A method and system for adjusting navigation information of a navigation robot

Abstract

The present invention discloses a method and system for adjusting navigation information for a navigation robot, wherein the method comprises: generating a navigation path based on an input tour location combined with the location information of the navigation robot; when the navigation robot moves to the corresponding tour location, dividing the obtained navigation area image into a plurality of image blocks; extracting features from each image block to obtain image block features, and encoding the features to obtain context information features; determining the corresponding navigation scene based on the context information features; comparing the similarities and differences between the scene label text and each navigation scene information, obtaining the target navigation scene information for display; the navigation robot obtains the operation instruction input by the user, obtains the corresponding operation information based on the operation instruction, and adjusts the navigation scene information based on the operation information. The present invention improves the reliability of the navigation service of the navigation robot, thereby being able to provide users with better navigation services.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a navigation information adjustment method and system for a navigation robot. Background Art

[0002] In the past, people usually use manual guides to enhance their travel experience during travel, but manual guides are extremely dependent on the quality of relevant personnel and will invest too high a labor cost. In order to solve the above problems, various companies choose to use guide robots to gradually replace manual guides. The use of guide robots can not only do what manual guides can do, but also bring people a more novel guide experience. However, there are still some problems with the current use of guide robots for guiding. Usually, the scene recognition processing of the guide robot is to extract features and recognize scenes for the entire image, but the operation on the entire image will make the extracted image features not have local robustness, which will have a greater impact on the determination of the guide scene and reduce the accuracy of the guide scene recognition. At the same time, in the acquisition of the guide scene information, the final guide scene information is usually determined only by the similarity comparison of text information, but only by similarity comparison, its accuracy will not reach the ideal effect, which may cause the guide scene and the guide scene information to not match, resulting in a decrease in the quality of the guide service. Summary of the invention

[0003] The purpose of the present invention is to overcome the deficiencies of the prior art. The present invention provides a method and system for adjusting navigation information of a navigation robot, which improves the reliability of the navigation service of the navigation robot, thereby being able to provide users with better navigation services.

[0004] In order to solve the above technical problems, the present invention provides a method for adjusting navigation information for a navigation robot, the method comprising:

[0005] The guide robot obtains a number of tourist spots input by the user, and generates a guide route based on the number of tourist spots combined with the position information of the guide robot;

[0006] The tour guide robot moves to each tourist spot according to the tour guide path. When the tour guide robot moves to the corresponding tourist spot, it acquires a tour area image based on preset image acquisition points, and divides the tour area image into a plurality of image blocks;

[0007] Extracting features from each image block to obtain image block features, and encoding the image block features to obtain context information features;

[0008] Determine a corresponding navigation scene in a scene database based on the context information feature;

[0009] Based on the corresponding navigation scene, a corresponding scene label text is obtained, and the similarity and difference between the scene label text and each navigation scene information in the information database are compared, and based on the similarity comparison result and the difference comparison result, the target navigation scene information is obtained, and the navigation robot displays the target navigation scene information;

[0010] When the navigation robot displays the target navigation scene information, the navigation robot obtains an operation instruction input by a user, obtains corresponding operation information based on the operation instruction, and adjusts the navigation scene information based on the operation information.

[0011] Optionally, generating a tour route based on a plurality of tour locations combined with the location information of the tour guide robot includes:

[0012] Acquire corresponding location information based on the positioning device of the tour guide robot, and acquire the recommended tour time and navigation data corresponding to each tour location;

[0013] A navigation path of the navigation robot is constructed according to the least time-consuming navigation route based on the location information, navigation data and recommended tour time.

[0014] Optionally, acquiring the navigation area image based on preset image acquisition points and dividing the navigation area image into a plurality of image blocks includes:

[0015] The navigation robot uses a shooting device at a preset image acquisition point to acquire an image of the navigation area at a preset angle;

[0016] Preprocessing the navigation area image to obtain a preprocessed navigation area image;

[0017] The preprocessed navigation area image is divided into several image blocks based on a sliding window.

[0018] Optionally, extracting features from each image block to obtain image block features includes:

[0019] Calculating the pixel grayscale value variance of the image block, and constructing a pixel grayscale value matrix based on the pixel grayscale value variance;

[0020] The image block is projected using a feature projection matrix based on the pixel grayscale matrix to obtain image block features.

[0021] Optionally, encoding the image block features to obtain context information features includes:

[0022] Generate sequence features based on the image block features, and obtain semantic correlation relationships of the image blocks using a multi-layer perceptron based on the sequence features;

[0023] Based on the long short-term memory network, the semantic correlation of image blocks is used to encode the context information of image blocks to obtain context information features.

[0024] Optionally, determining a corresponding navigation scene in a scene database based on the context information feature includes:

[0025] Calculating the similarity between the context information feature and each scene feature in the scene database, and selecting the target scene feature with the highest similarity;

[0026] A scene identifier is acquired based on the target scene feature, and a corresponding navigation scene is determined based on the scene identifier.

[0027] Optionally, comparing the scene label text with the navigation scene information in the information database for similarities and differences includes:

[0028] Obtain a label text string corresponding to the scene label text, and obtain a plurality of information text strings corresponding to each guide scene information;

[0029] Constructing a joint probability distribution between the label text string and each information text string;

[0030] Sampling the joint probability distribution to obtain a sampled string;

[0031] Calculating a first distance matrix between the label text string and the sampled string, and calculating a second distance matrix between each information text string and the sampled string;

[0032] Determine the similarity between the scene label text and each navigation scene information in the information database based on the first distance matrix and the second distance matrix;

[0033] Obtaining a difference string between the label text string and the information text string based on a hash function;

[0034] The edit distance of the difference character string is calculated based on the edit distance algorithm to obtain the corresponding edit distance, and the difference between the scene label text and each navigation scene information in the information database is determined based on the corresponding edit distance.

