Smart dynamic perception-based visual impairment group auxiliary navigation guide method and glasses

Abstract

The application relates to the technical field of intelligent wearing equipment and artificial intelligence assistance, and particularly discloses a visual impairment crowd auxiliary navigation guiding method and glasses based on intelligent dynamic sensing, which acquires the positions of personnel in real time based on position sharing permission, and determines the motion range of the visually impaired personnel; collects the building information in the motion range based on the pre-acquired collection permission, and constructs a scene model containing blocks with a preset size; reads the positions of the personnel in real time, determines the sensing accuracy based on the scene model and the positions of the personnel, acquires the scene information based on the sensing accuracy, generates guiding information, and feeds back the guiding information to the visually impaired personnel. Through the combination of various advanced algorithms, the application realizes the accurate identification of various road conditions and obstacles, early warning, greatly reduces the collision risk in the travel process of the blind, effectively reduces the probability of accidents, and improves the travel safety.

Description

Technical Field

[0001] This invention relates to the fields of smart wearable devices and artificial intelligence-assisted technology, specifically a navigation and guidance method and glasses for visually impaired people based on intelligent dynamic perception. Background Technology

[0002] Blind people face many inconveniences in their daily lives, and it is extremely difficult for them to travel and obtain information. Traditional guide tools, such as guide canes, can only detect obstacles within a range of about 1 meter in front of them through physical contact, and cannot detect suspended obstacles or obstacles at a distance. Therefore, how to provide a new assistive solution to help visually impaired people travel is the technical problem that this invention aims to solve. Summary of the Invention

[0003] The purpose of this invention is to provide a navigation and guidance method and glasses for visually impaired people based on intelligent dynamic perception, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A method for assistive navigation and guidance for visually impaired people based on intelligent dynamic perception, the method comprising:

[0006] Send an access request to a visually impaired person, receive location sharing permissions granted by the visually impaired person, obtain the person's location in real time based on the location sharing permissions, and determine the movement range of the visually impaired person; the movement range is a layer containing color values;

[0007] Based on pre-acquired collection permissions, building information within the movement range is collected to construct a scene model containing blocks of preset sizes; in the scene model, each block has a different update frequency;

[0008] The system reads the location of people in real time, determines the perception accuracy based on the scene model and the location of people, obtains scene information based on the perception accuracy, generates guidance information, and feeds it back to visually impaired people.

[0009] The scene model is updated based on the scene information, and the scene models of existing visually impaired individuals are linked together to obtain the global scene model, which is then fed back to the visually impaired individuals.

[0010] As a further aspect of the present invention: the steps of sending an access request to a visually impaired person, receiving location sharing permissions granted by the visually impaired person, obtaining the person's location in real time based on the location sharing permissions, and determining the movement range of the visually impaired person include:

[0011] Send permission requests to visually impaired individuals and receive location sharing permissions granted by them.

[0012] Based on location sharing permissions, the location of personnel is obtained in real time and inserted into a preset map. The insertion process is as follows: the feature value of the pixel area corresponding to the personnel location in the map is increased by a preset step size. The feature value of the pixel area is updated once every time the personnel location is obtained. The acquisition frequency of personnel location is a preset value. Among them, the feature value of all pixel areas decreases continuously according to a preset rate.

[0013] Statistically analyze the feature values ​​of each pixel region in the map, determine the expansion radius based on the proportionality of the feature values, and construct a circular region centered on the pixel region.

[0014] Calculate the union of the circular regions and fit the boundary to obtain the range of motion, and simultaneously determine the simulation feature values ​​of each position within the range of motion.

[0015] As a further aspect of the present invention: the step of collecting building information within the movement range based on pre-acquired collection permissions and constructing a scene model containing blocks of a preset size includes:

[0016] Query the passable area within the movement range, and determine the collection path based on the passable area;

[0017] The acquisition path is sent to the acquisition terminal, and images containing shooting parameters within the motion range are acquired based on the pre-acquired acquisition permissions; the shooting parameters include the shooting point and the shooting angle.

[0018] The image is used to identify and locate buildings, extract building features and image dimensions;

[0019] Based on the building features, identify images containing the building, query the shooting point and shooting angle of each image, and determine the building location and building model by combining the corresponding image size;

[0020] Based on the statistical building model of building location, a scene model is obtained;

[0021] Blocks are constructed in the scene model, and the update frequency of the blocks is determined synchronously.

[0022] As a further aspect of the present invention: the step of constructing blocks in the scene model and synchronously determining the update frequency of the blocks includes:

[0023] Insert a grid into the scene model according to the preset size, and treat each grid cell as a block;

[0024] For any building model, query the blocks that intersect with the building model and merge the queried blocks;

[0025] Query the type of the building model, and determine the update frequency of the merged blocks based on the type; wherein the correspondence between the update frequency and the type is a pre-set relationship.

