CN111340882A - Image-based indoor positioning method and device - Google Patents
Image-based indoor positioning method and device Download PDFInfo
- Publication number
- CN111340882A CN111340882A CN202010104620.5A CN202010104620A CN111340882A CN 111340882 A CN111340882 A CN 111340882A CN 202010104620 A CN202010104620 A CN 202010104620A CN 111340882 A CN111340882 A CN 111340882A
- Authority
- CN
- China
- Prior art keywords
- image
- indoor
- bim
- indoor image
- matrix
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 53
- 239000011159 matrix material Substances 0.000 claims abstract description 157
- 238000004364 calculation method Methods 0.000 claims abstract description 44
- 238000000605 extraction Methods 0.000 claims abstract description 19
- 239000013598 vector Substances 0.000 claims description 25
- 230000011218 segmentation Effects 0.000 claims description 21
- 238000004590 computer program Methods 0.000 claims description 3
- 238000004891 communication Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 238000012549 training Methods 0.000 description 5
- 230000003287 optical effect Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 238000011176 pooling Methods 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000013519 translation Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000000779 smoke Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/20—Instruments for performing navigational calculations
- G01C21/206—Instruments for performing navigational calculations specially adapted for indoor navigation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- Automation & Control Theory (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the invention provides an indoor positioning method and device based on images, wherein the method comprises the following steps: acquiring an indoor image of a building shot by a user, and performing feature extraction on the indoor image and a BIM image of the building in a BIM image library based on a convolution calculation model to acquire a feature matrix of the indoor image and a feature matrix of the BIM image; dividing the indoor image, acquiring a weight matrix of the indoor image according to a division result, and weighting the feature matrices of the indoor image and the BIM image respectively by using the weight matrix of the indoor image; matching the indoor image with each BIM image according to the weighting result of the characteristic matrix of the indoor image and each BIM image, and taking the position coordinate corresponding to the BIM image which is most matched with the indoor image as the position coordinate of the user; wherein, the BIM image and the position coordinate are stored in advance in an associated manner. The embodiment of the invention realizes automatic indoor positioning based on the image and has high positioning precision.
Description
Technical Field
The invention belongs to the technical field of indoor positioning, and particularly relates to an indoor positioning method and device based on images.
Background
Indoor positioning technology is widely applied to large places such as libraries, gymnasiums, underground garages, goods warehouses and the like. Quick positioning of personnel and articles can be achieved. The traditional indoor positioning method mostly adopts radio, infrared rays, ultrasonic waves and the like as a propagation medium.
The first method is that a positioning target uses an infrared (infrared) identification as a moving point, emits modulated infrared rays and is positioned by receiving the infrared rays through an optical sensor arranged indoors; the second is to weave infrared net to cover the space to be measured by a plurality of pairs of transmitters and receivers to directly position the moving target. The infrared technology is mature and relatively high in indoor positioning accuracy, but due to the fact that infrared rays can only be transmitted in a sight distance mode, penetrability is extremely poor, the infrared positioning device cannot normally work when an identification is shielded, and the infrared positioning device is extremely easily influenced by light, smoke and other environmental factors. The transmission distance of infrared rays is not long, so that the receiving end needs to be installed at the back of each shielding part or even at a corner in the layout in any mode, the layout is complex, the cost is improved, and the positioning effect is limited.
The ultrasonic indoor positioning system is developed based on an ultrasonic ranging system and consists of a plurality of transponders and a main range finder. The main distance meter is placed on a measured object, the same radio signal is transmitted to the transponder with a fixed position, the transponder transmits an ultrasonic signal to the main distance meter after receiving the signal, and the position of the object is determined by utilizing algorithms such as a reflection type distance measurement method and triangulation. The whole precision of ultrasonic indoor positioning is very high, and the structure is simple relatively, has certain penetrability and ultrasonic wave itself has very strong interference killing feature. However, the ultrasonic wave has large attenuation in the air, is not suitable for large-scale occasions, and is greatly influenced by multipath effect and non-line-of-sight propagation in reflection ranging, so that the investment of bottom hardware facilities needing accurate analysis and calculation is caused, and the cost is too high.
In summary, the conventional indoor positioning method requires a signal transmitter to be installed indoors in advance, which causes inconvenience in cost and use and poor positioning effect.
Disclosure of Invention
In order to overcome the problems of high cost and poor positioning effect of the conventional indoor positioning method or at least partially solve the problems, embodiments of the present invention provide an indoor positioning method and apparatus based on an image.
According to a first aspect of the embodiments of the present invention, there is provided an image-based indoor positioning method, including:
acquiring an indoor image of a building shot by a user, and performing feature extraction on the indoor image and a BIM image of the building acquired in advance in a BIM image library based on a convolution calculation model to acquire a feature matrix of the indoor image and a feature matrix of the BIM image;
dividing the indoor image, acquiring a weight matrix of the indoor image according to a division result, and weighting the feature matrix of the indoor image and the feature matrix of the BIM image by using the weight matrix of the indoor image;
matching the indoor image with each BIM image according to the weighting result of the characteristic matrix of the indoor image and each BIM image, and taking the position coordinate corresponding to the BIM image which is most matched with the indoor image as the position coordinate of the user; and the BIM image and the position coordinate are stored in a pre-associated mode.
Specifically, the step of extracting the features of the indoor image and the BIM image based on a convolution calculation model to obtain the feature matrix of the indoor image and the feature matrix of the BIM image includes:
scanning the indoor image and the BIM image by using a sliding window to obtain sub-images of the indoor image and the BIM image;
performing convolution calculation on the subgraph of the indoor image, the subgraph of the BIM image, the indoor image and the BIM image based on a convolution calculation model;
and taking the convolution calculation result of the sub-image of the indoor image and the indoor image as a characteristic matrix of the indoor image, and taking the convolution calculation result of the sub-image of the BIM image and the BIM image as the characteristic matrix of the BIM image.
