into a gray image, and then calculate its gray histogram. The gray histogram reflects the pixel distribution of different gray levels in the image. By analyzing the shape, peak position and distribution range of the histogram, the light intensity of the image can be roughly understood; in areas with sufficient light, more light is reflected from the object surface, and the pixel gray values captured by the camera are relatively high; while in areas with insufficient light, less reflected light and lower pixel gray values; generally speaking, if the peak of the histogram is concentrated in the high gray level area, it means that the overall image is brighter and the light intensity is stronger; on the contrary, if the peak is concentrated in the low gray level area, the light intensity is weaker; that is, the larger the light intensity value, the higher the light intensity of the captured image; otherwise, the light intensity of the captured image is lower; As a preferred embodiment of the present invention, the process of obtaining the gray histogram includes: Divide the captured image into several pixels, and obtain the gray values of each pixel according to the gray image; obtain all the gray values that appear in the gray image, and obtain the number of occurrences of each gray value, that is, the number of pixels; use the gray value as the abscissa and the number of pixels as the ordinate to establish a rectangular coordinate system; establish a histogram in the rectangular coordinate system according to each gray value and its corresponding number of pixels to obtain a gray histogram; As a preferred embodiment of the present invention, the process of obtaining the average value of the number of pixels includes: Obtain the number of pixels {N1, N2,..., N n} of each gray value in the gray histogram, where N n represents the number of pixels of the nth gray value, and n is the number of types of gray values; then obtain the average value of the number of pixels , where i ∈ [1, n] and i is a positive integer; As a preferred embodiment of the present invention, the process of setting the high gray level range and the low gray level range includes: The gray value setting range of the high gray level range is [240, 255], and the gray value setting range of the low gray level range is [0, 10]; It should be noted that for the histogram of a grayscale image, the abscissa represents the grayscale value, and the ordinate represents the number of pixels with that grayscale value. If the peak of the histogram is concentrated in the high-grayscale region, such as near 255, it indicates that there are more pixels with high grayscale values in the image, the overall brightness is higher, and the light intensity is stronger. On the contrary, if the peak is concentrated in the low-grayscale region, such as near 0, it means that the proportion of pixels with low grayscale values in the image is relatively large, the overall brightness is lower, and the light intensity is weaker. Step S3: When the user enters the underground warehouse and starts navigation, obtain the image directly in front of the current user, denoted as the current image; set a standard light intensity value, and adjust the current image and all the captured images in the two-dimensional image sequence so that the light intensity values of the current image and the captured images reach the standard light intensity value. Denote the adjusted current image as the current corrected image, and denote the adjusted captured images as corrected captured images. Compare the current corrected image with each corrected captured image, and obtain the image similarity between the current corrected image and each corrected captured image; select the corrected captured image with the highest image similarity to the current corrected image and with the image similarity exceeding a preset similarity threshold, denoted as the position image; determine the shooting point where the user is currently located in the underground warehouse in the offline three-dimensional model according to the position image, denoted as the current shooting point. It should be noted that during the process of adjusting the current image and all the captured images in the two-dimensional image sequence, the gray histogram of the image can be adjusted to make the pixels more evenly distributed across the entire gray level, thereby enhancing the contrast and brightness of the image. For an image with uneven illumination, histogram equalization can increase the brightness of the darker areas while avoiding overexposure of the brighter areas, achieving the purpose of improving the light intensity. It is worth noting that if the image similarities are all less than or equal to the preset similarity threshold, prompt the user to move forward or backward in a straight line by a preset distance, where the preset distance is less than or equal to the interval threshold; and during the movement, continuously obtain the image directly in front of the user, and continuously compare the image with the captured images in the two-dimensional image sequence until the current shooting point of the user is determined. In a preferred embodiment of the present invention, the process of obtaining the image similarity includes: Extract features from the current corrected image to obtain the image features of the current corrected image, and convert the image features into a feature vector, denoted as the corrected feature vector; and extract features from the captured image to obtain the feature vector of the captured image, denoted as the captured feature vector; obtain the cosine value of the included angle between the corrected feature vector and the captured feature vector, denoted as the image similarity. It can be understood that in order to measure the similarity between two feature vectors, the cosine value of the angle between them is calculated; the closer the cosine value of the angle is to 1, the more similar the two vectors are; the closer it is to -1, the less similar the two vectors are; by calculating the cosine value of the angle between the corrected feature vector and the captured feature vector, the image similarity is obtained, and this similarity value can be used to evaluate the matching degree between the current corrected image