[0026] As a further aspect of the present invention: the steps of real-time reading of personnel location, determining perception accuracy based on scene model and personnel location, acquiring scene information based on perception accuracy, generating guidance information, and feeding it back to visually impaired personnel include:

[0027] The personnel's location is read in real time, and a ray cluster is determined with the personnel's location as the center; the included angle of the ray cluster is a preset value.

[0028] Obtain the building model closest to the person's location on each ray and simultaneously determine its distance;

[0029] Determine the required clarity based on the type of building model, and determine the perception accuracy of the building model based on the required clarity and distance;

[0030] The perception accuracy of all building models is statistically analyzed, and the maximum value is selected as the final perception accuracy.

[0031] Based on the perception accuracy, scene information is acquired, guidance information is generated, and feedback is given to visually impaired people.

[0032] As a further aspect of the present invention: the step of updating the scene model based on scene information, concatenating the existing scene models of visually impaired individuals to obtain a global scene model, and feeding back the global scene model to visually impaired individuals includes:

[0033] Identify scene information and update the building model in the scene model based on the identification results;

[0034] Query existing scene models for visually impaired individuals and calculate the union of the scene models;

[0035] The updated results of the building models of each existing visually impaired person are read in real time, inserted into the union, and the global scene model is obtained.

[0036] The global scene model is fed back to visually impaired individuals.

[0037] The present invention also provides a navigation and guidance glasses for visually impaired people based on intelligent dynamic perception, the glasses comprising:

[0038] The motion range determination module is used to send permission acquisition requests to visually impaired individuals, receive location sharing permissions granted by visually impaired individuals, acquire the location of individuals in real time based on location sharing permissions, and determine the motion range of visually impaired individuals; the motion range is a layer containing color values.

[0039] The scene model building module is used to collect building information within the movement range based on pre-acquired collection permissions and build a scene model containing blocks of preset size; in the scene model, each block has a different update frequency;

[0040] The guidance information generation module is used to read the location of the person in real time, determine the perception accuracy based on the scene model and the person's location, obtain scene information based on the perception accuracy, generate guidance information, and feed it back to the visually impaired person.

[0041] The scene concatenation and feedback module is used to update the scene model based on scene information, concatenate the existing scene models of visually impaired individuals to obtain a global scene model, and then feed the global scene model back to the visually impaired individuals.

[0042] As a further aspect of the present invention: the motion range determination module includes:

[0043] The permission acquisition unit is used to send permission acquisition requests to visually impaired persons and receive location sharing permissions granted by visually impaired persons.

[0044] The location insertion unit is used to obtain personnel locations in real time based on location sharing permissions and insert the personnel locations into a preset map. The insertion process is as follows: the feature value of the pixel area corresponding to the personnel location in the map is increased by a preset step size, and the feature value of the pixel area is updated once every time the personnel location is obtained. The personnel location acquisition frequency is a preset value. Among them, the feature values ​​of all pixel areas continuously decrease according to a preset rate.

[0045] The circular region construction unit is used to statistically analyze the feature values ​​of each pixel region in the map, determine the expansion radius based on the proportionality of the feature values, and construct a circular region centered on the pixel region.

[0046] The simulation calculation unit is used to calculate the union of the circular regions and fit the boundary to obtain the motion range, and simultaneously determine the simulation feature values ​​of each position within the motion range.

[0047] As a further aspect of the present invention: the scene model construction module includes:

[0048] The path determination unit is used to query the passable area within the movement range and determine the collection path based on the passable area;

[0049] The path application unit is used to send the acquisition path to the acquisition end and acquire images containing shooting parameters within the motion range based on the pre-acquired acquisition permissions; the shooting parameters include the shooting point and the shooting angle.

[0050] The building positioning unit is used to identify the image, locate the building, extract building features and the image size of the building;

[0051] The information processing unit is used to determine images containing the building based on its features, query the shooting point and shooting angle of each image, and determine the building location and building model by combining the corresponding image size.

[0052] The building model statistics unit is used to statistically analyze the building model based on the building location to obtain the scene model;

[0053] Block building units are used to build blocks in the scene model and synchronously determine the update frequency of the blocks.

[0054] As a further aspect of the present invention: the guidance information generation module includes:

[0055] The ray cluster determination unit is used to read the personnel position in real time and determine the ray cluster with the personnel position as the center; wherein the included angle of the ray cluster is a preset value;

[0056] The distance determination unit is used to acquire the building model closest to the personnel's location on each ray and simultaneously determine its distance;

[0057] The perception accuracy determination unit is used to determine the required clarity based on the type of the building model, and to determine the perception accuracy of the building model based on the required clarity and distance.