Specifically, the step of scanning the indoor image and the BIM image by using a sliding window to acquire subgraphs of the indoor image and the BIM image comprises:
aligning the upper left corner of the sliding window with the upper left corners of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image;
aligning the upper right corner of the sliding window with the upper right corners of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image;
aligning the lower left corner of the sliding window with the lower left corners of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image;
aligning the lower right corner of the sliding window with the lower right corners of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image;
and aligning the central point of the sliding window with the central points of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image.
Specifically, the convolution calculation model is a VGG16 model;
correspondingly, the step of extracting the features of the indoor image and the BIM image based on the convolution calculation model to obtain the feature matrix of the indoor image and the feature matrix of the BIM image comprises the following steps:
taking the indoor image as an input of the VGG16 model, and taking an output of maxpool in a fourth layer of the VGG16 model as a feature matrix of the indoor image;
and taking the BIM image as an input of the VGG16 model, and taking an output of maxpool in a fourth layer of the VGG16 model as a feature matrix of the BIM image.
Specifically, the step of segmenting the indoor image and acquiring a weight matrix of the indoor image according to a segmentation result includes:
performing semantic segmentation on the indoor image based on a DeepLabv3+ model to obtain a segmentation result of the indoor image;
dividing the indoor image into grids, if the segmentation result of the indoor image in any grid comprises a foreground, setting the weight of the grid as a first preset weight, and otherwise, setting the weight of the grid as a second preset weight; wherein the first preset weight is smaller than the second preset weight;
and constructing a weight matrix of the indoor image according to the weights of all the grids.
Specifically, the step of weighting the feature matrix of the indoor image and the feature matrix of the BIM image respectively by using the weight matrix of the indoor image includes:
adjusting the size of the weight matrix of the indoor image so that the size of the weight matrix of the indoor image is the same as the size of the feature matrix of the indoor image;
and performing Hadamard multiplication on the adjusted weight matrix of the indoor image and the feature matrices of the indoor image and the BIM image respectively to obtain the weighting results of the feature matrices of the indoor image and each BIM image.
Specifically, the step of matching the indoor image with each of the BIM images according to the weighting result of the feature matrices of the indoor image and each of the BIM images, and using the position coordinate corresponding to the BIM image that is most matched with the indoor image as the position coordinate of the user includes:
expanding the weighting result of the feature matrix of the indoor image to obtain the feature vector of the indoor image, and expanding the weighting result of the feature matrix of each BIM image to obtain the feature vector of each BIM image;
calculating cosine similarity between the feature vectors of the indoor images and the feature vectors of the BIM images, and taking the BIM image corresponding to the maximum cosine similarity as the BIM image which is most matched with the indoor images;
and taking the position coordinate corresponding to the BIM image which is most matched with the indoor image as the position coordinate of the user.
According to a second aspect of the embodiments of the present invention, there is provided an image-based indoor positioning apparatus, including:
the extraction module is used for acquiring an indoor image of a building shot by a user, extracting features of the indoor image and a BIM image of the building acquired in advance in a BIM image library based on a convolution calculation model, and acquiring a feature matrix of the indoor image and a feature matrix of the BIM image;
the weighting module is used for segmenting the indoor image, acquiring a weight matrix of the indoor image according to a segmentation result, and weighting the feature matrix of the indoor image and the feature matrix of the BIM image by using the weight matrix of the indoor image;
the positioning module is used for matching the indoor image with each BIM image according to the weighting result of the characteristic matrix of the indoor image and each BIM image, and taking the position coordinate corresponding to the BIM image which is most matched with the indoor image as the position coordinate of the user; and the BIM image and the position coordinate are stored in a pre-associated mode.
According to a third aspect of the embodiments of the present invention, there is also provided an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor calls the program instruction to execute the image-based indoor positioning method provided in any one of the various possible implementations of the first aspect.
According to a fourth aspect of embodiments of the present invention, there is also provided a non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the image-based indoor positioning method provided in any one of the various possible implementations of the first aspect.