and the captured image; Step S4: Obtain the target station location, where the target station is the station location that the user needs to reach; in the offline three-dimensional model, obtain the optimal path according to the current capture point and the target station location, and prompt the user; It should be noted that the process of obtaining the optimal path includes: In the offline three-dimensional model, obtain several paths according to the current capture point and the target station location; sequentially obtain the lengths {L1, L2,..., L m}, where L m represents the length of the m-th path, and m is the total number of paths; For each capture point on the path, obtain the captured image at this time, denoted as the current captured image, and obtain the light intensity value of the current captured image; according to the light intensity values of all the current captured images on the path, obtain the average value of the light intensity values of all the current captured images, denoted as the light intensity mean of the path; obtain the light intensity means {Liv1, Liv2,..., Liv m} of each path, where Liv m represents the light intensity mean of the m-th path; Establish an obstacle image library, where the obstacle image library includes obstacle images, obstacle names, and obstacle volumes; according to the obstacle image library, perform image recognition on each current captured image to obtain the obstacle volume in the current captured image; if there is no obstacle in the current captured image, record the obstacle volume as 0; according to the obstacle volumes in all the current captured images on the path, obtain the total obstacle volume of the path; and then obtain the total obstacle volumes {V1, V2,..., V m} of each path, where V m represents the total obstacle volume of the m-th path; Sequentially obtain the turning times {Not1, Not2,..., Not m} of each path, where Not m represents the turning times of the m-th path; Obtain the shooting point closest to the user, denoted as the nearest shooting point, and obtain the user's shooting image through the nearest shooting point, denoted as the user image; obtain the carried items of the user in the user image based on image recognition technology, where the carried items are items carried by the user, and the carried items include backpacks and toolboxes; obtain the volume of the carried items, set a volume threshold and a volume determination coefficient K. If the volume is less than the volume threshold, record the volume determination coefficient K = 0; If the volume is greater than the volume threshold, record the volume determination coefficient K = 1; then obtain the path cost of the p-th path where w1 is a preset first weight coefficient, w2 is a preset second weight coefficient, and w1 > w2 > 0, and L p represents the length of the p-th path, and Liv p represents the average illumination of the p-th path, and V p represents the total volume of obstacles on the p-th path, and Not p represents the number of turns of the p-th path; select the path with the largest path cost value among all paths, denoted as the optimal path, and recommend the optimal path to the user. An AR-SLAM intelligent navigation system for underground station houses, comprising: Offline model establishment module: Obtain all underground garages in the urban area, select several shooting points in the underground garage, obtain the shooting images at the shooting points to obtain a two-dimensional image sequence; based on the two-dimensional image sequence, establish a virtual three-dimensional model of the underground garage using augmented reality technology; and in the virtual three-dimensional model of the underground garage, obtain all walking routes in the underground garage, and mark the station house positions on the walking routes to obtain an offline three-dimensional model; the station house position is the position where the station house is located in the underground garage; Illumination intensity determination module: Convert the shooting image into a grayscale image, obtain the grayscale histogram of the grayscale image, where the abscissa of the grayscale histogram is the grayscale value and the ordinate is the number of pixels; obtain the average number of pixels of each grayscale value in the grayscale histogram, and record the grayscale value with a pixel number higher than the average number of pixels as the peak grayscale value; Set a high grayscale interval and a low grayscale interval, obtain the number of peak grayscale values with grayscale values in the high grayscale interval, denoted as N high and obtain the number of peak grayscale values with grayscale values in the low grayscale interval, denoted as N down ; then obtain the illumination intensity value of the shooting image where λ is a preset correction coefficient and λ > 0; Shooting point determination module: When the user enters the underground warehouse and starts navigation, obtain the image directly in front of the current user, denoted as the current image; set a standard light intensity value, adjust the current image and all the captured images in the two-dimensional image sequence so that the light intensity values of the current image and the captured images reach the standard light intensity value, and denote the adjusted current image as the current corrected image, and denote the adjusted captured images as corrected captured images; Compare the current corrected image with each corrected captured image, and obtain the image similarity between the current corrected image and each corrected captured image; select the corrected captured image with the highest image similarity to the current corrected image and whose image similarity exceeds the preset similarity threshold, denoted as the position image; determine the shooting point where the user is currently located in the underground warehouse in the offline three-dimensional model according to the position image, denoted as the current shooting point; Navigation module: Obtain the target station location, where the target station is the station location that the user needs to reach; in the offline three-dimensional model, obtain the optimal path according to the current shooting point and the target station location, and prompt it to the user.