[0058] The maximum value selection unit is used to statistically analyze the perception accuracy of all building models and select the maximum value as the final perception accuracy.

[0059] The guidance information generation unit is used to acquire scene information based on perception accuracy, generate guidance information, and feed it back to visually impaired people.

[0060] Compared with existing technologies, the beneficial effects of the present invention are: the present invention combines a variety of advanced algorithms to achieve accurate identification of various road conditions and obstacles, provide early warnings, greatly reduce the collision risk of blind people during travel, effectively reduce the probability of accidents, and improve travel safety. Attached Figure Description

[0061] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention.

[0062] Figure 1 This is a flowchart of an assisted navigation and guidance method for visually impaired people based on intelligent dynamic perception.

[0063] Figure 2 This is the first sub-flowchart of an intelligent dynamic perception-based assisted navigation and guidance method for visually impaired people.

[0064] Figure 3 This is the second sub-flowchart of the intelligent dynamic perception-based assisted navigation and guidance method for visually impaired people.

[0065] Figure 4This is the third sub-flowchart of the intelligent dynamic perception-based assisted navigation and guidance method for visually impaired people.

[0066] Figure 5 This is the fourth sub-flowchart of the intelligent dynamic perception-based assisted navigation and guidance method for visually impaired people.

[0067] Figure 6 A block diagram showing the structural composition of navigation and guidance glasses for visually impaired people based on intelligent dynamic perception. Detailed Implementation

[0068] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0069] Figure 1 This is a flowchart of an assistive navigation and guidance method for visually impaired people based on intelligent dynamic perception. In this embodiment of the invention, an assistive navigation and guidance method for visually impaired people based on intelligent dynamic perception includes:

[0070] Step S100: Send an permission acquisition request to the visually impaired person, receive the location sharing permission granted by the visually impaired person, obtain the person's location in real time based on the location sharing permission, and determine the movement range of the visually impaired person; the movement range is a layer containing color values;

[0071] A permission request is sent to the visually impaired person. The visually impaired person will grant the method execution entity location sharing permission. Only then can the method execution entity obtain the location of the visually impaired person. If the visually impaired person does not grant permission, subsequent steps cannot be performed. In practical applications, the method execution entity will send a permission request to the visually impaired person when the visually impaired person indicates that they need guidance services. Based on the location sharing permission, the person's location is obtained in real time to determine the movement range of the visually impaired person. The conventional meaning of movement range is a set of coordinates used to determine a boundary range. However, in the technical solution of this invention, the movement range is limited to a region containing values ​​representing the location of the visually impaired person, which is then represented in the form of a layer. If only the location of the visually impaired person is considered, the color value in the layer can use a single value. If other parameters are to be introduced, more color values ​​can be used to reflect different states, such as movement speed and movement acceleration, or abnormal values ​​determined by movement speed and movement acceleration, which are specifically preset by the staff.

[0072] Step S200: Collect building information within the movement range based on the pre-acquired collection permissions, and construct a scene model containing blocks of preset size; wherein, in the scene model, each block has a different update frequency;

[0073] Once the movement range is determined, building information within that range is collected. In this invention, the building information refers only to the building's appearance. Of course, information cannot be obtained for certain special buildings. In such cases, staff need to construct some walls themselves as virtual building information. This invention will not elaborate on such special cases. The technical solution of this invention assumes that it occurs in a normal scenario. The execution subject of this method has already obtained the collection permissions for each building, which allows the collection process to proceed smoothly.

[0074] After collecting building information within the movement range, building models can be created under a preset scale. The collected building models are then used to construct a scene model. Once the scene model is obtained, it needs to be segmented to obtain blocks of different sizes in order to make the area of ​​the scene model easier to analyze.

[0075] Step S300: Real-time reading of personnel location, determination of perception accuracy based on scene model and personnel location, acquisition of scene information based on perception accuracy, generation of guidance information, and feedback to visually impaired personnel;

[0076] Based on existing location acquisition permissions, the system reads personnel locations in real time and determines their current state based on the scene model and personnel location, primarily assessing whether they are in a dangerous state, i.e., whether surrounding buildings will affect them, thereby determining the perception accuracy. In this invention's technical solution, the perception process includes: processing images captured by the camera using the Yolov12 algorithm to identify ground obstacles (vehicles, people, etc.), traffic light status, and road signs (such as signs) in real time. Through learning from a large number of traffic scene images, the model achieves an accuracy rate of over 90% in recognizing common obstacles. In this process, the higher the image accuracy, the better the recognition effect; however, higher image accuracy also consumes more acquisition and recognition resources. The adjustment target of this invention's technical solution is the image acquisition accuracy.