[0035] Optionally, determining the similarity between the scene label text and each navigation scene information in the information database based on the first distance matrix and the second distance matrix includes:

[0036] Performing feature extraction on the first distance matrix to obtain a first column representation vector, a first diagonal representation vector, and a first row representation vector;

[0037] Performing feature extraction on the second distance matrix to obtain a second column representation vector, a second diagonal representation vector, and a second row representation vector;

[0038] The vector similarity between the first column representation vector, the first diagonal representation vector and the first row representation vector and the second column representation vector, the second diagonal representation vector and the second row representation vector is calculated.

[0039] Optionally, acquiring corresponding operation information based on the operation instruction, and adjusting the navigation scene information based on the operation information includes:

[0040] Identify the statement type corresponding to the operation instruction, and determine the target operation type based on the corresponding statement type;

[0041] Acquire operation attribute information corresponding to the target operation type, and acquire corresponding operation information based on the target operation type and the operation attribute information;

[0042] The navigation robot adjusts the navigation scene information based on the corresponding operation information.

[0043] In addition, the present invention also provides a navigation information adjustment system for a navigation robot, the system comprising:

[0044] A navigation path generation module: used for obtaining a number of tourist attractions input by a user based on the navigation robot, and generating a navigation path based on the number of tourist attractions combined with the location information of the navigation robot;

[0045] A navigation area image division module: used for the navigation robot to move to each tourist spot according to the navigation path. When the navigation robot moves to the corresponding tourist spot, the navigation area image is acquired based on the preset image acquisition points, and the navigation area image is divided into a plurality of image blocks;

[0046] Feature acquisition module: used to extract features from each image block to obtain image block features, and encode the image block features to obtain context information features;

[0047] A navigation scene determination module: used to determine a corresponding navigation scene in a scene database based on the context information feature;

[0048] A navigation scene information acquisition module is used to acquire a corresponding scene label text based on a corresponding navigation scene, and compare the similarity and difference between the scene label text and each navigation scene information in the information database, and acquire target navigation scene information based on the similarity comparison result and the difference comparison result, and the navigation robot displays the target navigation scene information;

[0049] Navigation scene information adjustment module: used for the navigation robot to obtain the operation instruction input by the user during the process of displaying the target navigation scene information, obtain the corresponding operation information based on the operation instruction, and adjust the navigation scene information based on the operation information.

[0050] In an embodiment of the present invention, the navigation area image is divided into a number of image blocks, and feature extraction is performed using a feature projection matrix through the variance of the grayscale values ​​of the pixels of the image blocks, thereby improving the extraction effect of image features and avoiding the misclassification of similar scenes due to the unsegmented overall image. In obtaining the navigation scene information, the scene label text is compared with the navigation scene information in the information database for similarity and difference, and the final navigation scene information is determined from the two dimensions of similarity and difference, which not only improves the accuracy of matching the navigation scene information with the navigation scene, but also improves the calculation efficiency of the similarity by processing the distance matrix generated by the character string, thereby improving the reliability of the navigation service of the navigation robot, thereby providing users with a better navigation service. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0052] Figure 1 is a flow chart of a method for adjusting navigation information of a navigation robot in an embodiment of the present invention;

[0053] Figure 2 It is a schematic diagram of the structural composition of a navigation information adjustment system for a navigation robot in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] Embodiment 1

[0056] See also Figure 1 , Figure 1 It is a flow chart of a method for adjusting navigation information of a navigation robot in an embodiment of the present invention.

[0057] like Figure 1 As shown, a method for adjusting navigation information for a navigation robot, the method comprising:

[0058] S11: acquiring, by the tour guide robot, a number of tourist attractions input by a user, and generating a tour guide route based on the number of tourist attractions combined with location information of the tour guide robot;

[0059] In the specific implementation process of the present invention, the guide path is generated based on several tourist locations combined with the location information of the guide robot, including: obtaining the corresponding location information based on the positioning device of the guide robot, and obtaining the recommended tour time and navigation data corresponding to each tourist location; based on the location information, navigation data and recommended tour time, the guide path of the guide robot is constructed according to the least time-consuming guide route.

[0060] Specifically, the tour guide robot obtains several tourist spots input by the user, obtains corresponding position information based on the positioning device of the tour guide robot, and obtains recommended tour time and navigation data corresponding to each tourist spot. The tour guide robot can obtain the corresponding recommended tour time and navigation data by querying the server; based on the position information, navigation data and recommended tour time, the tour guide robot constructs a tour path according to the least time-consuming tour route, with the least time-consuming tour route as a constraint condition, the position of the tour guide robot as the starting point, the time required between the tourist spots is obtained through the navigation data and the recommended tour time, and the path with the least time-consuming is selected to construct the least time-consuming tour path.

[0061] S12: the tour guide robot moves to each tourist spot according to the tour guide path. When the tour guide robot moves to the corresponding tourist spot, an image of the tour guide area is acquired based on preset image acquisition points, and the image of the tour guide area is divided into a plurality of image blocks.

[0062] In the specific implementation process of the present invention, the navigation area image is obtained based on preset image acquisition points, and the navigation area image is divided into a number of image blocks, including: the navigation robot obtains the navigation area image at the preset image acquisition point using a shooting device at a preset angle; preprocessing the navigation area image to obtain a preprocessed navigation area image; and dividing the preprocessed navigation area image into a number of image blocks based on a sliding window.

