A blind navigation path screening method and system fusing semantics and real-time information

Abstract

The application belongs to the technical field of blind person auxiliary navigation, and relates to a blind person navigation path screening method and system fusing semantics and real-time information, which comprises the following steps: generating a city road data packet containing road basic attributes and street view interest points; generating a safe resource point list containing positions and attribute descriptions; generating a resilience skeleton path from a starting point to a destination; generating a user behavior state signal reflecting a user real-time travel state; generating an environment abnormal qualitative label; calculating and outputting a guide instruction guiding to a nearest available safe resource point; taking the safe resource point as a new starting point, re-planning a local update path segment to a destination coordinate; and forming a complete dynamically adjusted navigation path. The application solves the problems that the existing scheme provides options of returning to an original path or re-planning a global path, the former exposes the user to risks again, and the latter interrupts the continuity of navigation and causes delay due to calculation of a global optimal path.

Description

Technical Field

[0001] This invention belongs to the technical field of assisted navigation for the blind, and relates to a method and system for selecting navigation paths for the blind that integrates semantics and real-time information. Background Technology

[0002] While current electronic navigation applications for the visually impaired have made some progress in providing basic route guidance, their core functions still primarily rely on GPS positioning and static electronic map route planning. These solutions typically prioritize minimizing route distance or estimated travel time, resulting in fixed routes that lack consideration for the safety of the physical environment along the route, particularly the availability of temporary safe havens. In the complex urban environment, pre-planned static routes may pass through construction zones, temporarily controlled traffic sections, or areas with exceptionally high pedestrian and vehicular traffic. These dynamically changing environmental factors pose a major challenge to the independent and safe passage of visually impaired users.

[0003] The commonly adopted solutions in the industry are to adapt the voice broadcast interface for visually impaired users based on general navigation software, and to integrate real-time traffic information to avoid motor vehicle congestion. Some solutions attempt to introduce points of interest (POIs) as landmark references along the navigation route. The core logic of these methods is still based on the most efficient static or quasi-static path planning. The generation and adjustment of the path do not incorporate unforeseen immediate obstacles or dangerous environments during the user's journey into the decision-making model. Once the traditional navigation process begins, it guides the user along a predetermined route, lacking an effective perception and response mechanism for user hesitation, pauses, or reversals caused by sudden environmental changes.

[0004] Based on the above problems, existing technologies have many limitations in practical applications. When users need to temporarily change their routes to avoid risks, existing solutions often provide options to return to the original path or replan the global path. The former exposes users to risks again, while the latter interrupts the continuity of navigation and causes delays due to the calculation of the global optimal path, and needs to be improved in terms of timeliness. Summary of the Invention

[0005] In a first aspect, the present invention provides a method for selecting navigation paths for the blind that integrates semantic and real-time information, employing the following technical solution:

[0006] A method for selecting navigation paths for the blind that integrates semantics and real-time information includes the following steps:

[0007] S1. Obtain the coordinates of the starting point and destination of the route planning, and extract road network data and street view points of interest data from the electronic map to generate a city road data package containing basic road attributes and street view points of interest.

[0008] S2. Based on urban road data packages, identify and label physical spaces along roads that conform to preset geometric and functional characteristics, and generate a list of safe resource points containing location and attribute descriptions.

[0009] S3. Based on urban road data packages and a list of safe resource points, global path planning is performed with the distribution density of safe resource points along the path as one of the optimization objectives, generating a resilient skeleton path from the starting point to the destination.

[0010] S4. During navigation along the resilient skeleton path, continuously analyze the inertial measurement unit data to generate user behavior status signals that reflect the user's real-time travel status.

[0011] S5. During navigation, continuously analyze environmental sound data and generate qualitative markers for environmental anomalies by matching them with a preset abnormal sound pattern library.

[0012] S6. Receive user behavior status signals and environmental anomaly qualitative markers. When both meet the preset alarm triggering conditions, calculate and output a guidance instruction to guide to the nearest available security resource point based on the current location and the list of security resource points.

[0013] S7. After the user moves according to the guidance instructions and confirms that they have arrived at the target safe resource point, the local updated path segment to the destination coordinates is replanned using the safe resource point as the new starting point.

[0014] S8. The user's completed actual walking trajectory is seamlessly spliced ​​with the locally updated path segment to form a complete dynamically adjusted navigation path.

[0015] A further aspect of this invention involves generating a city road data package containing basic road attributes and street view points of interest, comprising the following steps:

[0016] The navigation application receives the user's route planning request and parses it to obtain the coordinates of the starting point and destination of the route planning.

[0017] Send a request to the online map service application programming interface to obtain data on basic road attributes within an area centered on the line connecting the origin and destination and covering potential paths. The basic road attributes include road name, lane width, sidewalk width, and road surface material.

[0018] The same map service acquires point-of-interest (POI) data for the area. Street view POIs include the names and locations of street-front shops, public facilities, plaza entrances, and bus stops.

[0019] Integrate basic road attribute data and its attributes with street view points of interest data to generate urban road data packages.

[0020] A further aspect of the present invention generates a list of secure resource points containing location and attribute descriptions, comprising the following steps:

[0021] By traversing the street view points of interest in the city road data package and combining them with the associated building outline data, physical spaces located within a preset distance of the road and whose inward concavity depth and width are not less than the first threshold and the second threshold, respectively, are identified.

[0022] The identified physical spaces are functionally filtered, and spaces of the preset safe location type are retained;

[0023] Create a record for each selected space, including its geographic coordinates and an attribute description consisting of space type and expected capacity, to generate a list of safe resource points.

[0024] A further aspect of the present invention generates a resilient skeleton path from the starting point to the destination, comprising the following steps:

[0025] Based on the road network of urban road data packets, a graph search algorithm is used to calculate multiple initial candidate paths from the origin to the destination;

[0026] For each initial candidate path, sampling points are set at equal intervals along the path line, and the number of safe resource points in the safe resource point list within a preset radius around each sampling point is counted.

[0027] Based on the statistical results of each sampling point, the average number of safe resource points per unit length of each path is calculated as a safe resource point density index.

[0028] The density of safe resource points, path length, and travel time are weighted and comprehensively evaluated, and the path with the best comprehensive evaluation is selected as the resilient skeleton path.

[0029] A further aspect of the present invention generates a user behavior status signal reflecting the user's real-time movement status, comprising the following steps:

[0030] Data streams from the accelerometers and gyroscopes in the inertial measurement unit are acquired at a fixed frequency;

[0031] The acceleration data stream is processed to calculate the user's real-time walking speed and step frequency variation coefficient within a continuous time window;

[0032] When the real-time walking speed is consistently lower than the user's historical average speed by a preset percentage exceeding the first time threshold, or when the step frequency variation coefficient exceeds the preset variation threshold and is accompanied by a small-range rotation movement indicated by gyroscope data, the user is determined to be in a state of obstructed movement.

[0033] Encapsulate the state of being blocked or moving normally into a user behavior state signal.

[0034] A further aspect of the present invention generates qualitative markers for environmental anomalies, comprising the following steps:

[0035] Frame the continuous environmental sound data and extract features, including average volume and energy percentage of specific frequency bands.

[0036] The extracted sound features are matched with a pre-set abnormal sound pattern library, which contains feature templates for continuous mechanical roaring, dense car horn sounds, and large vehicle idling sounds.