The embodiment of the invention provides an image-based indoor positioning method and device, wherein the method uses a convolution calculation model to extract the characteristics of a real indoor image and a BIM image shot by a user, so that the problem of cross-domain image matching of the real indoor image and the BIM image is solved; the method comprises the steps of obtaining a weight matrix by segmenting an indoor image, weighting the characteristic matrices of the indoor image and the BIM image by using the weight matrix, realizing concentrated distribution of attention, finally matching the indoor image and the BIM image according to a weighting result, and taking a position coordinate corresponding to the BIM image which is most matched with the indoor image as a position coordinate of a user, so that automatic indoor positioning based on the image is realized, and the positioning precision is high.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
Fig. 1 is a schematic overall flow chart of an image-based indoor positioning method according to an embodiment of the present invention;
fig. 2 is a schematic diagram of an indoor image actually photographed in the image-based indoor positioning method according to the embodiment of the present invention;
fig. 3 is a schematic diagram of a BIM image in the image-based indoor positioning method according to the embodiment of the present invention;
fig. 4 is a schematic diagram of a sliding window capture subgraph in the image-based indoor positioning method provided by the embodiment of the invention;
fig. 5 is a schematic diagram illustrating a weight matrix acquisition in the image-based indoor positioning method according to the embodiment of the present invention;
fig. 6 is a schematic view illustrating a complete flow of an image-based indoor positioning method according to another embodiment of the present invention;
fig. 7 is a schematic view of an overall structure of an image-based indoor positioning apparatus according to an embodiment of the present invention;
fig. 8 is a schematic view of an overall structure of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In an embodiment of the present invention, an image-based indoor positioning method is provided, and fig. 1 is a schematic overall flow chart of the image-based indoor positioning method provided in the embodiment of the present invention, where the method includes: s101, acquiring an indoor image of a Building shot by a user, extracting features of the indoor image and a Building Information Model (BIM) image of the Building acquired in advance from a BIM gallery based on a convolution calculation model, and acquiring a feature matrix of the indoor image and a feature matrix of the BIM image;
the indoor image is a picture of an indoor scene, which is shot by a user on a road network located indoors in a building by using a camera device such as a mobile phone. And comparing the indoor image shot by the user with the BIM image to realize indoor positioning. The present embodiment can acquire the position coordinates of the photo taken by the user, and also can acquire the direction angle of the taken photo and the BIM image that is most matched with the taken photo by inputting the taken indoor image and the BIM gallery. In this embodiment, the actually photographed indoor image and the BIM image belong to two images in different fields, and there is a certain difference between the two images, such as texture, structural ratio, and the like, as shown in fig. 2 and 3. Because the difference of the images in different fields can not achieve the expected matching effect by directly applying the traditional image matching method such as the SIFT (Scale Invariant Feature Transform) Feature descriptor, the embodiment extracts the features of the indoor image and the BIM image through the convolution calculation model, reduces the difference of the images in different fields, and performs indoor positioning according to the image features obtained by convolution.
Before feature extraction is performed on the indoor image and the BIM image, the indoor image and the BIM image are read to obtain a matrix I (x, y, z). Where x, y are the width and height of the image, respectively, and z is the R, G and B channels of the image. And (3) adjusting the sizes of the indoor image and the BIM image to be (m, n) by using a bilinear interpolation algorithm.
BIM image acquisition aims at building a BIM image and a matching position coordinate data set by means of a BIM model. Countless acquisition points exist in the BIM indoor model, if infinite acquisition points are acquired, on one hand, acquisition is difficult, on the other hand, the image characteristics of adjacent points are similar, unnecessary resource waste is caused, moreover, too many acquisition points can cause too many BIM images to be detected, and the operation time of position positioning is greatly reduced. How to reasonably set acquisition points of BIM data and convert an acquisition point set in an infinite space into a limited BIM image acquisition point set is a key step of indoor positioning of cooperation of BIM and computer vision.
The method comprises the steps of combining a BIM indoor map model formed by a building, and selecting a BIM image acquisition point set, wherein the position to be positioned is a passable area in the BIM indoor map model generally, so that the spatial position of the passable area can be ignored, and nodes in a topological road network are key positions in an indoor space, such as doors, stairway bends and the like.
S102, segmenting the indoor image, acquiring a weight matrix of the indoor image according to a segmentation result, and weighting the feature matrix of the indoor image and the feature matrix of the BIM image by using the weight matrix of the indoor image;
and segmenting the indoor image into a foreground part and a background part. Foreground portions in indoor images, such as tables and chairs, flowerpots, people, etc., are dynamic and are not present in the BIM model, which are factors that are not conducive to indoor positioning. While the view of indoor positioning is focused on background parts such as ceilings, floors, walls, etc. In order to distinguish different degrees of influence of the foreground part and the background part on indoor positioning, different weights are set for the foreground and the background in the indoor image, so that a weight matrix of the indoor image is obtained, and the feature matrix of the indoor image and the feature matrix of the BIM image are weighted according to the weight matrix.
S103, matching the indoor image with each BIM image according to the weighting result of the characteristic matrix of the indoor image and each BIM image, and taking the position coordinate corresponding to the BIM image which is most matched with the indoor image as the position coordinate of the user; and the BIM image and the position coordinate are stored in a pre-associated mode.
And comparing the weighting result of the feature matrix of the indoor image with the weighting result of the feature matrix of each BIM image, and acquiring the BIM image which is most matched with the indoor image according to the comparison result. And the position coordinates and the direction angle of the BIM image acquisition are taken as the position coordinates and the azimuth angle of indoor image shooting. And taking the position coordinates of BIM image acquisition as the coordinates of a user for shooting the indoor image.
The method has the advantages that the convolution calculation model is used for extracting the characteristics of the real indoor image and the BIM image shot by the user, so that the problem of cross-domain image matching of the real indoor image and the BIM image is solved; the method comprises the steps of obtaining a weight matrix by segmenting an indoor image, weighting the characteristic matrices of the indoor image and the BIM image by using the weight matrix, realizing concentrated distribution of attention, finally matching the indoor image and the BIM image according to a weighting result, and taking a position coordinate corresponding to the BIM image which is most matched with the indoor image as a position coordinate of a user, so that automatic indoor positioning based on the image is realized, and the positioning precision is high.
On the basis of the foregoing embodiment, in this embodiment, the step of extracting features of the indoor image and the BIM image based on a convolution calculation model, and acquiring a feature matrix of the indoor image and a feature matrix of the BIM image includes: scanning the indoor image and the BIM image by using a sliding window to obtain sub-images of the indoor image and the BIM image; performing convolution calculation on the subgraph of the indoor image, the subgraph of the BIM image, the indoor image and the BIM image based on a convolution calculation model; and taking the convolution calculation result of the sub-image of the indoor image and the indoor image as a characteristic matrix of the indoor image, and taking the convolution calculation result of the sub-image of the BIM image and the BIM image as the characteristic matrix of the BIM image.