[0063] Specifically, the tour guide robot moves to each tourist spot according to the tour guide path. When the tour guide robot moves to the corresponding tourist spot, the tour guide robot uses its built-in shooting device to obtain the tour area image according to the preset angle at the preset image acquisition point; the tour guide area image is pre-processed, the total number of pixels of the tour guide area image is calculated, the frequency of each pixel in the tour guide area image is counted, the frequency of each pixel and the total number of pixels are calculated, and the pixel distribution value is obtained. A corresponding histogram is established according to the pixel distribution value, and the corresponding histogram is linearly stretched to make the image more uniform in brightness and contrast, so as to obtain The navigation area image after brightness transformation is obtained, the navigation area image after brightness transformation is normalized to obtain a normalized navigation area image, the normalized navigation area image is enhanced, and after the image enhancement is completed, a pre-processed navigation area image is obtained; based on a sliding window, the pre-processed navigation area image is divided into a plurality of image blocks, and sliding window parameters are determined, wherein the sliding window parameters are the size and scaling ratio of the sliding window, and the positions of the sliding start point and the end point, and an expected sliding window is constructed according to the sliding window parameters, and the pre-processed navigation area image is segmented based on the expected sliding window to obtain a plurality of image blocks.

[0064] S13: extracting features from each image block to obtain image block features, and encoding the image block features to obtain context information features;

[0065] In the specific implementation process of the present invention, the feature extraction of each image block to obtain the image block features includes: calculating the variance of the pixel grayscale values ​​of the image block, and constructing a pixel grayscale value matrix based on the pixel grayscale value variance; projecting the image block using a feature projection matrix based on the pixel grayscale matrix to obtain the image block features.

[0066] Furthermore, the encoding process of the image block features to obtain context information features includes: generating sequence features based on the image block features, and obtaining semantic correlation relationships of the image blocks using a multi-layer perceptron based on the sequence features; and encoding the context information of the image blocks using the semantic correlation relationships of the image blocks based on a long short-term memory network to obtain context information features.

[0067] Specifically, the grayscale value of the pixel of the image block is calculated, and the pixel grayscale value is obtained by weighted average of the pixel values ​​of the red, green and blue channels of the pixel of the image block, and the average value of the pixel grayscale value of all the pixels is calculated, and the image block matrix of the image block is constructed. The pixel grayscale value variance is calculated based on the average value of the pixel grayscale value combined with the number of rows and columns of the image block matrix. The calculation expression of the pixel grayscale value variance is:

[0068]

[0069] Among them, σ is the variance of the pixel gray value, c is the pixel number, m is the number of rows in the image block matrix, n is the number of columns in the image block matrix, j is the number of pixels, and p is the pixel gray value. is the mean of the pixel grayscale values; construct a pixel grayscale value matrix based on the pixel grayscale value variance, calculate the matrix value through the pixel grayscale value variance statistics, obtain the co-occurrence matrix based on the matrix value, calculate the sum of all elements in the co-occurrence matrix based on the co-occurrence matrix, obtain the probability matrix through the sum of all elements in the co-occurrence matrix, extract the grayscale co-occurrence matrix through the probability matrix, and construct the pixel grayscale value matrix according to the grayscale co-occurrence matrix; project the image block based on the pixel grayscale matrix using the feature projection matrix, perform a two-dimensional discrete wavelet transform on the image block, obtain the image block after the two-dimensional discrete wavelet transform, vectorize the pixel grayscale matrix, obtain the vectorized pixel grayscale matrix, calculate the covariance matrix of the vectorized pixel grayscale matrix, perform eigenvalue decomposition on the covariance matrix, obtain the unit eigenvector, construct the feature projection matrix using the matrix transpose symbol and the unit eigenvector, project the image block with the constructed feature projection matrix, and obtain the image block feature; generate sequence features based on the image block feature, and generate sequence features through the cyclic unit module The block extracts sequence information from the image block features to obtain sequence features. The recurrent unit module is a module for extracting feature expressions from the sequences. Based on the sequence features, the semantic correlation of the image blocks is obtained by using a multi-layer perceptron. The probability features of the image blocks are obtained by using the multi-layer perceptron. The sequence features are used to iterate to a preset number of times according to the probability features of the image blocks. The multi-layer perceptron outputs the correlation of the image blocks at the semantic level, that is, the semantic correlation of the image blocks is obtained. The context information of the image blocks is encoded by using the semantic correlation of the image blocks based on the long short-term memory network. The characteristics of the long short-term memory network for processing time series information are used to filter out the features that do not have discriminative performance in the image blocks by using the semantic correlation through the long short-term memory network, encode the image blocks and their context information, and obtain context information features. The feature projection matrix is ​​used to extract features through the grayscale value variance of the pixel points of the image blocks, thereby improving the extraction effect of image features and avoiding the misclassification of similar scenes due to the unsegmented overall image.

[0070] S14: determining a corresponding navigation scene in a scene database based on the context information feature;

[0071] In the specific implementation process of the present invention, the corresponding navigation scene is determined in the scene database based on the context information feature, including: calculating the similarity between the context information feature and each scene feature in the scene database, and selecting the target scene feature with the highest similarity; obtaining a scene identifier based on the target scene feature, and determining the corresponding navigation scene based on the scene identifier.

[0072] Specifically, the similarity between the context information feature and each scene feature in the scene database is calculated. The similarity can be calculated by a cosine vector algorithm, the context information feature and each scene feature in the scene database are vectorized to obtain a first feature vector and a second feature vector, the cosine value of the angle between the first feature vector and the second feature vector is measured, the similarity between the two is measured by the cosine value of the angle, all similarities are compared, and the target scene feature with the highest similarity is selected; based on the target scene feature, a corresponding scene identifier is obtained in the database, and based on the scene identifier, a corresponding guide scene is determined, and each guide scene corresponds to a unique scene identifier.