[0037] When the matching degree between the sound features of multiple consecutive frames and a certain abnormal sound pattern exceeds the set matching degree threshold, it is determined that there is an abnormality of the corresponding type ahead, and an environmental abnormality qualitative marker is generated.

[0038] A further aspect of the present invention involves calculating and outputting a boot instruction to guide the user to the nearest available safe resource point, comprising the following steps:

[0039] When the alarm triggering conditions are met, obtain the user's current location coordinates;

[0040] The current location coordinates are used as a reference point. The nearest safe resource point that is located in front of the user's direction of travel is found in the list of safe resource points and is selected as the nearest available safe resource point.

[0041] Calculate the straight-line distance and direction from the current location to the nearest available safe resource point, and generate voice guidance instructions based on the attribute description of the safe resource point.

[0042] A further aspect of the present invention includes the following steps in determining the alarm triggering conditions:

[0043] The user behavior status signal is a state of obstructed movement, and the environmental anomaly qualitative marker is not empty.

[0044] A further aspect of the present invention, forming a complete dynamically adjusted navigation path, includes the following steps:

[0045] The splicing process specifically includes smoothing the connection points:

[0046] Select several coordinate points at the end of the actual walking trajectory and several coordinate points at the beginning of the locally updated path segment;

[0047] A transition curve is generated between the actual walking trajectory and the locally updated path segment using a spline interpolation algorithm;

[0048] The actual walking trajectory, transition curve, and locally updated path segments are connected in time sequence to form a complete dynamically adjusted navigation path.

[0049] Secondly, this invention provides a navigation path selection system for the blind that integrates semantic and real-time information, employing the following technical solution:

[0050] A navigation path selection system for the blind that integrates semantics and real-time information includes the following modules:

[0051] The data acquisition and preprocessing module acquires the coordinates of the starting point and destination of the route planning, and extracts road network data and street view points of interest data from the electronic map to generate a city road data package containing basic road attributes and street view points of interest.

[0052] The safety resource point identification and labeling module, based on urban road data packages, identifies and labels physical spaces along roads that conform to preset geometric and functional characteristics, and generates a list of safety resource points containing location and attribute descriptions.

[0053] The global resilient path planning module, based on urban road data packages and a list of safety resource points, performs global path planning with the distribution density of safety resource points along the path as one of the optimization objectives, generating a resilient skeleton path from the origin to the destination.

[0054] The user behavior status analysis module continuously analyzes inertial measurement unit data during navigation along the resilient skeleton path, generating user behavior status signals that reflect the user's real-time travel status.

[0055] The environmental anomaly sound analysis module continuously analyzes environmental sound data during navigation and generates qualitative markers of environmental anomalies by matching it with a preset abnormal sound pattern library.

[0056] The security guidance instruction generation module is used to receive user behavior status signals and environmental anomaly qualitative markers. When both meet the preset alarm triggering conditions, it calculates and outputs guidance instructions to guide to the nearest available security resource point based on the current location and the list of security resource points.

[0057] The local path replanning module allows users to move according to the guidance instructions and confirm that they have arrived at the target safe resource point. Using the safe resource point as the new starting point, the module replans the locally updated path segment to the destination coordinates.

[0058] The seamless path stitching and updating module seamlessly stitches together the user's completed actual walking trajectory with locally updated path segments to form a complete dynamically adjusted navigation path.

[0059] In summary, the present invention has the following beneficial technical effects:

[0060] 1. By identifying and marking safe resource points with specific geometric and functional characteristics in urban road data packages, and introducing the "average number of safe resource points per 100 meters" as a key optimization indicator for weighted comprehensive evaluation during the global route planning stage, this technical feature enables the generated "resilient skeleton routes" to improve the distribution density of safe spaces available for temporary refuge along the route while maintaining traditional traffic efficiency. This effect addresses the deficiency of "lack of safety considerations in route planning" pointed out in the background technology. By incorporating the accessibility of safe resources into the route generation logic in a quantitative manner, it provides users with a trunk route with higher safety redundancy in advance.

[0061] 2. By continuously analyzing inertial measurement unit data, a user behavior status signal reflecting the "impeded movement" state is generated, and environmental sound data is simultaneously analyzed to generate a qualitative marker of environmental anomalies. When both meet preset alarm conditions, a safety guidance process is triggered. This multi-sensor information fusion and collaborative judgment technology enables proactive and timely perception of users potentially encountering movement difficulties due to environmental anomalies. This addresses the shortcomings of the background technology, which was "slow to react to atypical user movement states and environmental risks," upgrading a single position deviation judgment to a comprehensive situational judgment combining behavioral patterns and environmental characteristics, thus improving the accuracy and timeliness of risk identification.

[0062] 3. After guiding the user to a safe resource point, this point is used as the new starting point. A local lightweight algorithm is invoked to replan the subsequent path to the final destination, generating a locally updated path segment, rather than performing a time-consuming global replanning. Subsequently, the user's actual travel trajectory is seamlessly stitched with this locally updated path segment to form a complete adjusted path. This technical feature enables dynamic, rapid local adjustments and continuity maintenance of the navigation path. Attached Figure Description

[0063] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. The drawings are used to provide a further understanding of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0064] Figure 1 A flowchart illustrating an embodiment of this application is disclosed.

[0065] Figure 2 Structural schematic diagrams of embodiments of this application are disclosed. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] The following is in conjunction with the appendix Figures 1-2 A preferred description of the present invention is provided below.

[0068] See attached document Figure 1 This invention proposes a method for selecting navigation paths for the blind that integrates semantic and real-time information, comprising the following steps:

[0069] S1. Obtain the coordinates of the starting point and destination of the route planning, and extract road network data and street view points of interest data from the electronic map to generate a city road data package containing basic road attributes and street view points of interest.

[0070] S2. Based on urban road data packages, identify and label physical spaces along roads that conform to preset geometric and functional characteristics, and generate a list of safe resource points containing location and attribute descriptions.

[0071] S3. Based on urban road data packages and a list of safe resource points, global path planning is performed with the distribution density of safe resource points along the path as one of the optimization objectives, generating a resilient skeleton path from the starting point to the destination.

[0072] S4. During navigation along the resilient skeleton path, continuously analyze the inertial measurement unit data to generate user behavior status signals that reflect the user's real-time travel status.

[0073] S5. During navigation, continuously analyze environmental sound data and generate qualitative markers for environmental anomalies by matching them with a preset abnormal sound pattern library.

[0074] S6. Receive user behavior status signals and environmental anomaly qualitative markers. When both meet the preset alarm triggering conditions, calculate and output a guidance instruction to guide to the nearest available security resource point based on the current location and the list of security resource points.

[0075] S7. After the user moves according to the guidance instructions and confirms that they have arrived at the target safe resource point, the local updated path segment to the destination coordinates is replanned using the safe resource point as the new starting point.

[0076] S8. The user's completed actual walking trajectory is seamlessly spliced ​​with the locally updated path segment to form a complete dynamically adjusted navigation path.

[0077] In one embodiment of the present invention, step S1 includes the following steps:

[0078] First, the system receives route planning requests from users via voice or text input through a navigation application for the blind. Then, it parses the content of the route planning request to obtain the coordinates of the user's starting point and final destination. Next, it sends a request to the online map service API to obtain vector data of all roads within a radius covering the potential route, centered on the line connecting the starting point and destination. The basic road attributes include road name, lane width, sidewalk width, and road surface material.