Specifically, a plurality of sub-images are respectively intercepted from the indoor image and the BIM image by using a sliding window, so that the error of the situation that the visual field width of the indoor photo which is really shot is slightly larger than that of the BIM screenshot is reduced. And performing convolution calculation on the subgraph of the indoor image and the original indoor image to obtain a characteristic matrix of the indoor image. And carrying out convolution calculation on the subgraph of the BIM image and the original BIM image to obtain a feature matrix of the BIM image.
On the basis of the above embodiment, in this embodiment, the step of scanning the indoor image and the BIM image by using the sliding window to obtain the subgraphs of the indoor image and the BIM image includes: aligning the upper left corner of the sliding window with the upper left corners of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image; aligning the upper right corner of the sliding window with the upper right corners of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image; aligning the lower left corner of the sliding window with the lower left corners of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image; aligning the lower right corner of the sliding window with the lower right corners of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image; and aligning the central point of the sliding window with the central points of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image.
Specifically, five sub-images are intercepted by using the sliding window of (m ', n'), wherein the five sub-images respectively comprise four corner points and a central point, and the original image, and six images are obtained. And performing feature extraction on the six indoor images and the six BIM images. As shown in fig. 4, a in fig. 4 is an original image, and the hatched portion represents the feature extraction performed on the whole original image. B, c, d, e and f in fig. 4 are respectively expressed as subgraphs intercepted by aligning four corner points and a central point, and the shaded part represents the subgraph intercepted by a sliding window, and the characteristics of the subgraphs are extracted.
On the basis of the above embodiments, the convolution calculation model in this embodiment is a VGG16 model; correspondingly, the step of extracting the features of the indoor image and the BIM image based on the convolution calculation model to obtain the feature matrix of the indoor image and the feature matrix of the BIM image comprises the following steps: taking the indoor image as an input of the VGG16 model, and taking an output of maxpool in a fourth layer of the VGG16 model as a feature matrix of the indoor image; and taking the BIM image as an input of the VGG16 model, and taking an output of maxpool in a fourth layer of the VGG16 model as a feature matrix of the BIM image.
Specifically, the convolutional network structure of the VGG16 model is shown in Table 1, where conv3 denotes that the convolution kernel is a matrix of 3 × 3, the Spatial compensation (Spatial Padding) of the convolutional layer is one pixel, and the type of maxpool is (2, 2). in the convolution operation value, convolution is applied to the image, and through sparse connection and parameter sharing, the convolutional network is made to have properties to translation and the like while the number of parameters is greatly reducedThe same way is changed, we say that it is equivalent. If the functions f (x) and g (x) satisfy f [ g (x)]＝g[f(x)]Then f (x) has an isodenaturing property for transform g. For convolution, if let g be an arbitrary translation function of the input, then the convolution function is equally degenerate for g. Pooling uses the overall statistical characteristics of neighboring outputs at a location instead of the output of the network at that location. Assume the input of the pooling layer is aijThen output Amax＝max(aij). The maximum pooling makes the convolutional network locally shift invariant to the image. Since the accuracy of the fourth layer is the highest, the embodiment obtains the output of the maxpool of the fourth layer to obtain the feature matrix O (i; j; 512) of the image, and outputs 512 feature maps.
TABLE 1 convolutional network architecture of VGG16
The embodiment trains the VGG16 model using ImageNet before feature extraction using the VGG16 model. Preprocessing of ImageNet images is required prior to training. The pixel values of all ImageNet images are counted to obtain the mean values (r, g, b) of the three channels of RGB of all ImageNet images, namely
Wherein p is the value of the pixel point, i and j represent the coordinates of the pixel point in the width and height directions of the image, R represents the R channel of the image, G represents the G channel of the image, B represents the B channel of the image, and n represents the total number of the images of all training sets in ImageNet.
R, g and b are subtracted from pixels of three channels of RGB of each image respectively, preprocessing of the image is carried out, so that the influence of illumination or photo brightness on feature extraction is reduced, extracted features are emphasized on natural characteristics of the image, and an extracted feature matrix I (m ', n', 3) is obtained. Inputting I (m ', n', 3) into a VGG16 model, wherein a training set is ImageNet, and performing convolution calculation as follows:
the above formula is applied to the convolution calculation of I (m ', n', 3). Wherein V is the tensor of the input image; k is a convolution kernel; i, j are the height and width of the image; k is the number of channels and s is the step length.
On the basis of the foregoing embodiments, in this embodiment, the step of segmenting the indoor image and acquiring the weight matrix of the indoor image according to the segmentation result includes: performing semantic segmentation on the indoor image based on a DeepLabv3+ model to obtain a segmentation result of the indoor image; dividing the indoor image into grids, if the segmentation result of the indoor image in any grid comprises a foreground, setting the weight of the grid as a first preset weight, and otherwise, setting the weight of the grid as a second preset weight; wherein the first preset weight is smaller than the second preset weight; and constructing a weight matrix of the indoor image according to the weights of all the grids.
Specifically, the DeepLabv3+ model was used for semantic segmentation. The DeepLabv3+ model emphasizes foreground rather than background in the training process on the training set. Based on the segmentation result, the interval of the grid is set to (l)m,ln) And carrying out grid division on the indoor image. If there is a semantically segmented part, i.e., foreground, in the mesh, the mesh weight is reduced to w (w)<1) And the weights of the rest grids are 1 to obtain a weight matrix W. In this case, the first predetermined potential is w and the second predetermined weight is 1. Thereby reducing the influence of dynamic objects and focusing attention on static parts. A in fig. 5 indicates a mesh division result based on the division result, white lines are the divided meshes, black parts are the background, and white parts are the foreground in b in fig. 5. Drawing (A)In c of 5, the black part is an area with the weight of the second preset weight, and the white part is an area with the weight of the first preset weight.