[0073] S15: based on the corresponding navigation scene, a corresponding scene label text is obtained, and the similarity and difference between the scene label text and each navigation scene information in the information database are compared, and based on the similarity comparison result and the difference comparison result, the target navigation scene information is obtained, and the navigation robot displays the target navigation scene information;

[0074] In the specific implementation process of the present invention, the comparison of the similarities and differences between the scene label text and each navigation scene information in the information database includes: obtaining a label text string corresponding to the scene label text, and obtaining a plurality of information text strings corresponding to each navigation scene information; constructing a joint probability distribution between the label text string and each information text string; sampling the joint probability distribution to obtain a sampled string; calculating a first distance matrix between the label text string and the sampled string, and calculating a second distance matrix between each information text string and the sampled string; determining the similarity between the scene label text and each navigation scene information in the information database based on the first distance matrix and the second distance matrix; obtaining a difference string between the label text string and the information text string based on a hash function; performing edit distance calculation on the difference string based on an edit distance algorithm to obtain a corresponding edit distance, and determining the difference between the scene label text and each navigation scene information in the information database based on the corresponding edit distance.

[0075] Furthermore, the determining of the similarity between the scene label text and each navigation scene information in the information database based on the first distance matrix and the second distance matrix includes: performing feature extraction on the first distance matrix to obtain a first column representation vector, a first diagonal representation vector and a first row representation vector; performing feature extraction on the second distance matrix to obtain a second column representation vector, a second diagonal representation vector and a second row representation vector; and calculating the vector similarity between the first column representation vector, the first diagonal representation vector and the first row representation vector and the second column representation vector, the second diagonal representation vector and the second row representation vector.

[0076] Specifically, based on the corresponding navigation scene, the corresponding scene label text is obtained in the label text database, the label text string corresponding to the scene label text is obtained, and several information text strings corresponding to each navigation scene information are obtained; by calculating the normal distribution and exponential distribution between the label text string and each information text string, the joint probability between the two is constructed through the normal distribution and the exponential distribution, that is, the joint probability distribution between the label text string and each information text string is constructed; the joint probability distribution is sampled, and the joint probability distribution is randomly down-sampled based on a preset sampling ratio, and the preset sampling ratio is inversely proportional to the string, that is, the string is also long. , the smaller the selected preset sampling ratio is, the better the optimization effect on data processing performance is, thereby obtaining a sampled string; calculating a first distance matrix between the label text string and the sampled string, defining a basic operation cost, wherein the basic operation cost includes a character deletion cost, a character insertion cost, and a character replacement cost, calculating the basic operation cost required from the label text string to the sampled string, that is, calculating the sum of its character deletion cost, character insertion cost, and character replacement cost, generating a distance representation vector through the calculated sum, that is, generating a first distance matrix, calculating a second distance matrix between each information text string and the sampled string, and calculating the second distance matrix as shown in the first The steps for generating a distance matrix are consistent and will not be repeated here; feature extraction is performed on the first distance matrix, that is, the representation vector is extracted from the first distance matrix, the last row sequence in the first distance matrix is ​​extracted as the first row representation vector, the last column sequence in the first distance matrix is ​​extracted as the first table representation vector, and the diagonal sequence in the first distance matrix is ​​extracted as the second object representation vector to obtain the first column representation vector, the first diagonal representation vector and the first row representation vector; feature extraction is performed on the second distance matrix to obtain the second column representation vector, the second diagonal representation vector and the second row representation vector, and the second column representation vector, the second diagonal representation vector and the second row representation vector are the same The extraction process is performed from the second distance matrix. The extraction process steps of the first distance matrix are the same and will not be repeated here. The vector similarity between the first column representation vector, the first diagonal representation vector and the first row representation vector and the second column representation vector, the second diagonal representation vector and the second row representation vector is calculated. The first cosine similarity between the first column representation vector and the second column representation vector is calculated. The second cosine similarity between the first diagonal representation vector and the second diagonal representation vector is calculated. The third cosine similarity between the first diagonal representation vector and the second diagonal representation vector is calculated. The target similarity is generated according to the first cosine similarity, the second cosine similarity and the third cosine similarity according to a preset ratio.Based on the hash function, a difference string between the label text string and the information text string is obtained, a hash calculation is performed on the label text string and the information text string through the hash function, a hash table is generated according to the hash calculation result, and each string is compared with the hash table to obtain a corresponding difference string; based on the edit distance algorithm, an edit distance calculation is performed on the difference string, and an edit sub-distance is calculated for the corresponding difference string, and then the edit sub-distances are added to obtain a corresponding edit distance, and based on the corresponding edit distance, the difference between the scene label text and each navigation scene information in the information database is determined, and the edit distance between the two is compared to obtain the difference between the two characters. The difference between the character strings, that is, determining the difference between the scene label text and each navigation scene information in the information database; multiplying the target similarity and the edit distance, selecting the navigation scene information with the largest product result as the final target navigation scene information, and the navigation robot displays the target navigation scene information, and determines the final navigation scene information from the two dimensions of similarity and difference, which not only improves the accuracy of matching the navigation scene information with the navigation scene, but also improves the calculation efficiency of the similarity by processing the distance matrix generated by the character string, and improves the reliability of the navigation service of the navigation robot, so as to provide users with better navigation services. ;

[0077] S16: When the navigation robot displays the target navigation scene information, the navigation robot obtains an operation instruction input by a user, obtains corresponding operation information based on the operation instruction, and adjusts the navigation scene information based on the operation information.

[0078] In the specific implementation process of the present invention, the corresponding operation information is obtained based on the operation instruction, and the navigation scene information is adjusted based on the operation information, including: identifying the sentence type corresponding to the operation instruction, and determining the target operation type based on the corresponding sentence type; obtaining the operation attribute information corresponding to the target operation type, and obtaining the corresponding operation information based on the target operation type and the operation attribute information; the navigation robot adjusts the navigation scene information based on the corresponding operation information.

[0079] Specifically, when the navigation robot displays the target navigation scene information, the navigation robot obtains an operation instruction input by the user. The operation instruction may be a voice instruction or a touch screen instruction, etc., and identifies the sentence type corresponding to the operation instruction through a preset field, and determines the target operation type based on the corresponding sentence type. After determining the corresponding sentence type, the target operation type can be directly determined in the database; a semantic tree object is constructed through a discrete character set and a semantic structure, and operation attribute information corresponding to the target operation type is obtained through the semantic tree object, and corresponding operation information is obtained based on the target operation type and the operation attribute information. The navigation robot calls the corresponding program based on the corresponding operation information to adjust the navigation scene information.