[0079] The system retrieves point-of-interest (POI) data for the area from the same map service. Street view POIs include the names and locations of shops, public facilities, plaza entrances, and bus stops along the street, generating a city road data package that contains both basic road attributes and street view POI information.

[0080] Specifically, users activate the route planning function through a navigation application for the blind installed on their mobile devices. Users express their travel needs from one location to another via voice or text input. The navigation application's built-in voice recognition or text parsing module receives this input, performs semantic understanding, and extracts information representing the starting and ending locations. The application then calls an integrated online map service interface to convert this location information into latitude and longitude coordinates, which serve as the starting and destination coordinates for route planning, respectively.

[0081] The route planning start and destination coordinates are two pairs of two-dimensional coordinate points composed of longitude and latitude values. These are used to identify the start and end points of a trip in geographic space. The coordinates are determined by the location description input by the user via voice or text, converted using the geocoding function of the online map service. After obtaining the coordinates, the application constructs a data request. By calling the application programming interface provided by the online map service, it requests vector data of all roads within a circular geographic area centered on the straight line connecting the start and destination coordinates, with a preset radius. The length of this preset radius must ensure coverage of all potential reasonable routes from the start to the destination. Based on statistical analysis of typical urban block sizes, this radius can be set to 1.5 times the straight-line distance between the start and destination, with a minimum of 500mm.

[0082] The online map service interface responds to the data request, returning the geometry, topological connections, and associated basic road attribute data for all roads within the circular geographic area. This basic road attribute data includes at least the road name, lane width, sidewalk width, and pavement material classification for each road. The basic road attributes are a set of data fields describing the physical and functional characteristics of the road itself, including: the road name as a text string; lane width as a floating-point number in meters, representing the width of the motor vehicle lane; sidewalk width as a floating-point number in meters, representing the width of the pedestrian crossing area on both sides of the road; and pavement material as a category label, such as "asphalt," "concrete," or "masonry."

[0083] Simultaneously, the application requests and retrieves point-of-interest (POI) data within the same geographic area through the same map service interface. These POIs specifically refer to point-like geographic features distributed along streets that have landmark or potential stopping points significance, collectively referred to as Street View POIs. Their data includes at least the names and latitude / longitude coordinates of street-front shops, the identification and coordinates of public facilities, the location coordinates of the main entrance to the square, and the names and coordinates of bus stops. Street View POIs are a collection of data on point-like geographic entities located along roads that have identifying or potential auxiliary functions, including: street-front shops (commercial storefronts open to the public on both sides of the road, with data including their names and coordinates); public facilities (service points managed by the government or public institutions, such as post offices and public toilets, with data including their type identification and coordinates); square entrances (the location coordinates of the main passage connecting the city square and the road); and bus stops (fixed stops for public transportation vehicles, with data including their route names and coordinates).

[0084] The application integrates and packages the acquired road infrastructure data and its associated road infrastructure data, along with the acquired street view point of interest (POI) data, according to a unified geospatial reference system, to generate a structured data file or data stream, namely the urban road data package. The urban road data package contains road data and POI data linked through shared geospatial coordinates, allowing for convenient querying and analysis of the POI distribution around each road segment.

[0085] Online map services refer to map data services provided by professional map service providers that include detailed road networks and related information.

[0086] For example, a user inputs "navigate from the east entrance of Wangfujing Pedestrian Street to the National Centre for the Performing Arts" into a navigation application for the blind via voice. After parsing, the application performs geocoding through the online map service application programming interface, obtaining the starting point coordinates as approximately 116.417 degrees east longitude and 39.917 degrees north latitude, and the destination coordinates as approximately 116.397 degrees east longitude and 39.909 degrees north latitude, with a straight-line distance of approximately 2 kilometers between the two points.

[0087] The application constructs a request, using the midpoint of the line connecting the two points as the center and a radius of 3000 meters as the radius, to request road and point-of-interest (POI) data within that area from the map service's application programming interface (API). The API returns road data for that area, such as "East Chang'an Avenue," with attributes including a lane width of 3.5 meters, a sidewalk width of 4 meters, and a pavement material of "asphalt." The returned POI data includes coordinates for locations such as "Wangfujing Department Store," "Dongdan Subway Station Exit C," "East Entrance of Tiananmen Square," and "Tiananmen East Bus Station." The application integrates this data to generate a city road data package, in which the "East Chang'an Avenue" road segment and the "Wangfujing Department Store" POI are associated due to their spatial proximity.

[0088] In one embodiment of the present invention, step S2 includes the following steps:

[0089] The system iterates through the street view points of interest in the city road data package and combines them with the associated building outline data to identify physical spaces located within a preset distance from the road and whose inward recess depth and width are not less than the first threshold and the second threshold, respectively. The identified physical spaces are functionally filtered, and spaces of the preset safe place type are retained. A record is created for each space that passes the filter, and the record contains its geographical coordinates and attribute description consisting of space type and expected number of people, generating a list of safe resource points.

[0090] Specifically, the system receives the city road data packet generated in step S1 as input, parses the data packet, extracts all street view points of interest (PHOs) data entries, and processes them one by one. For each PHO, the system obtains the building outline data associated with the PHO based on its geographic coordinates by calling the additional data interface of the online map service or accessing the locally stored building dataset. The building outline data associated with the PHO describes the ground projection shape of the building in vector polygon form. The system calculates the vertical distance from the PHO coordinates to its nearest road edge. This distance is calculated by drawing perpendicular lines from the PHO coordinates to all road segments in the city road data packet and taking the shortest perpendicular line length. The preset distance range for roads is... meters, if the shortest distance is less than or equal to Meters, determining that the point of interest is located within a preset distance range of the road.

[0091] For points of interest that meet the distance criteria, the system analyzes the associated building outline polygons. The system identifies whether there are recessed sections in the building outlines facing the road, i.e., geometric shapes formed by continuous line segments within the polygon boundaries that recede inwards. For each recessed area, the system measures two key dimensions: the depth and width of the recess. The depth of the recess refers to the straight-line distance from the road edge along the perpendicular direction to the innermost point of the recessed area, expressed in meters. It represents the straight-line distance from the road boundary to the innermost point of the space, and is based on field observation data of the minimum setback space required for temporary shelter for the blind, ensuring effective shielding. The width refers to the span of the opening of the recessed area along the road direction, i.e., the straight-line distance between the two boundary points of the recessed area, expressed in meters. It represents the lateral dimension of the opening, ensuring the minimum passage width that allows at least a blind person and their guide dog or assistant to enter simultaneously.

[0092] The system sets a recognition threshold; recognition only occurs when the measured indentation depth is not less than a first threshold. Meters, and the width value is not less than the second threshold, i.e. Only when the distance reaches a certain threshold can the recessed area be identified as a physical space conforming to preset geometric features. After geometric recognition, the system performs functional filtering on the identified physical spaces. Based on the original type label of the street view point of interest, combined with the detailed semantic classification information of the point of interest obtained from the map service, the system determines the functional type of the physical space and compares the space type with a preset list of reserved types. This list includes convenience store entrances, bank entrances, police station entrances, community guard booths, and fixed street seating areas. Only when the functional type of a physical space completely matches any item in this list is the space reserved.