In addition, the weight matrix of each sub-image of the indoor image can be acquired by adopting the method. When weighting is carried out, the feature matrix of the indoor image and the feature matrix of the BIM image are weighted by using the weight matrix of the indoor image, and the feature matrix of each subgraph of the indoor image and the feature matrix of each subgraph of the BIM image are weighted by using the weight matrix of each subgraph of the indoor image.
On the basis of the foregoing embodiments, in this embodiment, the step of weighting the feature matrix of the indoor image and the feature matrix of the BIM image by using the weight matrix of the indoor image includes: adjusting the size of the weight matrix of the indoor image so that the size of the weight matrix of the indoor image is the same as the size of the feature matrix of the indoor image; and performing Hadamard multiplication on the adjusted weight matrix of the indoor image and the feature matrices of the indoor image and the BIM image respectively to obtain the weighting results of the feature matrices of the indoor image and each BIM image.
Specifically, the weight matrix is scaled to be the same as the feature matrix, resulting in W'. Hadamard products are carried out on the feature matrix and the weight matrix to obtain a feature matrix O' (i, j,512), so that the attention of the indoor image and the BIM image is adjusted.
On the basis of the foregoing embodiments, in this embodiment, the step of matching the indoor image with each of the BIM images according to the weighting result of the feature matrices of the indoor image and each of the BIM images, and taking the position coordinate corresponding to the BIM image that is most matched with the indoor image as the position coordinate of the user includes: expanding the weighting result of the feature matrix of the indoor image to obtain the feature vector of the indoor image, and expanding the weighting result of the feature matrix of each BIM image to obtain the feature vector of each BIM image; calculating cosine similarity between the feature vectors of the indoor images and the feature vectors of the BIM images, and taking the BIM image corresponding to the maximum cosine similarity as the BIM image which is most matched with the indoor images; and taking the position coordinate corresponding to the BIM image which is most matched with the indoor image as the position coordinate of the user.
Specifically, the feature matrix O' (I, j,512) is expanded into feature vectors. And the feature vectors of all BIM images in the BIM image library form a vector library. The method comprises the steps of calculating cosine similarity between a feature vector of an indoor image and each feature vector in a vector library, selecting the feature vector in the vector library corresponding to the maximum cosine similarity, taking a BIM image to which the selected feature vector belongs as a BIM image which is most matched with the indoor image, and taking a position coordinate associated with the BIM image as a position coordinate for shooting the indoor image, so that the position of a user is positioned. A complete flow chart of the image-based indoor positioning method is shown in fig. 6.
In another embodiment of the present invention, an image-based indoor positioning apparatus is provided, which is used for implementing the method in the foregoing embodiments. Therefore, the description and definition in the embodiments of the image-based indoor positioning method described above may be used for understanding the respective execution modules in the embodiments of the present invention. Fig. 7 is a schematic diagram of an overall structure of an image-based indoor positioning apparatus according to an embodiment of the present invention, where the apparatus includes an extraction module 701, a weighting module 702, and a positioning module 703, where:
the extraction module 701 is configured to acquire an indoor image of a building, which is shot by a user, perform feature extraction on the indoor image and a Building Information Model (BIM) image of the building, which is acquired in advance in a BIM gallery, based on a convolution calculation model, and acquire a feature matrix of the indoor image and a feature matrix of the BIM image;
the weighting module 702 is configured to segment the indoor image, obtain a weight matrix of the indoor image according to a segmentation result, and weight the feature matrix of the indoor image and the feature matrix of the BIM image respectively by using the weight matrix of the indoor image;
the positioning module 703 is configured to match the indoor image with each of the BIM images according to a weighting result of the feature matrices of the indoor image and each of the BIM images, and use a position coordinate corresponding to the BIM image that is most matched with the indoor image as a position coordinate of the user; and the BIM image and the position coordinate are stored in a pre-associated mode.
The method has the advantages that the convolution calculation model is used for extracting the characteristics of the real indoor image and the BIM image shot by the user, so that the problem of cross-domain image matching of the real indoor image and the BIM image is solved; the method comprises the steps of obtaining a weight matrix by segmenting an indoor image, weighting the characteristic matrices of the indoor image and the BIM image by using the weight matrix, realizing concentrated distribution of attention, finally matching the indoor image and the BIM image according to a weighting result, and taking a position coordinate corresponding to the BIM image which is most matched with the indoor image as a position coordinate of a user, so that automatic indoor positioning based on the image is realized, and the positioning precision is high.
On the basis of the foregoing embodiment, the extraction module in this embodiment is specifically configured to: scanning the indoor image and the BIM image by using a sliding window to obtain sub-images of the indoor image and the BIM image; performing convolution calculation on the subgraph of the indoor image, the subgraph of the BIM image, the indoor image and the BIM image based on a convolution calculation model; and taking the convolution calculation result of the sub-image of the indoor image and the indoor image as a characteristic matrix of the indoor image, and taking the convolution calculation result of the sub-image of the BIM image and the BIM image as the characteristic matrix of the BIM image.
On the basis of the foregoing embodiment, the extraction module in this embodiment is further configured to: aligning the upper left corner of the sliding window with the upper left corners of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image; aligning the upper right corner of the sliding window with the upper right corners of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image; aligning the lower left corner of the sliding window with the lower left corners of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image; aligning the lower right corner of the sliding window with the lower right corners of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image; and aligning the central point of the sliding window with the central points of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image.
On the basis of the above embodiment, the convolution calculation model in this embodiment is a VGG16 model; correspondingly, the extraction module is specifically configured to: taking the indoor image as an input of the VGG16 model, and taking an output of maxpool in a fourth layer of the VGG16 model as a feature matrix of the indoor image; and taking the BIM image as an input of the VGG16 model, and taking an output of maxpool in a fourth layer of the VGG16 model as a feature matrix of the BIM image.