[0077] After acquiring images based on perception accuracy, the images are recognized. The recognition results are called scene information, specifically the distribution of surrounding objects. This information is then converted into guidance information and fed back to visually impaired individuals. This guidance information is generally in Braille and can be read by visually impaired individuals.

[0078] Step S400: Update the scene model according to the scene information, connect the existing scene models of visually impaired people to obtain the global scene model, and feed the global scene model back to the visually impaired people.

[0079] During use, visually impaired users can continuously acquire scene information. The scene information and the data collected in step S200 are actually from the same source. Therefore, the scene model can also be updated based on the scene information. The execution subject of this method is simultaneously facing multiple visually impaired users. By linking the scene models of existing visually impaired users, a global scene model can be obtained. This panoramic model has extremely high real-time performance and can serve as an up-to-date map to be fed back to the visually impaired users. It should be noted that the feedback process also needs to be converted into Braille before feedback.

[0080] Figure 2 The first sub-flowchart of the intelligent dynamic perception-based assisted navigation and guidance method for visually impaired people includes the following steps: sending an access request to the visually impaired person, receiving location sharing permissions granted by the visually impaired person, obtaining the person's location in real time based on the location sharing permissions, and determining the movement range of the visually impaired person.

[0081] Step S101: Send an permission request to the visually impaired person and receive location sharing permission granted by the visually impaired person;

[0082] Step S102: Obtain personnel location in real time based on location sharing permission, and insert the personnel location into a preset map; the insertion process is as follows: increase the feature value of the pixel area corresponding to the personnel location in the map by a preset step size, and update the feature value of the pixel area once every time the personnel location is obtained; the personnel location acquisition frequency is a preset value; wherein, the feature value of all pixel areas continuously decreases according to a preset rate;

[0083] Step S103: Calculate the feature values ​​of each pixel area in the map, determine the expansion radius based on the proportionality of the feature values, and construct a circular area centered on the pixel area;

[0084] Step S104: Calculate the union of the circular regions and fit the boundary to obtain the motion range, and simultaneously determine the simulation feature values ​​of each position within the motion range.

[0085] In one example of the technical solution of this invention, the process of determining the range of motion is described: a permission acquisition request is sent to the visually impaired person, location sharing permission is received from the visually impaired person, the person's location is obtained in real time based on the location sharing permission, and the person's location is inserted into a preset map. The insertion process is as follows:

[0086] First, the feature values ​​of all pixels in the preset map are set to zero. Then, the location of a person is acquired every preset time interval (the acquisition frequency of the person's location is a preset value). Each time the location of a person is acquired, the feature value of the pixel area will increase once. The increase method is to increase the feature value of the pixel area corresponding to the location of the person in the map by a preset step size, that is, increase a certain value at a time. Finally, the feature values ​​of all pixel areas will continuously decrease according to a preset rate. According to this insertion process, the feature value of the places on the map where visually impaired people have passed will increase, but the feature value will continuously decrease over time. Therefore, it reflects the location of visually impaired people over a period of time.

[0087] It should be noted that the above content refers to pixel regions, not individual pixels. This is because visually impaired individuals have a volume, corresponding to an area, such as a circular area, rather than a single pixel. Of course, these circular areas are relatively small, and in extreme cases, using individual pixels directly is feasible. The feature values ​​of each pixel region are statistically analyzed on the map, and the expansion radius is determined based on the proportionality of the feature values. A circular region centered on the pixel region is constructed, the union of the circular regions is calculated, and the boundary is fitted to obtain the range of motion. Simultaneously, the simulation feature values ​​of each position within the range of motion are determined.

[0088] The boundary between the two circular regions is formed by generating an arc tangent to both circular edges. This is a standard fitting process and will not be elaborated upon here.

[0089] Specifically, the process for determining any eigenvalue is as follows:

[0090] ;

[0091] In the formula, for Time and location Updated feature values, for Time and location Updated feature values, The preset interval for acquiring personnel locations, The coefficient is the position. When it belongs to the pixel area corresponding to the person's location, =1, when position When it does not belong to the pixel area corresponding to the person's location, =0; The step size is mentioned; Let be the rate of descent, and represent the amount of descent per unit time; where, Generally, a very small value is chosen. A minimum value is usually set, and the decrease stops when the characteristic value decreases to the minimum value over time.

[0092] The process of determining the simulation feature value is as follows: for a location within the movement range where no person has passed, query its distance from each existing feature value point, determine a coefficient by the inverse ratio of the distance, multiply the coefficient by the corresponding feature value, and then add up all the multiplied values ​​to obtain the final value. Of course, this is only one simulation method. In fact, directly using the nearest feature value is also a feasible technical solution.