[0093] For all physical spaces that pass the functional screening, the system creates an independent data record for each space. This record contains two core fields: geographic coordinates and attribute description. The geographic coordinates are typically taken as the coordinates of the geometric center point of the recessed area of ​​the physical space, obtained by calculating the average of the polygon vertices. The attribute description field consists of two sub-items: space type and estimated capacity. The space type is directly derived from the screening results. The estimated capacity is estimated based on the dimensions of the physical space, using a preset mapping relationship. For example, for a space with a recessed depth of not less than 1.5 meters and a width of not less than 2 meters, the basic capacity is set to 2 people; for every 1 meter increase in width, the capacity increases by 1 person.

[0094] All records are organized according to a uniform field order and data structure, and the output is an ordered list or table file, namely the list of safe resource points. The list of safe resource points is a structured dataset, and the data structure of each record includes a coordinate field and a text description field, which is used for fast retrieval in the subsequent path planning module.

[0095] Physical space refers to an area with a clear three-dimensional boundary formed by the recesses of building facades or the layout of street facilities. The data structure of physical space is defined by a set of geographical boundary points and functional labels.

[0096] Functional screening is a logical judgment process. The basis for functional screening is the social functional attributes of the space. The designated types of spaces that are retained are based on the fact that these spaces usually have good accessibility, lighting conditions, or staffing characteristics, and can provide temporary safe places for blind people to stay.

[0097] The preset distance range for roads refers to the vertical distance threshold from the street view point of interest to the nearest road edge. It is used to initially screen potential spaces near roads. The setting of this parameter is mainly based on the accessibility needs of blind people in emergency avoidance. That is, when a sudden danger occurs, the user needs to be able to enter the roadside sheltered space in a short time. It usually refers to urban pedestrian emergency behavior research and accessibility design specifications. The value range is 5 meters to 30 meters, and the typical value can be set to 10 meters to balance space availability and recognition efficiency.

[0098] The first threshold corresponds to the "inward recess depth" of the physical space, which is the straight-line distance from the road edge to the innermost point of the recessed area. Its setting is based on on-site observations and human-scale analysis of the minimum safe space required for temporary shelter for blind people, ensuring it can accommodate one person and their belongings while providing effective shelter. The value typically ranges from 1 meter to 3 meters, with a typical value of 1.5 meters, to meet basic retreat and protection requirements.

[0099] The second threshold corresponds to the "width" of the physical space, which is the opening span of the recessed area along the road. When setting it, it is necessary to ensure that the space allows at least a blind person and their guide dog or assistant to enter at the same time. Referring to the minimum width standard for barrier-free access and the need for multiple people to walk side by side, the value range is 1.5 meters to 4 meters, with a typical value of 2 meters, to ensure the feasibility of basic passage and temporary stop.

[0100] Space type refers to the category of venues retained through functional filtering. It is set based on the security characteristics typically present in such venues, including staff presence, surveillance coverage, long opening hours, or emergency assistance facilities. The preset list includes typical security-related venues such as convenience store entrances, bank entrances, police station entrances, neighborhood guard posts, and fixed street seating areas. This parameter is a discrete enumeration value, and filtering must match the type in the list.

[0101] The estimated capacity is a temporary shelter capacity estimated based on the geometric dimensions of the physical space, namely the depth and width. The setting is based on the standard area occupied per person, usually 0.5-1 square meters per person, and the characteristics of the spatial layout. A mapping rule is used to calculate it. For example, when the depth is ≥1.5 meters and the width is ≥2 meters, the basic capacity is 2 people. For every 1 meter increase in width, the capacity increases by 1 person. This parameter is an integer value, ranging from 1 to 10 people. The specific value is dynamically calculated according to the space size.

[0102] For example, based on the urban road data package generated in step S1, which contains a point of interest "China Construction Bank Dongdan Branch", the coordinates of this point are approximately 116.418 degrees east longitude and 39.916 degrees north latitude. The associated building outline data is obtained as a polygon, and analysis reveals a recessed area facing "East Chang'an Avenue". The vertical distance from this point of interest to the edge of "East Chang'an Avenue" is calculated to be 2.2 meters, satisfying the condition of being within 3 meters. The recessed area is measured to have an inward depth of 1.8 meters and a width of 2.5 meters, both satisfying the thresholds of not less than 1.5 meters and not less than 2 meters, respectively.

[0103] The semantic type of this point of interest is "bank," and its physical space function type is determined to be "bank entrance." This type is in the preset reserved type list, so it passes the functional filtering. The system creates a record for this space, taking the center point coordinates of the recessed area as approximately 116.4181 degrees east longitude and 39.9161 meters north latitude. Based on the size estimation, the width is 2.5 meters, with a basic capacity of 2 people. Increasing the width by 0.5 meters does not reach a 1-meter increment, so the expected capacity remains 2 people. The attribute description is recorded as space type "bank entrance" and expected capacity of 2 people. This record is added to the safe resource point list. For the point of interest "Wangfujing Department Store," its type is "department store." Although the physical space may be identified, "department store entrance" does not match the preset reserved type, so it is filtered out in this example and no record is generated. The final generated safe resource point list contains multiple similar records, such as "bank entrance" and "convenience store entrance," etc.

[0104] In one embodiment of the present invention, step S3 includes the following steps:

[0105] Based on the road network in the urban road data package, a graph search algorithm is used to calculate multiple initial candidate paths from the origin to the destination. For each initial candidate path, sampling points are set at equal intervals along the path line, and the number of safe resource points in the safe resource point list within a preset radius around each sampling point is counted.

[0106] Based on the statistical results of each sampling point, the average number of safe resource points per unit length of each path is calculated as the safe resource point density index. The safe resource point density index is then weighted and comprehensively evaluated with the path length index and the travel time index, and the path with the best comprehensive evaluation is selected as the resilient skeleton path.

[0107] Specifically, the system receives the generated urban road data packet and the generated list of safe resource points as input, initiates global path planning, and extracts the road network topology from the urban road data packet. Road intersections are treated as nodes, and road segments as weighted edges, with the initial weights set according to the road length. The system uses a graph search algorithm, such as Dijkstra's algorithm or A algorithm, with the starting coordinates as the search starting node and the destination coordinates as the search target node. It calculates and outputs multiple initial candidate paths from the starting point to the destination in the road network graph. The graph search algorithm generates multiple differentiated initial candidate paths by setting different heuristic functions or introducing slight random perturbations. The initial candidate paths are multiple feasible routes calculated in the road network data by the graph search algorithm, and their generation depends on the topological connectivity of the road network.

[0108] For each initial candidate path, a security resource coverage density analysis is performed. Specifically, sampling points are set at equal intervals along the geometric centerline of the path from the starting point to the endpoint. For each sampling point, a circular buffer area with a preset radius is constructed centered on its coordinates. The system queries the security resource point list generated in step S2, retrieves and counts the number of security resource points whose location coordinates fall within this circular buffer area. After traversing all sampling points on the path, the system obtains the sequence of security resource point counts for that path.