On the basis of the above embodiments, the weighting module in this embodiment is specifically configured to: performing semantic segmentation on the indoor image based on a DeepLabv3+ model to obtain a segmentation result of the indoor image; dividing the indoor image into grids, if the segmentation result of the indoor image in any grid comprises a foreground, setting the weight of the grid as a first preset weight, and otherwise, setting the weight of the grid as a second preset weight; wherein the first preset weight is smaller than the second preset weight; and constructing a weight matrix of the indoor image according to the weights of all the grids.
On the basis of the above embodiments, the weighting module in this embodiment is specifically configured to: adjusting the size of the weight matrix of the indoor image so that the size of the weight matrix of the indoor image is the same as the size of the feature matrix of the indoor image; and performing Hadamard multiplication on the adjusted weight matrix of the indoor image and the feature matrices of the indoor image and the BIM image respectively to obtain the weighting results of the feature matrices of the indoor image and each BIM image.
On the basis of the foregoing embodiments, the positioning module in this embodiment is specifically configured to: expanding the weighting result of the feature matrix of the indoor image to obtain the feature vector of the indoor image, and expanding the weighting result of the feature matrix of each BIM image to obtain the feature vector of each BIM image; calculating cosine similarity between the feature vectors of the indoor images and the feature vectors of the BIM images, and taking the BIM image corresponding to the maximum cosine similarity as the BIM image which is most matched with the indoor images; and taking the position coordinate corresponding to the BIM image which is most matched with the indoor image as the position coordinate of the user.
Fig. 8 illustrates a physical structure diagram of an electronic device, and as shown in fig. 8, the electronic device may include: a processor (processor)801, a communication Interface (Communications Interface)802, a memory (memory)803 and a communication bus 804, wherein the processor 801, the communication Interface 802 and the memory 803 complete communication with each other through the communication bus 804. The processor 801 may call logic instructions in the memory 803 to perform the following method: acquiring an indoor image of a building shot by a user, and performing feature extraction on the indoor image and a BIM image of the building in a BIM image library based on a convolution calculation model to acquire a feature matrix of the indoor image and a feature matrix of the BIM image; dividing the indoor image, acquiring a weight matrix of the indoor image according to a division result, and weighting the feature matrices of the indoor image and the BIM image respectively by using the weight matrix of the indoor image; matching the indoor image with each BIM image according to the weighting result of the characteristic matrix of the indoor image and each BIM image, and taking the position coordinate corresponding to the BIM image which is most matched with the indoor image as the position coordinate of the user; wherein, the BIM image and the position coordinate are stored in advance in an associated manner.
In addition, the logic instructions in the memory 803 may be implemented in the form of software functional units and stored in a computer readable storage medium when the logic instructions are sold or used as independent products. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The present embodiments provide a non-transitory computer-readable storage medium storing computer instructions that cause a computer to perform the methods provided by the above method embodiments, for example, including: acquiring an indoor image of a building shot by a user, and performing feature extraction on the indoor image and a BIM image of the building in a BIM image library based on a convolution calculation model to acquire a feature matrix of the indoor image and a feature matrix of the BIM image; dividing the indoor image, acquiring a weight matrix of the indoor image according to a division result, and weighting the feature matrices of the indoor image and the BIM image respectively by using the weight matrix of the indoor image; matching the indoor image with each BIM image according to the weighting result of the characteristic matrix of the indoor image and each BIM image, and taking the position coordinate corresponding to the BIM image which is most matched with the indoor image as the position coordinate of the user; wherein, the BIM image and the position coordinate are stored in advance in an associated manner.
Those of ordinary skill in the art will understand that: all or part of the steps for implementing the method embodiments may be implemented by hardware related to program instructions, and the program may be stored in a computer readable storage medium, and when executed, the program performs the steps including the method embodiments; and the aforementioned storage medium includes: various media that can store program codes, such as ROM, RAM, magnetic or optical disks.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. An image-based indoor positioning method, comprising:
acquiring an indoor image of a building shot by a user, and performing feature extraction on the indoor image and a BIM image of the building acquired in advance in a BIM image library based on a convolution calculation model to acquire a feature matrix of the indoor image and a feature matrix of the BIM image;
dividing the indoor image, acquiring a weight matrix of the indoor image according to a division result, and weighting the feature matrix of the indoor image and the feature matrix of the BIM image by using the weight matrix of the indoor image;
matching the indoor image with each BIM image according to the weighting result of the characteristic matrix of the indoor image and each BIM image, and taking the position coordinate corresponding to the BIM image which is most matched with the indoor image as the position coordinate of the user; and the BIM image and the position coordinate are stored in a pre-associated mode.
2. The image-based indoor positioning method according to claim 1, wherein the step of performing feature extraction on the indoor image and the BIM image based on a convolution calculation model to obtain the feature matrix of the indoor image and the feature matrix of the BIM image comprises:
scanning the indoor image and the BIM image by using a sliding window to obtain sub-images of the indoor image and the BIM image;
performing convolution calculation on the subgraph of the indoor image, the subgraph of the BIM image, the indoor image and the BIM image based on a convolution calculation model;
and taking the convolution calculation result of the sub-image of the indoor image and the indoor image as a characteristic matrix of the indoor image, and taking the convolution calculation result of the sub-image of the BIM image and the BIM image as the characteristic matrix of the BIM image.