[0093] Figure 3 The second sub-flowchart of the intelligent dynamic perception-based assisted navigation and guidance method for visually impaired people includes the following steps: collecting building information within the movement range based on pre-acquired acquisition permissions, and constructing a scene model containing blocks of a preset size.

[0094] Step S201: Query the passable area within the movement range, and determine the collection path based on the passable area;

[0095] Step S202: Send the acquisition path to the acquisition terminal, and acquire images containing shooting parameters within the motion range based on the pre-acquired acquisition permissions; the shooting parameters include the shooting point and the shooting angle;

[0096] Step S203: Recognize the image, locate the building, extract the building features and the image size of the building;

[0097] Step S204: Determine the images containing the building based on its features, query the shooting point and shooting angle of each image, and determine the building location and building model by combining the corresponding image size;

[0098] Step S205: Based on the building location, statistically analyze the building model to obtain the scene model;

[0099] Step S206: Construct blocks in the scene model and synchronously determine the update frequency of the blocks.

[0100] In one example of the technical solution of this invention, the movement range is essentially a map. The system queries the passable areas within the movement range, determines the acquisition path based on the passable areas, sends the acquisition path to the acquisition terminal, and acquires images containing shooting parameters within the movement range based on the pre-acquired acquisition permissions. The acquisition terminal is a device with movement and shooting functions, which can be used to capture images, such as an acquisition vehicle, for acquiring information in the preprocessing stage. The shooting parameters include the shooting point and the shooting angle.

[0101] The image is identified, and the building is located. After the location is completed, the building features and image size can be extracted. The building features are the image features corresponding to the building, and the image size is the length, width and height of the building in the image. The same building is located in the acquired image based on the building features. That is, the image containing the building is determined based on the building features. This only requires a simple image traversal process. The shooting point and shooting angle of each image are queried. Combined with the corresponding image size, the building position and building model are determined. This is an image-based modeling process. When the shooting point, shooting angle, building location result and image size are known, the modeling process is very easy.

[0102] Based on the building location statistical building model, a scene model is obtained. Then, the scene model is divided into partitions, and blocks are constructed in the scene model. The update frequency of each block is determined synchronously.

[0103] As a preferred embodiment of the technical solution of the present invention, the step of constructing blocks in the scene model and synchronously determining the update frequency of the blocks includes:

[0104] Insert a grid into the scene model according to the preset size, and treat each grid cell as a block;

[0105] For any building model, query the blocks that intersect with the building model and merge the queried blocks;

[0106] Query the type of the building model, and determine the update frequency of the merged blocks based on the type; wherein the correspondence between the update frequency and the type is a pre-set relationship.

[0107] In one example of the technical solution of this invention, the partitioning process is described. A grid is inserted into the scene model according to a preset size, and each grid cell is taken as a block. For any building model, blocks that intersect with the building model are queried, and the queried blocks are merged. The type of the building model is queried, and the update frequency of the merged blocks is determined according to the type. The correspondence between the update frequency and the type is a preset relationship. This process is actually adjusting the working process of the acquisition end and simplifying its workload. For some relatively conventional buildings, it may be possible to acquire data once a month. For buildings under construction, the acquisition frequency should be higher.

[0108] Figure 4 The third sub-flowchart of the intelligent dynamic perception-based assisted navigation and guidance method for visually impaired people includes the following steps: real-time reading of the person's position, determining the perception accuracy based on the scene model and the person's position, acquiring scene information based on the perception accuracy, generating guidance information, and feeding it back to the visually impaired person:

[0109] Step S301: Read the personnel position in real time, and determine the ray cluster with the personnel position as the center; wherein the included angle of the ray cluster is a preset value;

[0110] Step S302: Obtain the building model closest to the personnel's location on each ray, and simultaneously determine its distance;

[0111] Step S303: Determine the required clarity based on the type of building model, and determine the perception accuracy of the building model based on the required clarity and distance;

[0112] Step S304: Calculate the perception accuracy of all building models and select the maximum value as the final perception accuracy;

[0113] Step S305: Obtain scene information based on perception accuracy, generate guidance information, and feed it back to visually impaired personnel.

[0114] In one example of the technical solution of this invention, the perception process is described. The location of the person is read in real time. A ray cluster is determined with the person's location as the center. The angle between the ray clusters is a preset value, for example, the angle between two adjacent rays is 30 degrees. The building model closest to the person's location on each ray is obtained, and its distance is determined synchronously. The required sharpness is determined according to the type of building model. The perception accuracy of the building model is determined according to the required sharpness and the distance. The perception accuracy is the sharpness of the camera. The required sharpness is the required sharpness of each building, which is defined as the number of pixels per unit area. The more pixels, the higher the sharpness.