[0080] In an embodiment of the present invention, the navigation area image is divided into a number of image blocks, and feature extraction is performed using a feature projection matrix through the variance of the grayscale values ​​of the pixels of the image blocks, thereby improving the extraction effect of image features and avoiding the misclassification of similar scenes due to the unsegmented overall image. In obtaining the navigation scene information, the scene label text is compared with the navigation scene information in the information database for similarity and difference, and the final navigation scene information is determined from the two dimensions of similarity and difference, which not only improves the accuracy of matching the navigation scene information with the navigation scene, but also improves the calculation efficiency of the similarity by processing the distance matrix generated by the character string, thereby improving the reliability of the navigation service of the navigation robot, thereby providing users with a better navigation service.

[0081] Embodiment 2

[0082] See also Figure 2 , Figure 2 It is a schematic diagram of the structural composition of a navigation information adjustment system for a navigation robot in an embodiment of the present invention.

[0083] like Figure 2 As shown, a navigation information adjustment system for a navigation robot, the system comprising:

[0084] A guide path generation module 21 is used to obtain a number of tourist spots input by a user based on the guide robot, and generate a guide path based on the number of tourist spots combined with the position information of the guide robot;

[0085] In the specific implementation process of the present invention, the guide path is generated based on several tourist locations combined with the location information of the guide robot, including: obtaining the corresponding location information based on the positioning device of the guide robot, and obtaining the recommended tour time and navigation data corresponding to each tourist location; based on the location information, navigation data and recommended tour time, the guide path of the guide robot is constructed according to the least time-consuming guide route.

[0086] Specifically, the tour guide robot obtains several tourist spots input by the user, obtains corresponding position information based on the positioning device of the tour guide robot, and obtains recommended tour time and navigation data corresponding to each tourist spot. The tour guide robot can obtain the corresponding recommended tour time and navigation data by querying the server; based on the position information, navigation data and recommended tour time, the tour guide robot constructs a tour path according to the least time-consuming tour route, with the least time-consuming tour route as a constraint condition, the position of the tour guide robot as the starting point, the time required between the tourist spots is obtained through the navigation data and the recommended tour time, and the path with the least time-consuming is selected to construct the least time-consuming tour path.

[0087] A navigation area image division module 22 is used for the navigation robot to move to each tourist spot according to the navigation path. When the navigation robot moves to the corresponding tourist spot, the navigation area image is acquired based on the preset image acquisition points, and the navigation area image is divided into a plurality of image blocks;

[0088] In the specific implementation process of the present invention, the navigation area image is obtained based on preset image acquisition points, and the navigation area image is divided into a number of image blocks, including: the navigation robot obtains the navigation area image at the preset image acquisition point using a shooting device at a preset angle; preprocessing the navigation area image to obtain a preprocessed navigation area image; and dividing the preprocessed navigation area image into a number of image blocks based on a sliding window.

[0089] Specifically, the tour guide robot moves to each tourist spot according to the tour guide path. When the tour guide robot moves to the corresponding tourist spot, the tour guide robot uses its built-in shooting device to obtain the tour area image according to the preset angle at the preset image acquisition point; the tour guide area image is pre-processed, the total number of pixels of the tour guide area image is calculated, the frequency of each pixel in the tour guide area image is counted, the frequency of each pixel and the total number of pixels are calculated, and the pixel distribution value is obtained. A corresponding histogram is established according to the pixel distribution value, and the corresponding histogram is linearly stretched to make the image more uniform in brightness and contrast, so as to obtain The navigation area image after brightness transformation is obtained, the navigation area image after brightness transformation is normalized to obtain a normalized navigation area image, the normalized navigation area image is enhanced, and after the image enhancement is completed, a pre-processed navigation area image is obtained; based on a sliding window, the pre-processed navigation area image is divided into a plurality of image blocks, and sliding window parameters are determined, wherein the sliding window parameters are the size and scaling ratio of the sliding window, and the positions of the sliding start point and the end point, and an expected sliding window is constructed according to the sliding window parameters, and the pre-processed navigation area image is segmented based on the expected sliding window to obtain a plurality of image blocks.

[0090] Feature acquisition module 23: used to extract features from each image block to obtain image block features, and encode the image block features to obtain context information features;

[0091] In the specific implementation process of the present invention, the feature extraction of each image block to obtain the image block features includes: calculating the variance of the pixel grayscale values ​​of the image block, and constructing a pixel grayscale value matrix based on the pixel grayscale value variance; projecting the image block using a feature projection matrix based on the pixel grayscale matrix to obtain the image block features.

[0092] Furthermore, the encoding process of the image block features to obtain context information features includes: generating sequence features based on the image block features, and obtaining semantic correlation relationships of the image blocks using a multi-layer perceptron based on the sequence features; and encoding the context information of the image blocks using the semantic correlation relationships of the image blocks based on a long short-term memory network to obtain context information features.