3. The image-based indoor positioning method of claim 1, wherein the indoor image and the BIM image are scanned by using a sliding window, and the step of acquiring the subgraphs of the indoor image and the BIM image comprises:
aligning the upper left corner of the sliding window with the upper left corners of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image;
aligning the upper right corner of the sliding window with the upper right corners of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image;
aligning the lower left corner of the sliding window with the lower left corners of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image;
aligning the lower right corner of the sliding window with the lower right corners of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image;
and aligning the central point of the sliding window with the central points of the indoor image and the BIM image, and intercepting sub-images of the indoor image and the BIM image.
4. The image-based indoor positioning method of claim 1, wherein the convolution calculation model is a VGG16 model;
correspondingly, the step of extracting the features of the indoor image and the BIM image based on the convolution calculation model to obtain the feature matrix of the indoor image and the feature matrix of the BIM image comprises the following steps:
taking the indoor image as an input of the VGG16 model, and taking an output of maxpool in a fourth layer of the VGG16 model as a feature matrix of the indoor image;
and taking the BIM image as an input of the VGG16 model, and taking an output of maxpool in a fourth layer of the VGG16 model as a feature matrix of the BIM image.
5. The image-based indoor positioning method of any one of claims 1 to 4, wherein the step of segmenting the indoor image and obtaining the weight matrix of the indoor image according to the segmentation result comprises:
performing semantic segmentation on the indoor image based on a DeepLabv3+ model to obtain a segmentation result of the indoor image;
dividing the indoor image into grids, if the segmentation result of the indoor image in any grid comprises a foreground, setting the weight of the grid as a first preset weight, and otherwise, setting the weight of the grid as a second preset weight; wherein the first preset weight is smaller than the second preset weight;
and constructing a weight matrix of the indoor image according to the weights of all the grids.
6. The image-based indoor positioning method of any one of claims 1 to 4, wherein the step of weighting the feature matrix of the indoor image and the feature matrix of the BIM image respectively using the weight matrix of the indoor image comprises:
adjusting the size of the weight matrix of the indoor image so that the size of the weight matrix of the indoor image is the same as the size of the feature matrix of the indoor image;
and performing Hadamard multiplication on the adjusted weight matrix of the indoor image and the feature matrices of the indoor image and the BIM image respectively to obtain the weighting results of the feature matrices of the indoor image and each BIM image.
7. The image-based indoor positioning method according to any one of claims 1 to 4, wherein the step of matching the indoor image with each of the BIM images according to the weighting result of the feature matrices of the indoor image and each of the BIM images, and taking the position coordinate corresponding to the BIM image that is most matched with the indoor image as the position coordinate of the user comprises:
expanding the weighting result of the feature matrix of the indoor image to obtain the feature vector of the indoor image, and expanding the weighting result of the feature matrix of each BIM image to obtain the feature vector of each BIM image;
calculating cosine similarity between the feature vectors of the indoor images and the feature vectors of the BIM images, and taking the BIM image corresponding to the maximum cosine similarity as the BIM image which is most matched with the indoor images;
and taking the position coordinate corresponding to the BIM image which is most matched with the indoor image as the position coordinate of the user.
8. An image-based indoor positioning apparatus, comprising:
the extraction module is used for acquiring an indoor image of a building shot by a user, extracting features of the indoor image and a BIM image of the building acquired in advance in a BIM image library based on a convolution calculation model, and acquiring a feature matrix of the indoor image and a feature matrix of the BIM image;
the weighting module is used for segmenting the indoor image, acquiring a weight matrix of the indoor image according to a segmentation result, and weighting the feature matrix of the indoor image and the feature matrix of the BIM image by using the weight matrix of the indoor image;
the positioning module is used for matching the indoor image with each BIM image according to the weighting result of the characteristic matrix of the indoor image and each BIM image, and taking the position coordinate corresponding to the BIM image which is most matched with the indoor image as the position coordinate of the user; and the BIM image and the position coordinate are stored in a pre-associated mode.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor, when executing the program, carries out the steps of the image based indoor positioning method as claimed in any one of claims 1 to 7.
10. A non-transitory computer readable storage medium, having stored thereon a computer program, which, when being executed by a processor, carries out the steps of the image based indoor positioning method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010104620.5A CN111340882B (en) | 2020-02-20 | 2020-02-20 | Indoor positioning method and device based on image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010104620.5A CN111340882B (en) | 2020-02-20 | 2020-02-20 | Indoor positioning method and device based on image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111340882A true CN111340882A (en) | 2020-06-26 |
| CN111340882B CN111340882B (en) | 2024-02-20 |
Family
ID=71181649
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010104620.5A Active CN111340882B (en) | 2020-02-20 | 2020-02-20 | Indoor positioning method and device based on image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111340882B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112146660A (en) * | 2020-09-25 | 2020-12-29 | 电子科技大学 | Indoor map positioning method based on dynamic word vector |