[0115] After each building model obtains a perception accuracy, the maximum value is selected as the final perception accuracy. Based on the perception accuracy, scene information is acquired, guidance information is generated, and feedback is given to visually impaired people.

[0116] The process of generating and feeding back guidance information is explained below:

[0117] The model utilizes the Yolov12 algorithm to process images captured by cameras, enabling real-time identification of ground obstacles (vehicles, people, etc.), traffic light status, and road signs (such as directional signs). Through learning from a large number of traffic scene images, the model achieves an accuracy rate of over 90% in identifying common obstacles. Furthermore, the Deep Anything v2 algorithm can be applied to segment and perform depth analysis on objects in complex scenes, distinguishing between walls and suspended obstacles to improve recognition accuracy. Additionally, combined with the intelligent analysis capabilities of the Qwen model, navigation information is generated and promptly communicated to the user via a voice module. Pytts3 speech synthesis technology is used to deliver the route information to the user in clear, real-time speech.

[0118] Figure 5The fourth sub-flowchart of the intelligent dynamic perception-based assisted navigation and guidance method for visually impaired people includes the following steps: updating the scene model based on scene information, connecting the existing scene models of visually impaired people to obtain a global scene model, and feeding back the global scene model to the visually impaired people.

[0119] Step S401: Identify the scene information and update the building model in the scene model based on the identification results;

[0120] Step S402: Query the existing scene models of visually impaired individuals and calculate the union of the scene models;

[0121] Step S403: Read the update results of the building model of each existing visually impaired person in real time, insert them into the union, and obtain the global scene model;

[0122] Step S404: Feed back the global scene model to the visually impaired person.

[0123] In one embodiment of the technical solution of this invention, an extended solution is also provided. When there are many visually impaired people, for each visually impaired person, their corresponding scene information is read, the scene information is identified, the building model in the scene model is updated according to the identification result, the scene models of existing visually impaired people are queried, the union of the scene models is calculated, the update result of the building model of each existing visually impaired person is read in real time, and it is inserted into the union to obtain the global scene model. Thus, a comprehensive scene model jointly determined by multiple visually impaired people is constructed, called the global scene model. The global scene model is fed back to the visually impaired people. Of course, the feedback method is broadcast, and it is sent to the visually impaired people who provided the scene information at the same time.

[0124] It is worth mentioning that when feeding back the global scene model, it needs to be simplified into a Braille map. At this time, since it is not a navigation scenario but only an information sharing scenario, it is generally not converted into voice form.

[0125] Figure 6 The diagram shows the structural composition of an intelligent dynamic perception-based assistive navigation and guidance glasses for visually impaired individuals. In this embodiment of the invention, an assistive navigation and guidance glasses for visually impaired individuals based on intelligent dynamic perception is provided, wherein the glasses 10 comprises:

[0126] The motion range determination module 11 is used to send an permission acquisition request to the visually impaired person, receive the location sharing permission granted by the visually impaired person, obtain the person's location in real time based on the location sharing permission, and determine the motion range of the visually impaired person; the motion range is a layer containing color values.

[0127] The scene model construction module 12 is used to collect building information within the movement range based on the pre-acquired collection permissions, and construct a scene model containing blocks of preset size; wherein, in the scene model, each block has a different update frequency;

[0128] The guidance information generation module 13 is used to read the location of the person in real time, determine the perception accuracy based on the scene model and the person's location, obtain scene information based on the perception accuracy, generate guidance information, and feed it back to the visually impaired person.

[0129] The scene connection feedback module 14 is used to update the scene model according to the scene information, connect the existing scene models of visually impaired people to obtain the global scene model, and feed the global scene model back to the visually impaired people.

[0130] Furthermore, the motion range determination module 11 includes:

[0131] The permission acquisition unit is used to send permission acquisition requests to visually impaired persons and receive location sharing permissions granted by visually impaired persons.

[0132] The location insertion unit is used to obtain personnel locations in real time based on location sharing permissions and insert the personnel locations into a preset map. The insertion process is as follows: the feature value of the pixel area corresponding to the personnel location in the map is increased by a preset step size, and the feature value of the pixel area is updated once every time the personnel location is obtained. The personnel location acquisition frequency is a preset value. Among them, the feature values ​​of all pixel areas continuously decrease according to a preset rate.

[0133] The circular region construction unit is used to statistically analyze the feature values ​​of each pixel region in the map, determine the expansion radius based on the proportionality of the feature values, and construct a circular region centered on the pixel region.