[0093] Specifically, the grayscale value of the pixel of the image block is calculated, and the pixel grayscale value is obtained by weighted average of the pixel values ​​of the red, green and blue channels of the pixel of the image block, and the average value of the pixel grayscale value of all the pixels is calculated, and the image block matrix of the image block is constructed. The pixel grayscale value variance is calculated based on the average value of the pixel grayscale value combined with the number of rows and columns of the image block matrix. The calculation expression of the pixel grayscale value variance is:

[0094]

[0095] Among them, σ is the variance of the pixel gray value, c is the pixel number, m is the number of rows in the image block matrix, n is the number of columns in the image block matrix, j is the number of pixels, and p is the pixel gray value. is the mean of the pixel grayscale values; construct a pixel grayscale value matrix based on the pixel grayscale value variance, calculate the matrix value through the pixel grayscale value variance statistics, obtain the co-occurrence matrix based on the matrix value, calculate the sum of all elements in the co-occurrence matrix based on the co-occurrence matrix, obtain the probability matrix through the sum of all elements in the co-occurrence matrix, extract the grayscale co-occurrence matrix through the probability matrix, and construct the pixel grayscale value matrix according to the grayscale co-occurrence matrix; project the image block based on the pixel grayscale matrix using the feature projection matrix, perform a two-dimensional discrete wavelet transform on the image block, obtain the image block after the two-dimensional discrete wavelet transform, vectorize the pixel grayscale matrix, obtain the vectorized pixel grayscale matrix, calculate the covariance matrix of the vectorized pixel grayscale matrix, perform eigenvalue decomposition on the covariance matrix, obtain the unit eigenvector, construct the feature projection matrix using the matrix transpose symbol and the unit eigenvector, project the image block with the constructed feature projection matrix, and obtain the image block feature; generate sequence features based on the image block feature, and generate sequence features through the cyclic unit module The block extracts sequence information from the image block features to obtain sequence features. The recurrent unit module is a module for extracting feature expressions from the sequences. Based on the sequence features, the semantic correlation of the image blocks is obtained by using a multi-layer perceptron. The probability features of the image blocks are obtained by using the multi-layer perceptron. The sequence features are used to iterate to a preset number of times according to the probability features of the image blocks. The multi-layer perceptron outputs the correlation of the image blocks at the semantic level, that is, the semantic correlation of the image blocks is obtained. The context information of the image blocks is encoded by using the semantic correlation of the image blocks based on the long short-term memory network. The characteristics of the long short-term memory network for processing time series information are used to filter out the features that do not have discriminative performance in the image blocks by using the semantic correlation through the long short-term memory network, encode the image blocks and their context information, and obtain context information features. The feature projection matrix is ​​used to extract features through the grayscale value variance of the pixel points of the image blocks, thereby improving the extraction effect of image features and avoiding the misclassification of similar scenes due to the unsegmented overall image.

[0096] A navigation scene determination module 24 is used to determine a corresponding navigation scene in a scene database based on the context information feature;

[0097] In the specific implementation process of the present invention, the corresponding navigation scene is determined in the scene database based on the context information feature, including: calculating the similarity between the context information feature and each scene feature in the scene database, and selecting the target scene feature with the highest similarity; obtaining a scene identifier based on the target scene feature, and determining the corresponding navigation scene based on the scene identifier.

[0098] Specifically, the similarity between the context information feature and each scene feature in the scene database is calculated. The similarity can be calculated by a cosine vector algorithm, the context information feature and each scene feature in the scene database are vectorized to obtain a first feature vector and a second feature vector, the cosine value of the angle between the first feature vector and the second feature vector is measured, the similarity between the two is measured by the cosine value of the angle, all similarities are compared, and the target scene feature with the highest similarity is selected; based on the target scene feature, a corresponding scene identifier is obtained in the database, and based on the scene identifier, a corresponding guide scene is determined, and each guide scene corresponds to a unique scene identifier.

[0099] The navigation scene information acquisition module 25 is used to acquire the corresponding scene label text based on the corresponding navigation scene, and compare the similarity and difference between the scene label text and each navigation scene information in the information database, and acquire the target navigation scene information based on the similarity comparison result and the difference comparison result, and the navigation robot displays the target navigation scene information;

[0100] In the specific implementation process of the present invention, the comparison of the similarities and differences between the scene label text and each navigation scene information in the information database includes: obtaining a label text string corresponding to the scene label text, and obtaining a plurality of information text strings corresponding to each navigation scene information; constructing a joint probability distribution between the label text string and each information text string; sampling the joint probability distribution to obtain a sampled string; calculating a first distance matrix between the label text string and the sampled string, and calculating a second distance matrix between each information text string and the sampled string; determining the similarity between the scene label text and each navigation scene information in the information database based on the first distance matrix and the second distance matrix; obtaining a difference string between the label text string and the information text string based on a hash function; performing edit distance calculation on the difference string based on an edit distance algorithm to obtain a corresponding edit distance, and determining the difference between the scene label text and each navigation scene information in the information database based on the corresponding edit distance.

[0101] Furthermore, the determining of the similarity between the scene label text and each navigation scene information in the information database based on the first distance matrix and the second distance matrix includes: performing feature extraction on the first distance matrix to obtain a first column representation vector, a first diagonal representation vector and a first row representation vector; performing feature extraction on the second distance matrix to obtain a second column representation vector, a second diagonal representation vector and a second row representation vector; and calculating the vector similarity between the first column representation vector, the first diagonal representation vector and the first row representation vector and the second column representation vector, the second diagonal representation vector and the second row representation vector.