| CN112182279A (en) * | 2020-12-03 | 2021-01-05 | 武大吉奥信息技术有限公司 | Indoor self-positioning method and equipment based on discrete grid and image matching |
| CN114627426A (en) * | 2022-05-13 | 2022-06-14 | 盈嘉互联（北京）科技有限公司 | Image detection method and device fusing building digital twin semantic graph |
| CN117058473A (en) * | 2023-10-12 | 2023-11-14 | 深圳易行机器人有限公司 | Warehouse material management method and system based on image recognition |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107133913A (en) * | 2017-05-08 | 2017-09-05 | 重庆邮电大学 | A kind of image split-joint method of automatic flattening |
| CN108229468A (en) * | 2017-06-28 | 2018-06-29 | 北京市商汤科技开发有限公司 | Vehicle appearance feature recognition and vehicle retrieval method, apparatus, storage medium, electronic equipment |
| CN109726746A (en) * | 2018-12-20 | 2019-05-07 | 浙江大华技术股份有限公司 | A kind of method and device of template matching |
| CN109784291A (en) * | 2019-01-23 | 2019-05-21 | 电子科技大学 | Pedestrian detection method based on multiple dimensioned convolution feature |
| KR20200012379A (en) * | 2018-07-27 | 2020-02-05 | 연세대학교 산학협력단 | Image-based indoor position detection apparatus and detection method |
-
2020
- 2020-02-20 CN CN202010104620.5A patent/CN111340882B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107133913A (en) * | 2017-05-08 | 2017-09-05 | 重庆邮电大学 | A kind of image split-joint method of automatic flattening |
| CN108229468A (en) * | 2017-06-28 | 2018-06-29 | 北京市商汤科技开发有限公司 | Vehicle appearance feature recognition and vehicle retrieval method, apparatus, storage medium, electronic equipment |
| KR20200012379A (en) * | 2018-07-27 | 2020-02-05 | 연세대학교 산학협력단 | Image-based indoor position detection apparatus and detection method |
| CN109726746A (en) * | 2018-12-20 | 2019-05-07 | 浙江大华技术股份有限公司 | A kind of method and device of template matching |
| CN109784291A (en) * | 2019-01-23 | 2019-05-21 | 电子科技大学 | Pedestrian detection method based on multiple dimensioned convolution feature |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112146660A (en) * | 2020-09-25 | 2020-12-29 | 电子科技大学 | Indoor map positioning method based on dynamic word vector |
| CN112182279A (en) * | 2020-12-03 | 2021-01-05 | 武大吉奥信息技术有限公司 | Indoor self-positioning method and equipment based on discrete grid and image matching |
| CN112182279B (en) * | 2020-12-03 | 2021-03-19 | 武大吉奥信息技术有限公司 | Indoor self-positioning method and equipment based on discrete grid and image matching |
| CN114627426A (en) * | 2022-05-13 | 2022-06-14 | 盈嘉互联（北京）科技有限公司 | Image detection method and device fusing building digital twin semantic graph |
| CN114627426B (en) * | 2022-05-13 | 2022-08-23 | 盈嘉互联（北京）科技有限公司 | Image detection method and device fusing building digital twin semantic graph |
| CN117058473A (en) * | 2023-10-12 | 2023-11-14 | 深圳易行机器人有限公司 | Warehouse material management method and system based on image recognition |
| CN117058473B (en) * | 2023-10-12 | 2024-01-16 | 深圳易行机器人有限公司 | Warehouse material management method and system based on image recognition |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111340882B (en) | 2024-02-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111340882B (en) | Indoor positioning method and device based on image | |
| US10930005B1 (en) | Profile matching of buildings and urban structures | |
| CA2826534C (en) | Backfilling points in a point cloud | |
| CN107004297B (en) | Three-dimensional automatic stereo modeling method and program based on two-dimensional plane diagram | |
| CN108121931B (en) | Two-dimensional code data processing method and device and mobile terminal | |
| CN112700552A (en) | Three-dimensional object detection method, three-dimensional object detection device, electronic apparatus, and medium | |
| CN105303514A (en) | Image processing method and apparatus | |
| CN111291768A (en) | Image feature matching method and device, equipment and storage medium | |
| CN112560619B (en) | Multi-focus image fusion-based multi-distance bird accurate identification method | |
| US10438405B2 (en) | Detection of planar surfaces for use in scene modeling of a captured scene | |
| CN111260687B (en) | Aerial video target tracking method based on semantic perception network and related filtering | |
| CN112634340A (en) | Method, device, equipment and medium for determining BIM (building information modeling) model based on point cloud data | |
| CN113724379B (en) | Three-dimensional reconstruction method and device for fusing image and laser point cloud | |
| CN112085031A (en) | Target detection method and system | |
| CN109766896B (en) | Similarity measurement method, device, equipment and storage medium | |
| CN108629742B (en) | True ortho image shadow detection and compensation method, device and storage medium | |
| CN111353325A (en) | Key point detection model training method and device | |
| CN111161138B (en) | Target detection method, device, equipment and medium for two-dimensional panoramic image | |
| CN117671031A (en) | Binocular camera calibration method, device, equipment and storage medium | |
| CN115393538A (en) | Visual SLAM method and system for indoor dynamic scene based on deep learning | |
| CN116579920A (en) | Image stitching method and system based on heterogeneous multimode panoramic stereoscopic imaging system | |
| CN116091871B (en) | Physical countermeasure sample generation method and device for target detection model | |
| CN111091117B (en) | Target detection method, device, equipment and medium for two-dimensional panoramic image | |
| CN113240656A (en) | Visual positioning method and related device and equipment | |
| Uskenbayeva et al. | Contour analysis of external images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20240402 Address after: 1308, 13th Floor, Shougang Sports Building, No. 6 Jinyuanzhuang Road, Shijingshan District, Beijing, 100043 Patentee after: BIM WINNER (BEIJING) TECHNOLOGY CO.,LTD. Country or region after: China Address before: 100041 1008, building 16, xishanhui, Shijingshan Park, Zhongguancun Science Park, Haidian District, Beijing Patentee before: BIM WINNER (BEIJING) TECHNOLOGY CO.,LTD. Country or region before: China Patentee before: BIM WINNER (SHANGHAI) TECHNOLOGY Co.,Ltd. Patentee before: JIAXING WUZHEN YINGJIA QIANZHEN TECHNOLOGY Co.,Ltd. Patentee before: SHENZHEN QIANHAI YINGJIA DATA SERVICE Co.,Ltd. Patentee before: SHENZHEN BIM WINNER TECHNOLOGY Co.,Ltd. |