[0134] The simulation calculation unit is used to calculate the union of the circular regions and fit the boundary to obtain the motion range, and simultaneously determine the simulation feature values ​​of each position within the motion range.

[0135] Specifically, the scene model construction module 12 includes:

[0136] The path determination unit is used to query the passable area within the movement range and determine the collection path based on the passable area;

[0137] The path application unit is used to send the acquisition path to the acquisition end and acquire images containing shooting parameters within the motion range based on the pre-acquired acquisition permissions; the shooting parameters include the shooting point and the shooting angle.

[0138] The building positioning unit is used to identify the image, locate the building, extract building features and the image size of the building;

[0139] The information processing unit is used to determine images containing the building based on its features, query the shooting point and shooting angle of each image, and determine the building location and building model by combining the corresponding image size.

[0140] The building model statistics unit is used to statistically analyze the building model based on the building location to obtain the scene model;

[0141] Block building units are used to build blocks in the scene model and synchronously determine the update frequency of the blocks.

[0142] Furthermore, the guidance information generation module 13 includes:

[0143] The ray cluster determination unit is used to read the personnel position in real time and determine the ray cluster with the personnel position as the center; wherein the included angle of the ray cluster is a preset value;

[0144] The distance determination unit is used to acquire the building model closest to the personnel's location on each ray and simultaneously determine its distance;

[0145] The perception accuracy determination unit is used to determine the required clarity based on the type of the building model, and to determine the perception accuracy of the building model based on the required clarity and distance.

[0146] The maximum value selection unit is used to statistically analyze the perception accuracy of all building models and select the maximum value as the final perception accuracy.

[0147] The guidance information generation unit is used to acquire scene information based on perception accuracy, generate guidance information, and feed it back to visually impaired people.

[0148] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for assistive navigation and guidance for visually impaired people based on intelligent dynamic perception, characterized in that, The method includes: Send an access request to a visually impaired person, receive location sharing permissions granted by the visually impaired person, obtain the person's location in real time based on the location sharing permissions, and determine the movement range of the visually impaired person; the movement range is a layer containing color values; Based on pre-acquired collection permissions, building information within the movement range is collected to construct a scene model containing blocks of preset sizes; in the scene model, each block has a different update frequency; The system reads the location of people in real time, determines the perception accuracy based on the scene model and the location of people, obtains scene information based on the perception accuracy, generates guidance information, and feeds it back to visually impaired people. The scene model is updated based on the scene information, and the existing scene models of visually impaired individuals are linked together to obtain the global scene model, which is then fed back to the visually impaired individuals. The step of collecting building information within the movement range based on pre-acquired collection permissions and constructing a scene model containing blocks of preset sizes includes: Query the passable area within the movement range, and determine the collection path based on the passable area; The acquisition path is sent to the acquisition terminal, and images containing shooting parameters within the motion range are acquired based on the pre-acquired acquisition permissions; the shooting parameters include the shooting point and the shooting angle. The image is used to identify and locate buildings, extract building features and image dimensions; Based on the building features, identify images containing the building, query the shooting point and shooting angle of each image, and determine the building location and building model by combining the corresponding image size; Based on the statistical building model of building location, a scene model is obtained; Construct blocks in the scene model and synchronously determine the update frequency of the blocks; The steps of real-time reading of personnel location, determining perception accuracy based on scene model and personnel location, acquiring scene information based on perception accuracy, generating guidance information, and feeding it back to visually impaired personnel include: The personnel's location is read in real time, and a ray cluster is determined with the personnel's location as the center; the included angle of the ray cluster is a preset value. Obtain the building model closest to the person's location on each ray and simultaneously determine its distance; Determine the required clarity based on the type of building model, and determine the perception accuracy of the building model based on the required clarity and distance; The perception accuracy of all building models is statistically analyzed, and the maximum value is selected as the final perception accuracy. Based on the perception accuracy, scene information is acquired, guidance information is generated, and feedback is given to visually impaired people.

2. The assisted navigation and guidance method for visually impaired people based on intelligent dynamic perception according to claim 1, characterized in that, The steps of sending an access request to a visually impaired person, receiving location sharing permissions granted by the visually impaired person, obtaining the person's location in real time based on the location sharing permissions, and determining the visually impaired person's range of motion include: Send permission requests to visually impaired individuals and receive location sharing permissions granted by them. Based on location sharing permissions, the location of personnel is obtained in real time and inserted into a preset map. The insertion process is as follows: the feature value of the pixel area corresponding to the personnel location in the map is increased by a preset step size. The feature value of the pixel area is updated once every time the personnel location is obtained. The acquisition frequency of personnel location is a preset value. Among them, the feature value of all pixel areas decreases continuously according to a preset rate. Statistically analyze the feature values ​​of each pixel region in the map, determine the expansion radius based on the proportionality of the feature values, and construct a circular region centered on the pixel region. Calculate the union of the circular regions and fit the boundary to obtain the range of motion, and simultaneously determine the simulation feature values ​​of each position within the range of motion.