[0019] The above has described an embodiment of the present invention in detail, but the content is only a preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of the application of the present invention should still fall within the scope covered by the patent of the present invention.
Claims
1. An AR-SLAM intelligent navigation method for underground station buildings, characterized in that: The following steps are involved: Step S1: Acquire all underground garages in the urban area, select a number of shooting points in the underground garage, acquire the shooting images at the shooting points, and obtain a two-dimensional image sequence; and establish a virtual three-dimensional model of the underground garage based on the two-dimensional image sequence based on augmented reality technology; In the virtual three-dimensional model of the underground garage, all walking routes in the underground garage are obtained, and the station building position is marked on the walking route to obtain an offline three-dimensional model; the station building position is the position of the station building in the underground garage; Step S2: converting the captured image into a grayscale image, obtaining a grayscale histogram of the grayscale image, wherein the abscissa of the grayscale histogram is the grayscale value and the ordinate is the number of pixels; obtaining the mean number of pixels for each grayscale value in the grayscale histogram, and recording the grayscale value whose number of pixels is higher than the mean number of pixels as the peak grayscale value; Set the high grayscale interval and the low grayscale interval, and obtain the number of peak grayscale values in the high grayscale interval, recorded as N high , and obtain the number of peak grayscale values in the low grayscale interval, recorded as N down ; then the light intensity value of the captured image is obtained ,in λ is the preset correction factor and λ >0; Step S3: When the user enters the underground garage and starts navigating, an image directly in front of the user is obtained and recorded as the current image; Setting a standard light intensity value, adjusting the current image and all captured images in the two-dimensional image sequence so that the light intensity values of the current image and the captured images reach the standard light intensity value, and recording the adjusted current image as a current corrected image, and recording the adjusted captured image as a corrected captured image; The current corrected image is compared with each corrected shot image, and the image similarity between the current corrected image and each corrected shot image is obtained; a corrected shot image having the highest image similarity with the current corrected image and an image similarity exceeding a preset similarity threshold is selected and recorded as a position image; based on the position image, the shooting point where the user is currently located in the underground warehouse is determined in the offline three-dimensional model and recorded as the current shooting point; Step S4: obtaining the location of a target station building, where the target station building is the location of the station building that the user needs to reach; In the offline three-dimensional model, the optimal path is obtained according to the current shooting point and the target station building position, and is prompted to the user.
2. The AR-SLAM intelligent navigation method for underground station building according to claim 1 is characterized in that: In step S1, the process of selecting the shooting point includes: The plane area of the underground warehouse is obtained, and an interval threshold is set, wherein the setting range of the interval threshold is [0.5, 1]. The plane area is gridded according to the interval threshold to obtain a plurality of squares, and the center point of each square is obtained and recorded as a shooting point.
3. The AR-SLAM intelligent navigation method for underground station building according to claim 1 is characterized in that: In step S1, the process of establishing the virtual three-dimensional model includes: The feature points of the captured image are extracted based on a feature point extraction algorithm, wherein the feature points include corner points, edge points and texture feature points; camera parameters are acquired, and three-dimensional point cloud data are generated according to the feature points and the camera parameters; the generated three-dimensional point cloud data is gridded, and the three-dimensional point cloud data is converted into a three-dimensional model to obtain a virtual three-dimensional model.