[0109] The system calculates the "average number of safe resource points per 100 meters" index for this route, satisfying the following formula: D represents the average number of safe resource points per 100 meters along the path, used to quantify the distribution density of safe resources along the path. This value is set by transforming discrete sampling point statistics into comparable indicators over a continuous length. N represents the total number of safe resource points within the buffer zones of all sampling points along the path, obtained through summation. M represents the total number of sampling points, obtained by dividing the total path length by the fixed sampling interval and rounding up. L represents the sampling point spacing, typically set to 10 meters. The 10-meter interval is chosen to balance computational accuracy and efficiency at an urban scale.

[0110] The system performs a multi-objective comprehensive evaluation of each candidate path, assigning weights to three key planning optimization indicators: path length (L), travel time (T), and the calculated safe resource point density (D). Since the indicators have different dimensions, a normalization process is first performed. For the set of indicator values ​​for all candidate paths, a linear normalization method is used: I represents the original index value of a certain path. and These are the maximum and minimum values ​​of the indicator across all candidate paths. After normalization, each indicator value is mapped to the interval between 0 and 1. The system has preset weight coefficients. ,satisfy . These are weighting coefficients for path length, travel time, and safety resource point density, respectively. The values ​​can be configured based on the priority given to navigation safety and efficiency. The weighting coefficients can be set based on expert experience or user preferences; for example, in navigation scenarios for the blind, the weighting coefficients can be set... It is 0.5. and Each is set at 0.25 to emphasize safety.

[0111] The formula for calculating the overall evaluation score S for each path is:

[0112]

[0113] in, and These are normalized path length and travel time metrics, typically used to achieve shorter paths and less travel time. and To make it positive; This is a normalized safety resource point density index; a higher value indicates better safety coverage. The system compares the comprehensive evaluation score S of all candidate paths and selects the path with the highest score as the final recommended resilient backbone path. The final recommended resilient backbone path constitutes the main traffic route connecting the origin and destination, achieving an optimal balance between traditional efficiency indicators and the spatial distribution density of safety resources.

[0114] The sampling points are geographical locations selected at equal intervals along the path centerline for spatial statistical analysis. The average number of safe resource points per 100 meters is a derived quantitative indicator used to characterize the abundance of safe refuge spaces along the path; its calculation logic ensures comparability between paths of different lengths. The planning optimization index is a set of quantitative parameters used to evaluate and compare the merits of different paths, specifically referring to path length, travel time, and the density of safe resource points.

[0115] Weighted comprehensive evaluation is a decision-making process that linearly weights and sums multiple normalized planning optimization indicators according to preset weights to obtain a single comprehensive score. The resilient skeleton path is a recommended path selected through multi-objective optimization, which emphasizes the density of safety resource coverage along the route while ensuring basic traffic efficiency, and serves as the benchmark for subsequent navigation execution and dynamic adjustment.

[0116] For example, starting from the east entrance of Wangfujing Pedestrian Street and ending at the National Centre for the Performing Arts, the A algorithm is used to calculate two initial candidate paths:

[0117] Route A is 2000 meters long and the estimated travel time is 30 minutes.

[0118] Route B is 2100 meters long and the estimated travel time is 28 minutes.

[0119] Sampling points are set every 10 meters along path A, for a total of 200 points. The number of safe resource points within a 25-meter buffer zone for each point is counted. Assume the total number of points is... Calculate the density index of 60 samples. Each hundred meters. Perform the same calculation for path B, assuming a total... There are 90 samples, with approximately 210 sampling points, and the density index... Each hundred meters.

[0120] After normalizing the metrics for all candidate paths, assuming the shortest path length is 2000 meters and the longest is 2100 meters, then path A's... The value is 0, for path B. The value is 1. Travel time indicator, minimum 28 minutes, maximum 30 minutes, for route A. The value is 1, for path B. The value is 0. The minimum safety density index is 3.0, and the maximum is 4.29. Path A... The value is 0, for path B. The value is 1.

[0121] Use example weights ,but:

[0122] Overall score for path A ,

[0123] Overall score for path B .

[0124] Path B scored higher and was therefore selected as the toughness skeleton path.

[0125] In one embodiment of the present invention, step S4 includes the following steps:

[0126] The system acquires accelerometer and gyroscope data streams from the inertial measurement unit at a fixed frequency; processes the acceleration data streams to calculate the user's real-time walking speed and step frequency variation coefficient within a continuous time window; when the real-time walking speed is consistently lower than the user's historical average speed by a preset proportion exceeding a first time threshold, or when the step frequency variation coefficient exceeds a preset variation threshold and is accompanied by a small-range rotational movement indicated by the gyroscope data, the system determines that the user is in a state of obstructed movement; and encapsulates the obstructed movement state or normal movement state into a user behavior state signal.

[0127] Specifically, after the user begins navigation based on the generated resilient skeleton path, the system starts a background service thread on the user's mobile terminal. This background service thread continuously monitors the real-time data output of the phone's built-in inertial measurement unit (IMU). The system synchronously reads the raw data streams generated by the triaxial accelerometer and triaxial gyroscope from the IMU at a fixed sampling frequency, such as 50Hz. The triaxial accelerometer data is a sequence of instantaneous acceleration values ​​in three orthogonal directions, measured in m / s². The triaxial gyroscope data is a sequence of instantaneous angular velocities in three axes, measured in rad / s.

[0128] The acceleration data is preprocessed, including using a low-pass digital filter to eliminate high-frequency noise and subtracting the gravitational acceleration component to obtain a linear acceleration reflecting the user's body movement. The system performs gait event detection on the processed acceleration data within a short, continuously sliding time window, for example, a window length of 2 seconds. By identifying periodic peaks in the acceleration magnitude sequence, the system identifies the user's stride and records the time interval of each gait cycle. Based on a pre-calibrated stride length model or one estimated based on the user's height, combined with the detected real-time stride frequency, the speed value is calculated using the following formula: , Real-time walking speed, used to reflect the user's macroscopic movement speed; The estimated step size is set to a fixed value based on statistical experience, such as 0.7 meters, or according to the user's height h using a formula. Dynamic calculations can be based on average stride length data from demographics or estimated in conjunction with user height. f represents the real-time stride frequency in Hz, obtained by taking the reciprocal of the gait event time intervals derived from acceleration signal analysis.

[0129] The system collects a series of continuous instantaneous step frequency values ​​within an analysis time window, such as the past 30 seconds. Calculate the average step frequency within this window: , The average step frequency within the analysis window is used to characterize the user's typical step frequency level. Then, the standard deviation of the step frequency is calculated: , The standard deviation of the step frequency within the analysis window is used to measure the dispersion of the step frequency. Step frequency variation coefficient. The calculation formula is: , The step frequency variation coefficient is set based on experimental observation data of step frequency stability during normal walking and hesitant walking.

[0130] The real-time walking speed v and step frequency variation coefficient were obtained through continuous monitoring and calculation. The system maintains the user's historical average walking speed. It is obtained by collecting speed data and averaging it during multiple normal navigation cycles by the user. Two conditions for parallel decision-making are set:

[0131] Condition 1 is that the real-time walking speed v is consistently lower than the historical average speed. 50%, that is, satisfying And this state is maintained for more than 10 seconds.