3. The assisted navigation and guidance method for visually impaired people based on intelligent dynamic perception according to claim 1, characterized in that, The steps of constructing blocks in the scene model and synchronously determining the update frequency of the blocks include: Insert a grid into the scene model according to the preset size, and treat each grid cell as a block; For any building model, query the blocks that intersect with the building model and merge the queried blocks; Query the type of the building model, and determine the update frequency of the merged blocks based on the type; wherein the correspondence between the update frequency and the type is a pre-set relationship.

4. The assisted navigation and guidance method for visually impaired people based on intelligent dynamic perception according to claim 1, characterized in that, The steps of updating the scene model based on scene information, connecting the existing scene models of visually impaired individuals to obtain a global scene model, and feeding back the global scene model to visually impaired individuals include: Identify scene information and update the building model in the scene model based on the identification results; Query existing scene models for visually impaired individuals and calculate the union of the scene models; The updated results of the building models of each existing visually impaired person are read in real time, inserted into the union, and the global scene model is obtained. The global scene model is fed back to visually impaired individuals.

5. A navigation and guidance glasses for visually impaired people based on intelligent dynamic perception, characterized in that, The eyeglasses include: The motion range determination module is used to send permission acquisition requests to visually impaired individuals, receive location sharing permissions granted by visually impaired individuals, acquire the location of individuals in real time based on location sharing permissions, and determine the motion range of visually impaired individuals; the motion range is a layer containing color values. The scene model building module is used to collect building information within the movement range based on pre-acquired collection permissions and build a scene model containing blocks of preset size; in the scene model, each block has a different update frequency; The guidance information generation module is used to read the location of the person in real time, determine the perception accuracy based on the scene model and the person's location, obtain scene information based on the perception accuracy, generate guidance information, and feed it back to the visually impaired person. The scene concatenation and feedback module is used to update the scene model based on scene information, concatenate the existing scene models of visually impaired individuals to obtain a global scene model, and then feed the global scene model back to the visually impaired individuals. The scene model construction module includes: The path determination unit is used to query the passable area within the movement range and determine the collection path based on the passable area; The path application unit is used to send the acquisition path to the acquisition end and acquire images containing shooting parameters within the motion range based on the pre-acquired acquisition permissions; the shooting parameters include the shooting point and the shooting angle. The building positioning unit is used to identify the image, locate the building, extract building features and the image size of the building; The information processing unit is used to determine images containing the building based on its features, query the shooting point and shooting angle of each image, and determine the building location and building model by combining the corresponding image size. The building model statistics unit is used to statistically analyze the building model based on the building location to obtain the scene model; Block building unit, used to build blocks in scene model and synchronously determine the block update frequency; The guidance information generation module includes: The ray cluster determination unit is used to read the personnel position in real time and determine the ray cluster with the personnel position as the center; wherein the included angle of the ray cluster is a preset value; The distance determination unit is used to acquire the building model closest to the personnel's location on each ray and simultaneously determine its distance; The perception accuracy determination unit is used to determine the required clarity based on the type of the building model, and to determine the perception accuracy of the building model based on the required clarity and distance. The maximum value selection unit is used to statistically analyze the perception accuracy of all building models and select the maximum value as the final perception accuracy. The guidance information generation unit is used to acquire scene information based on perception accuracy, generate guidance information, and feed it back to visually impaired people.

6. The intelligent dynamic perception-based auxiliary navigation and guidance glasses for visually impaired people according to claim 5, characterized in that, The motion range determination module includes: The permission acquisition unit is used to send permission acquisition requests to visually impaired persons and receive location sharing permissions granted by visually impaired persons. The location insertion unit is used to obtain personnel locations in real time based on location sharing permissions and insert the personnel locations into a preset map. The insertion process is as follows: the feature value of the pixel area corresponding to the personnel location in the map is increased by a preset step size, and the feature value of the pixel area is updated once every time the personnel location is obtained. The personnel location acquisition frequency is a preset value. Among them, the feature values ​​of all pixel areas continuously decrease according to a preset rate. The circular region construction unit is used to statistically analyze the feature values ​​of each pixel region in the map, determine the expansion radius based on the proportionality of the feature values, and construct a circular region centered on the pixel region. The simulation calculation unit is used to calculate the union of the circular regions and fit the boundary to obtain the motion range, and simultaneously determine the simulation feature values ​​of each position within the motion range.

Images