[0102] Specifically, based on the corresponding navigation scene, the corresponding scene label text is obtained in the label text database, the label text string corresponding to the scene label text is obtained, and several information text strings corresponding to each navigation scene information are obtained; by calculating the normal distribution and exponential distribution between the label text string and each information text string, the joint probability between the two is constructed through the normal distribution and the exponential distribution, that is, the joint probability distribution between the label text string and each information text string is constructed; the joint probability distribution is sampled, and the joint probability distribution is randomly down-sampled based on a preset sampling ratio, and the preset sampling ratio is inversely proportional to the string, that is, the string is also long. , the smaller the selected preset sampling ratio is, the better the optimization effect on data processing performance is, thereby obtaining a sampled string; calculating a first distance matrix between the label text string and the sampled string, defining a basic operation cost, wherein the basic operation cost includes a character deletion cost, a character insertion cost, and a character replacement cost, calculating the basic operation cost required from the label text string to the sampled string, that is, calculating the sum of its character deletion cost, character insertion cost, and character replacement cost, generating a distance representation vector through the calculated sum, that is, generating a first distance matrix, calculating a second distance matrix between each information text string and the sampled string, and calculating the second distance matrix as shown in the first The steps for generating a distance matrix are consistent and will not be repeated here; feature extraction is performed on the first distance matrix, that is, the representation vector is extracted from the first distance matrix, the last row sequence in the first distance matrix is ​​extracted as the first row representation vector, the last column sequence in the first distance matrix is ​​extracted as the first table representation vector, and the diagonal sequence in the first distance matrix is ​​extracted as the second object representation vector to obtain the first column representation vector, the first diagonal representation vector and the first row representation vector; feature extraction is performed on the second distance matrix to obtain the second column representation vector, the second diagonal representation vector and the second row representation vector, and the second column representation vector, the second diagonal representation vector and the second row representation vector are the same The extraction process is performed from the second distance matrix. The extraction process steps of the first distance matrix are the same and will not be repeated here. The vector similarity between the first column representation vector, the first diagonal representation vector and the first row representation vector and the second column representation vector, the second diagonal representation vector and the second row representation vector is calculated. The first cosine similarity between the first column representation vector and the second column representation vector is calculated. The second cosine similarity between the first diagonal representation vector and the second diagonal representation vector is calculated. The third cosine similarity between the first diagonal representation vector and the second diagonal representation vector is calculated. The target similarity is generated according to the first cosine similarity, the second cosine similarity and the third cosine similarity according to a preset ratio.Based on the hash function, a difference string between the label text string and the information text string is obtained, a hash calculation is performed on the label text string and the information text string through the hash function, a hash table is generated according to the hash calculation result, and each string is compared with the hash table to obtain a corresponding difference string; based on the edit distance algorithm, an edit distance calculation is performed on the difference string, and an edit sub-distance is calculated for the corresponding difference string, and then the edit sub-distances are added to obtain a corresponding edit distance, and based on the corresponding edit distance, the difference between the scene label text and each navigation scene information in the information database is determined, and the edit distance between the two is compared to obtain the difference between the two characters. The difference between the character strings, that is, determining the difference between the scene label text and each navigation scene information in the information database; multiplying the target similarity and the edit distance, selecting the navigation scene information with the largest product result as the final target navigation scene information, and the navigation robot displays the target navigation scene information, and determines the final navigation scene information from the two dimensions of similarity and difference, which not only improves the accuracy of matching the navigation scene information with the navigation scene, but also improves the calculation efficiency of the similarity by processing the distance matrix generated by the character string, and improves the reliability of the navigation service of the navigation robot, so as to provide users with better navigation services. ;

[0103] The navigation scene information adjustment module 26 is used for the navigation robot to obtain the operation instruction input by the user when the navigation robot displays the target navigation scene information, obtain the corresponding operation information based on the operation instruction, and adjust the navigation scene information based on the operation information.

[0104] In the specific implementation process of the present invention, the corresponding operation information is obtained based on the operation instruction, and the navigation scene information is adjusted based on the operation information, including: identifying the sentence type corresponding to the operation instruction, and determining the target operation type based on the corresponding sentence type; obtaining the operation attribute information corresponding to the target operation type, and obtaining the corresponding operation information based on the target operation type and the operation attribute information; the navigation robot adjusts the navigation scene information based on the corresponding operation information.

[0105] Specifically, when the navigation robot displays the target navigation scene information, the navigation robot obtains an operation instruction input by the user. The operation instruction may be a voice instruction or a touch screen instruction, etc., and identifies the sentence type corresponding to the operation instruction through a preset field, and determines the target operation type based on the corresponding sentence type. After determining the corresponding sentence type, the target operation type can be directly determined in the database; a semantic tree object is constructed through a discrete character set and a semantic structure, and operation attribute information corresponding to the target operation type is obtained through the semantic tree object, and corresponding operation information is obtained based on the target operation type and the operation attribute information. The navigation robot calls the corresponding program based on the corresponding operation information to adjust the navigation scene information.

[0106] In an embodiment of the present invention, the navigation area image is divided into a number of image blocks, and feature extraction is performed using a feature projection matrix through the variance of the grayscale values ​​of the pixels of the image blocks, thereby improving the extraction effect of image features and avoiding the misclassification of similar scenes due to the unsegmented overall image. In obtaining the navigation scene information, the scene label text is compared with the navigation scene information in the information database for similarity and difference, and the final navigation scene information is determined from the two dimensions of similarity and difference, which not only improves the accuracy of matching the navigation scene information with the navigation scene, but also improves the calculation efficiency of the similarity by processing the distance matrix generated by the character string, thereby improving the reliability of the navigation service of the navigation robot, thereby providing users with a better navigation service.

[0107] A person skilled in the art may understand that all or part of the steps in the various methods of the above embodiments may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the storage medium may include: a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc.

[0108] In addition, the above is a detailed introduction to a method and system for adjusting navigation information for a navigation robot provided by an embodiment of the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for a person skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for adjusting navigation information for a navigation robot, characterized in that: The method comprises: The guide robot obtains a number of tourist spots input by the user, and generates a guide route based on the number of tourist spots combined with the position information of the guide robot; The tour guide robot moves to each tourist spot according to the tour guide path. When the tour guide robot moves to the corresponding tourist spot, it acquires a tour area image based on preset image acquisition points, and divides the tour area image into a plurality of image blocks; Extracting features from each image block to obtain image block features, and encoding the image block features to obtain context information features; Determine a corresponding navigation scene in a scene database based on the context information feature; Based on the corresponding navigation scene, a corresponding scene label text is obtained, and the similarity and difference between the scene label text and each navigation scene information in the information database are compared, and based on the similarity comparison result and the difference comparison result, the target navigation scene information is obtained, and the navigation robot displays the target navigation scene information; During the process of the navigation robot displaying the target navigation scene information, the navigation robot obtains an operation instruction input by a user, obtains corresponding operation information based on the operation instruction, and adjusts the navigation scene information based on the operation information; Among them, the comparison of the similarities and differences between the scene label text and each navigation scene information in the information database includes: obtaining a label text string corresponding to the scene label text, and obtaining a number of information text strings corresponding to each navigation scene information; constructing a joint probability distribution between the label text string and each information text string; sampling the joint probability distribution to obtain a sampled string; calculating a first distance matrix between the label text string and the sampled string, and calculating a second distance matrix between each information text string and the sampled string; determining the similarity between the scene label text and each navigation scene information in the information database based on the first distance matrix and the second distance matrix; obtaining a difference string between the label text string and the information text string based on a hash function; performing edit distance calculation on the difference string based on an edit distance algorithm to obtain a corresponding edit distance, and determining the difference between the scene label text and each navigation scene information in the information database based on the corresponding edit distance.