[0132] Condition 2 is the step frequency variation coefficient. If the variability exceeds a preset threshold, such as 20%, and within the same analysis time window, the gyroscope data indicates that the user has a rotational motion, the determination of the rotational motion is based on the yaw angle change obtained by integrating the angular velocity of the gyroscope around the vertical axis. Within a short period, such as 5 seconds, the yaw angle change exhibits an oscillation pattern of first increasing in the positive direction and then decreasing in the negative direction, and the total angle change amplitude is between 30 degrees and 120 degrees.

[0133] If either condition one or condition two is met, the user is determined to be in a "travel obstructed" state. If neither condition is met, the user is determined to be in a "normal travel" state. The system encapsulates the current state determination result into a discrete logical signal, namely the user behavior state signal, which can be a data structure containing state enumeration values ​​and timestamps.

[0134] The inertial measurement unit (IMU) is a miniature electronic sensor assembly integrated into the mobile terminal, used to measure the acceleration and rotational angular velocity of an object. The triaxial accelerometer is the component within the IMU used to measure linear acceleration, outputting a sequence of acceleration data along three axes. The gyroscope is the component within the IMU used to measure angular velocity, outputting a sequence of angular velocity data along three axes. A data stream refers to the collection of time-series data continuously generated by sensors at fixed time intervals. Real-time walking speed is a dynamically changing scalar value representing the user's current average walking rate. The coefficient of variation (COP) measures the stability of the user's gait rhythm; a higher value indicates a more disordered gait.

[0135] The "obstructed progress" state is one of the logical states determined by the system, indicating that the user is unable to proceed normally along the expected path due to obstacles, confusion, or danger. The "normal progress" state is another logical state determined by the system, indicating that the user's progress is in line with expectations. The user behavior state signal is a logical variable that represents the user's real-time progress state, and its data structure includes state type and time information.

[0136] For example, a user is walking along the selected resilient skeleton path B. The system backend collects data from the inertial measurement unit at a frequency of 50 Hz. At a certain time period, the user's real-time walking speed is calculated. The speed is 0.4 m / s. The system queries the user's historical average walking speed. It is 1.0 m / s. Because... The condition is met, and this low-speed state has lasted for 15 seconds, satisfying condition one. The system calculates the step frequency data over the past 30 seconds to obtain the average step frequency. 1.0 Hz, standard deviation If the frequency is 0.15Hz, then the step frequency variation coefficient is... The variation threshold of 20% was not exceeded, and the gyroscope data showed a stable reading with no gyroscope rotation.

[0137] Although condition two is not met, condition one has been triggered. The system determines that the user is currently in a "progress blocked" state and generates a corresponding user behavior status signal output. The signal content is "Status: Progress blocked; Timestamp: 2024-10-27 10:05:15".

[0138] In one embodiment of the present invention, step S5 includes the following steps:

[0139] The system performs frame segmentation and feature extraction on continuous ambient sound data collected by the microphone. The extracted features include average volume and energy percentage of specific frequency bands. The extracted sound features are matched with a preset abnormal sound pattern library, which contains feature templates for continuous mechanical roaring, dense car horn sounds, and large vehicle idling sounds. When the matching degree between the sound features of multiple consecutive frames and a certain abnormal sound pattern exceeds a set matching degree threshold, it is determined that there is an abnormality of the corresponding type ahead, such as "abnormal construction noise" or "abnormal traffic congestion". The determined abnormality type is output as a qualitative marker of environmental abnormality.

[0140] Specifically, during the user's navigation along the generated resilient skeleton path, the system synchronously starts an audio processing thread on the user's mobile terminal. This thread continuously captures continuous ambient sound analog signals from the phone's built-in microphone. First, it uses an analog-to-digital converter to convert the analog audio signals into discrete digital audio data streams at a predetermined sampling rate, such as 16,000 samples per second. The continuous audio data stream is then processed by frame segmentation, dividing the data into a series of short analysis units, for example, each frame is 20ms long, with a 10ms overlap between adjacent frames. For each frame of audio data, the system applies a window function, such as a Hamming window, to reduce spectral leakage. Finally, a Fast Fourier Transform is performed on the windowed frame data to transform it from the time domain to the frequency domain, obtaining the spectrum of that frame.

[0141] Two key acoustic features are extracted from the spectrum. The first feature is the average volume, which is calculated by first calculating the root mean square of the temporal sample values ​​of the audio data frame, satisfying the formula: L represents the average volume of the frame, a dimensionless relative amplitude value used to measure the overall loudness level of ambient sound, obtained by calculating the root mean square of the samples within the frame. N is the number of audio samples per frame, determined by the sampling rate and frame duration, such as... One sample. This represents the quantized amplitude value of the i-th sample point.

[0142] The second feature is the energy percentage of a specific frequency band. At least two frequency bands of interest are preset. For example, band one is 0Hz to 500Hz, used to capture low-pitched, continuous mechanical roaring sounds; band two is 1000Hz to 3000Hz, used to capture sharp car horns. For each frame, the total energy within that frequency band, E_band, and the total energy across the entire audible spectrum, E_total, are calculated. The formula for the energy percentage P of a specific frequency band is: P = (E_band / E_total) * 100%; P represents the energy percentage of a specific frequency band, used to identify the spectral characteristics of the sound. The setting is based on the fact that different types of abnormal noise have significant energy concentration areas in the spectrum; for example, construction noise energy is concentrated in the low frequencies.

[0143] E_band represents the total energy within a specified frequency band, obtained by summing the squares of the amplitudes of all frequency components within that band. E_total represents the total energy of the signal, obtained by summing the squares of the amplitudes of all frequency components. After feature extraction, the system performs real-time matching between the feature vector containing the average volume L and the energy percentage P of a specific frequency band and a pre-set abnormal sound pattern library. The abnormal sound pattern library stores feature vector templates of various typical environmental abnormal sounds and their corresponding semantic labels. It is pre-constructed by collecting and labeling a large number of typical abnormal environmental sound samples. Each record in the abnormal sound pattern library is a feature template associated with an abnormal type. For example, the "continuous mechanical roaring sound" template is characterized by a high average volume and a sustained energy percentage of over 60% in the 0-500Hz frequency band; the "dense car horn sound" template is characterized by a moderate average volume but intermittent spikes in the energy percentage of the 1000-3000Hz frequency band.

[0144] The matching process is achieved by calculating the similarity between the input feature vector and the feature vectors of each template. For example, the reciprocal of the Euclidean distance or the cosine similarity can be used as the matching score. The system sets a matching score threshold, for example, 0.8. This threshold is based on a balance point determined from a large amount of test data to achieve an acceptable trade-off between the false positive rate and the false negative rate. When the system detects multiple consecutive frames where the matching score with a certain anomalous sound template exceeds the matching score threshold, it determines that there is an anomaly of the corresponding type in the foreground environment. The system encapsulates the determined anomaly type, such as "abnormal construction noise" or "abnormal traffic congestion," into a text label, i.e., an environmental anomaly qualitative marker. This label contains the anomaly type and the time information of the detection.

[0145] Continuous ambient sound data refers to the continuous and digitized ambient audio signal stream collected by a mobile phone microphone. Framing is a fundamental step in dividing the continuous audio stream into short, fixed-duration segments for processing to analyze the short-term stationary characteristics of the signal. Feature extraction is the process of calculating quantitative parameters characterizing the acoustic properties of audio frames, specifically the average volume and the energy proportion of a particular frequency band. Matching degree indicates the closeness between the extracted sound features and template features in the library, used to determine the type of anomaly. Environmental anomaly qualitative labeling is the result of semantic classification of abnormal conditions in the foreground environment; its data structure includes an anomaly type string.