4. The AR-SLAM intelligent navigation method for underground station building according to claim 1, characterized in that: In step S2, the grayscale histogram acquisition process includes: The captured image is divided into a number of pixels, and the grayscale value of each pixel is obtained according to the grayscale image; all grayscale values appearing in the grayscale image are obtained, and the number of occurrences of each grayscale value, that is, the number of pixels, is obtained; a rectangular coordinate system is established with the grayscale value as the horizontal coordinate and the number of pixels as the vertical coordinate; a histogram is established in the rectangular coordinate system according to each grayscale value and its corresponding number of pixels to obtain a grayscale histogram.
5. The AR-SLAM intelligent navigation method for underground station building according to claim 1, characterized in that: In step S2, the process of obtaining the mean value of the number of pixels includes: Get the number of pixels of each gray value in the gray histogram {N1, N2, ..., N n }, where N n represents the number of pixels of the nth grayscale value, where n is the number of grayscale values; then the mean value of the number of pixels is obtained , where i∈[1,n] and i is a positive integer.
6. The AR-SLAM intelligent navigation method for underground station building according to claim 1, characterized in that: In step S2, the process of setting the high grayscale interval and the low grayscale interval includes: The grayscale value setting range of the high grayscale interval is [240, 255], and the grayscale value setting range of the low grayscale interval is [0, 10].
7. The AR-SLAM intelligent navigation method for underground station building according to claim 1, characterized in that: In step S3, the process of obtaining the image similarity includes: Perform feature extraction on the current corrected image to obtain image features of the current corrected image, and convert the image features into feature vectors, recorded as corrected feature vectors; perform feature extraction on the captured image to obtain feature vectors of the captured image, recorded as captured feature vectors; obtain the cosine value of the angle between the corrected feature vector and the captured feature vector, recorded as image similarity.
8. The AR-SLAM intelligent navigation system for underground station buildings is characterized by: include: Offline model building module: obtain all underground garages in the urban area, select several shooting points in the underground garage, obtain the shooting images at the shooting points, and obtain a two-dimensional image sequence; according to the two-dimensional image sequence, build a virtual three-dimensional model of the underground garage based on augmented reality technology; In the virtual three-dimensional model of the underground garage, all walking routes in the underground garage are obtained, and the station building position is marked on the walking route to obtain an offline three-dimensional model; the station building position is the position of the station building in the underground garage; Light intensity determination module: convert the captured image into a grayscale image, obtain a grayscale histogram of the grayscale image, wherein the abscissa of the grayscale histogram is the grayscale value and the ordinate is the number of pixels; obtain the mean number of pixels for each grayscale value in the grayscale histogram, and record the grayscale value whose number of pixels is higher than the mean number of pixels as the peak grayscale value; Set the high grayscale interval and the low grayscale interval, and obtain the number of peak grayscale values in the high grayscale interval, recorded as N high , and obtain the number of peak grayscale values in the low grayscale interval, recorded as N down ; then the light intensity value of the captured image is obtained ,in λ is the preset correction factor and λ >0; Shooting point determination module: When the user enters the underground garage and starts navigating, the image directly in front of the user is obtained and recorded as the current image; Setting a standard light intensity value, adjusting the current image and all captured images in the two-dimensional image sequence so that the light intensity values of the current image and the captured images reach the standard light intensity value, and recording the adjusted current image as a current corrected image, and recording the adjusted captured image as a corrected captured image; The current corrected image is compared with each corrected shot image, and the image similarity between the current corrected image and each corrected shot image is obtained; a corrected shot image having the highest image similarity with the current corrected image and an image similarity exceeding a preset similarity threshold is selected and recorded as a position image; based on the position image, the shooting point where the user is currently located in the underground warehouse is determined in the offline three-dimensional model and recorded as the current shooting point; Navigation module: obtain the location of the target station building, where the target station building is the location of the station building that the user needs to reach; In the offline three-dimensional model, the optimal path is obtained according to the current shooting point and the target station building position, and is prompted to the user.