2. The method for adjusting navigation information for a navigation robot according to claim 1, characterized in that: The generating of the guide route based on a plurality of tour locations combined with the position information of the guide robot comprises: Acquire corresponding location information based on the positioning device of the tour guide robot, and acquire the recommended tour time and navigation data corresponding to each tour location; A navigation path of the navigation robot is constructed according to the least time-consuming navigation route based on the location information, navigation data and recommended tour time.

3. The method for adjusting navigation information for a navigation robot according to claim 1, characterized in that: The step of acquiring the navigation area image based on the preset image acquisition points and dividing the navigation area image into a plurality of image blocks includes: The navigation robot uses a shooting device at a preset image acquisition point to acquire an image of the navigation area at a preset angle; Preprocessing the navigation area image to obtain a preprocessed navigation area image; The preprocessed navigation area image is divided into several image blocks based on a sliding window.

4. The method for adjusting navigation information for a navigation robot according to claim 1, characterized in that: The step of extracting features from each image block to obtain image block features includes: Calculating the pixel grayscale value variance of the image block, and constructing a pixel grayscale value matrix based on the pixel grayscale value variance; Based on the pixel gray value matrix, the image block is projected using a feature projection matrix to obtain image block features.

5. The method for adjusting navigation information for a navigation robot according to claim 1, characterized in that: The encoding process of the image block features to obtain context information features includes: Generate sequence features based on the image block features, and obtain semantic correlation relationships of the image blocks using a multi-layer perceptron based on the sequence features; Based on the long short-term memory network, the semantic correlation of image blocks is used to encode the context information of image blocks to obtain context information features.

6. The method for adjusting navigation information for a navigation robot according to claim 1, characterized in that: The determining a corresponding navigation scene in a scene database based on the context information feature includes: Calculating the similarity between the context information feature and each scene feature in the scene database, and selecting the target scene feature with the highest similarity; A scene identifier is acquired based on the target scene feature, and a corresponding navigation scene is determined based on the scene identifier.

7. The method for adjusting navigation information for a navigation robot according to claim 1, characterized in that: The determining the similarity between the scene label text and each navigation scene information in the information database based on the first distance matrix and the second distance matrix includes: Performing feature extraction on the first distance matrix to obtain a first column representation vector, a first diagonal representation vector, and a first row representation vector; Performing feature extraction on the second distance matrix to obtain a second column representation vector, a second diagonal representation vector, and a second row representation vector; The vector similarity between the first column representation vector, the first diagonal representation vector and the first row representation vector and the second column representation vector, the second diagonal representation vector and the second row representation vector is calculated.

8. The method for adjusting navigation information for a navigation robot according to claim 1, characterized in that: The acquiring corresponding operation information based on the operation instruction, and adjusting the navigation scene information based on the operation information, includes: Identify the statement type corresponding to the operation instruction, and determine the target operation type based on the corresponding statement type; Acquire operation attribute information corresponding to the target operation type, and acquire corresponding operation information based on the target operation type and the operation attribute information; The navigation robot adjusts the navigation scene information based on the corresponding operation information.

9. A navigation information adjustment system for a navigation robot, characterized in that: The system comprises: A navigation path generation module: used for obtaining a number of tourist attractions input by a user based on the navigation robot, and generating a navigation path based on the number of tourist attractions combined with the location information of the navigation robot; A navigation area image division module: used for the navigation robot to move to each tourist spot according to the navigation path. When the navigation robot moves to the corresponding tourist spot, the navigation area image is acquired based on the preset image acquisition points, and the navigation area image is divided into a plurality of image blocks; Feature acquisition module: used to extract features from each image block to obtain image block features, and encode the image block features to obtain context information features; A navigation scene determination module: used to determine a corresponding navigation scene in a scene database based on the context information feature; A navigation scene information acquisition module is used to acquire a corresponding scene label text based on a corresponding navigation scene, and compare the similarity and difference between the scene label text and each navigation scene information in the information database, and acquire target navigation scene information based on the similarity comparison result and the difference comparison result, and the navigation robot displays the target navigation scene information; A navigation scene information adjustment module: used for the navigation robot to obtain an operation instruction input by a user during the process of the navigation robot displaying the target navigation scene information, obtain corresponding operation information based on the operation instruction, and adjust the navigation scene information based on the operation information; Among them, the comparison of the similarities and differences between the scene label text and each navigation scene information in the information database includes: obtaining a label text string corresponding to the scene label text, and obtaining a number of information text strings corresponding to each navigation scene information; constructing a joint probability distribution between the label text string and each information text string; sampling the joint probability distribution to obtain a sampled string; calculating a first distance matrix between the label text string and the sampled string, and calculating a second distance matrix between each information text string and the sampled string; determining the similarity between the scene label text and each navigation scene information in the information database based on the first distance matrix and the second distance matrix; obtaining a difference string between the label text string and the information text string based on a hash function; performing edit distance calculation on the difference string based on an edit distance algorithm to obtain a corresponding edit distance, and determining the difference between the scene label text and each navigation scene information in the information database based on the corresponding edit distance.

Images