[0146] For example, when a user navigates along path B to a certain section of the road, the system's audio processing thread starts working, and the microphone picks up a continuous ambient sound. The system processes this sound frame by frame and analyzes five consecutive frames.

[0147] The average volume L for these 5 frames was calculated to be 0.7, 0.72, 0.71, 0.69, and 0.7, respectively, and the energy percentage P in the 0-500Hz frequency band was calculated to be 68%, 70%, 65%, 72%, and 67%, respectively. These feature vectors were matched with templates in the abnormal sound pattern library. The matching scores with the "continuous mechanical roaring sound" template were 0.85, 0.87, 0.83, 0.88, and 0.84, respectively, all exceeding the matching score threshold of 0.8. However, the matching scores with the "dense car horn sound" template were all below 0.3. The system determined that there was "abnormal construction noise" ahead, and immediately generated and output an environmental anomaly qualitative marker with the content: "Anomaly type: abnormal construction noise; timestamp: 2024-10-27 10:05:30".

[0148] In one embodiment of the present invention, step S6 includes the following steps:

[0149] It receives generated user behavior status signals and generated environmental anomaly qualitative markers. When both meet the preset alarm triggering conditions, it calculates and outputs the nearest available security resource point guidance instruction based on the user's current location and the generated security resource point list.

[0150] The alarm trigger condition is set as follows: the user behavior status signal is "obstructed" and the environmental anomaly qualitative marker is not empty. When the generated user behavior status signal and the generated environmental anomaly qualitative marker meet the above alarm trigger condition, the provided current position coordinates are obtained. Using the current position coordinates as the center, the nearest safe resource point located in front of the user's direction of travel is found in the generated list of safe resource points. The straight-line distance and rough direction from the current position to the nearest safe resource point are calculated, and a voice guidance command containing distance, direction and target safe resource point attribute description is generated.

[0151] Specifically, the system receives and listens to the user behavior status signal output from step S4 and the environmental anomaly qualitative marker output from step S5, and presets an alarm trigger conditions based on logical judgment. These conditions require the simultaneous fulfillment of two sub-conditions: first, the content of the user behavior status signal must be the string "Movement Blocked"; second, the environmental anomaly qualitative marker must be non-empty, meaning its contained anomaly type string must not be null. When the system detects that the data from both inputs simultaneously meets the above conditions, it triggers the safety boot process. After triggering, the system calls the mobile terminal's built-in GPS module to obtain the module's current real-time geographic location coordinates, represented as floating-point numbers of longitude and latitude.

[0152] Using the current location coordinates as the center or reference point, the system performs spatial querying and filtering on the list of safe resource points generated in step S2. The system needs to estimate the user's current direction of travel, which is estimated by analyzing a sequence of continuous location points provided by the Global Positioning System (GPS) over a recent period, such as the past 10 seconds. The earliest and last points in the sequence are used to construct a direction vector, and the direction of this vector represents the user's approximate direction of travel. The system iterates through each record in the list of safe resource points.

[0153] For each record, the system performs the following calculations: First, it calculates the straight-line distance from the current location coordinates to the coordinates of the safe resource point. Since the coordinates use latitude and longitude, a planar approximation formula can be used to calculate the distance within a small geographical area. The formula is:

[0154]

[0155] Where d is the straight-line distance, which is a measure used to quantify spatial proximity. Its calculation is based on the difference in latitude and longitude and the geographical conversion factor. and These are the differences in longitude and latitude, obtained by subtracting the coordinates of the two points. and This is a scaling factor that converts latitude and longitude differences into meters. Its value depends on the geographical location and is set based on the Earth ellipsoid model. It is an approximation for a specific latitude and is used to convert angular differences into linear distances, such as around 39 degrees latitude. Take 111,000 meters per degree. Desirable Meters per degree.

[0156] The second step is to determine whether the safe resource point is located to the side or front of the user's direction of travel. The system constructs two vectors: vector A is the previously estimated user direction of travel vector; vector B is the vector pointing from the current real-time position to the coordinates of the safe resource point. The angle θ between the two vectors is calculated, which can be obtained using the dot product formula. A·B is the dot product of vectors A and B, where |A| and |B| are the magnitudes of vectors A and B, respectively.

[0157] If the absolute value of the calculated angle θ is less than or equal to 45 degrees, the safe resource point is determined to be located to the side and front. A 45-degree angle threshold is set to ensure that the target point is in a frontal and lateral area that is easily perceived and navigable by the user. From all safe resource points that meet the side-front condition, the one with the smallest straight-line distance d is selected as the nearest available safe resource point. The system generates a voice guidance command, which must include three elements: movement direction, approximate distance, and a description of the target point's attributes. The movement direction is determined based on the target point's position relative to the direction of travel. If the target point vector is to the right of the direction of travel vector, the direction is "right front"; otherwise, it is "left front".

[0158] The approximate distance is the calculated straight-line distance d, rounded to the nearest multiple of 5 meters. The target point attribute description is directly obtained from the attribute description field recorded in the list of safe resource points, such as "convenience store entrance". The final generated instruction text is, for example: "Please move about 15 meters to your right front and stop at the nearest convenience store entrance". This text is converted into a speech signal by the mobile terminal's text-to-speech engine and played to the user.

[0159] The alarm trigger condition is a judgment rule composed of two Boolean expressions connected by a logical AND. It is set based on the premise that an emergency requiring active guidance to a safe point is only determined when the user's movement is obstructed and an abnormal environment exists. The current location coordinates are real-time, timestamped geographic latitude and longitude data provided by the Global Positioning System (GPS) module. The nearest available safe resource point refers to a safe resource point record that meets preset filtering criteria in both spatial distance and orientation (closest and located to the side and front). The rough direction is a qualitative directional description derived from vector angle analysis, used to generate directional guidance that the user can understand. The voice guidance command is a structured natural language string used to convey clear action commands to the user via voice, containing distance, direction, and target point descriptions.

[0160] For example, the system receives a user behavior status signal of "obstructed movement" and an environmental anomaly qualitative label of "abnormal construction noise." Both are not empty and their statuses match, meeting the alarm triggering conditions. The current location coordinates are obtained from the Global Positioning System as 116.4185 degrees East longitude and 39.9165 degrees North latitude. Assuming the user's direction of travel is estimated to be due north based on the nearest trajectory, the system queries the list of safe resource points generated in step S2. The list contains a record with the spatial type "convenience store entrance," coordinates 116.4188 degrees East longitude and 39.9170 degrees North latitude. The latitude and longitude difference is calculated as follows: = 0.0003 degrees, = 0.0005 degrees.

[0161] Pick =111,000 meters per degree =86000 meters per degree, calculate the straight-line distance. The distance is approximately 60 meters after rounding. Construct a vector B pointing from the current position to the target point, and calculate the angle θ between it and the true north vector. The calculated angle θ is approximately 30 degrees, less than 45 degrees, satisfying the side-front condition. Because other points in the list are farther away or have different directions, this point is determined as the nearest available safe resource point. The target point is located to the right front of the direction of travel, and a voice guidance command is generated: "Please move approximately 60 meters to your right front to reach the nearest convenience store entrance and stop temporarily."

[0162] In one embodiment of the present invention, step S7 includes the following steps:

[0163] As the user moves according to the guidance instructions, the system continuously compares the user's GPS location with the coordinates of the target safe resource point. When the distance between the two is less than a set distance threshold, or when the mobile phone receives the identification signal from the Bluetooth beacon deployed at the safe resource point, it confirms that the user has arrived. The coordinates of the confirmed safe resource point are used as the starting point for a new route planning, and the coordinates of the obtained final destination are used as the unchanging target endpoint. The local lightweight route planning algorithm is invoked to recalculate a new route from the current safe resource point to the final destination based on the generated urban road data packets. However, this calculation only generates subsequent road segments and does not change the historical routes that have already been traveled. The calculated subsequent new route segments are marked as locally updated route segments.

[0164] Specifically, after starting to move towards the target security resource point according to the voice guidance command, the system initiates a location arrival confirmation monitoring process. This process acquires the real-time location coordinates provided by the user's mobile terminal's GPS module at a high frequency, for example, once per second, and reads the coordinates of the target security resource point determined in step S6. The system continuously calculates the real-time straight-line distance between the two points, using the same planar approximation formula as in step S6. It then initiates a scan for Bluetooth Low Energy signals to listen for the wireless signal broadcast by the Bluetooth beacon deployed at the specific security resource point. When any of the following conditions are met, the system confirms that the user has arrived at the target security resource point:

[0165] Condition 1: The calculated real-time straight-line distance is consistently less than a set distance threshold, such as 3 meters, and this state is maintained for more than 5 seconds to filter out positioning jitter. Condition 2: The mobile terminal scans and parses a Bluetooth beacon signal, and the identification code carried in the signal is completely consistent with the beacon identification code pre-bound to the target security resource point in the list. Once arrival is confirmed, the system records the coordinates of this security resource point as the new path planning starting point. The final destination coordinates obtained in step S1 remain unchanged and are still used as the ultimate target endpoint of the path planning.

[0166] The system invokes a locally running lightweight path planning algorithm instance. This algorithm, based on the road network described by the city road data package generated in step S1 and cached locally, re-executes path calculation with a new starting point and an unchanged ending point as input. The core constraint of this calculation is that the planning scope is limited to subsequent road segments from the current safe resource point to the final destination; the algorithm must not recalculate or modify the historical trajectory already traveled by the user from the original starting point to the current safe resource point. This lightweight algorithm can be a simplified Dijkstra's algorithm or a Focus-D search algorithm, with optimization objectives focusing on path length or turning complexity, but without repeating the complex global optimization involving safe resource point density in step S3. After calculation, the algorithm outputs a continuous path sequence from the new starting point to the final destination.

[0167] This path sequence is encapsulated and marked as a locally updated path segment, which is an independent data structure containing a series of ordered geographic coordinates to describe the subsequent recommended walking route starting from the current safe resource point.

[0168] Location matching confirmation is a process of continuously comparing the user's real-time location with the target location or verifying wireless signals to determine whether the user has arrived at the predetermined location. The Bluetooth beacon identification signal is a wireless radio frequency signal containing a unique identifier, periodically broadcast by Bluetooth Low Energy devices pre-deployed in physical locations. The new path planning starting point is set after confirming the user's arrival at a safe resource point; the coordinates of this point are then set as the starting point for subsequent path calculations, replacing the original starting point coordinates. The local lightweight path planning algorithm is a path search algorithm that runs locally on the mobile terminal, consumes low computational resources, and focuses on fast point-to-point path solving. The locally updated path segment is routing data generated by the path planning algorithm, covering only the portion from the currently confirmed location to the final destination; its data structure is an ordered list of coordinate points.

[0169] For example, a user begins moving according to the instruction "Please move approximately 60 meters to your right front and pause at the nearest convenience store entrance." The target safe resource point is the "convenience store entrance," and its coordinates have been recorded. During the movement, the distance between the user's GPS location and the coordinates of this point is calculated every second. When the user approaches the convenience store entrance, the distance values ​​calculated for five consecutive times are 2.8m, 2.5m, 2.2m, 2.0m, and 1.9m, all less than the 3m threshold, thus satisfying condition one.

[0170] The system confirms the user's arrival and sets the coordinates of the "convenience store entrance" (116.4188°E, 39.9170°N) as the new starting point for route planning. The final destination remains the National Centre for the Performing Arts. The system invokes a local lightweight route planning algorithm, based on cached Beijing urban road data, to calculate a new route from the convenience store entrance to the National Centre for the Performing Arts. The algorithm's output path does not include the section the user has already traversed from the east entrance of Wangfujing Pedestrian Street to this convenience store entrance; it only includes the subsequent direction after starting from this point, such as "walk 100 meters east along the auxiliary road of East Chang'an Avenue, turn right onto Nanheyan Street..." This new route segment is encapsulated as a partially updated route segment.

[0171] In one embodiment of the present invention, step S8 includes the following steps:

[0172] The actual trajectory of the user from the starting point to the current safe resource point is smoothed by connecting the beginning and end of the generated local update path segment. The processed actual trajectory and the local update path segment are then spliced ​​together in chronological order to form a complete new path that starts from the original starting point, passes through the safe resource point, and finally reaches the destination.

[0173] The new complete path is set as the currently active navigation path, updating the path display and voice broadcast logic in the navigation application, guiding the user to continue towards the destination from the current safe resource point along the locally updated path segment.

[0174] Specifically, the system receives the locally updated path segment output in step S7 and simultaneously accesses the historical coordinate point sequence stored locally on the mobile terminal, which records the user's actual walking trajectory from the starting point of the resilient skeleton path in step S3 to the current safe resource point. This historical trajectory is generated by the GPS module at fixed intervals during navigation. The system first performs connection point smoothing processing, which targets the connection between the last coordinate point of the historical trajectory and the first coordinate point of the locally updated path segment. Due to positioning errors or path geometric discontinuities, direct connections may result in sharp angle turns in the path.

[0175] The smoothing process involves fitting a smooth curve between several points at the end of the historical trajectory and several points at the beginning of the locally updated path segment (e.g., three points each). For example, a transition curve is generated using cubic spline interpolation. This transition curve connects the end of the historical trajectory and the beginning of the locally updated path segment, ensuring the path is geometrically continuous and its direction changes smoothly. After smoothing, the system concatenates the processed actual walking trajectory coordinate sequence with the locally updated path segment coordinate sequence in chronological and spatial order. The resulting complete new path is a single, ordered list of geographic coordinate points. This list describes the trajectory from the original planned starting point, through the user's actual walking path to the safe resource point, and then connects to the subsequent recommended path replanned from the safe resource point to the final destination.

[0176] The system sets this new, complete route as the only active navigation route within the navigation application, updates the application's route display logic, and highlights this new route on the map interface, replacing the previously displayed rigid skeleton route. The voice prompt logic is also updated, recalculating and generating subsequent turn-by-turn voice commands based on the geometry and key turn points of the new route. The navigation state is seamlessly switched, and the system begins guiding the user from their current safe resource point location, following the route described by the locally updated route segment, towards their final destination.

[0177] Seamless stitching refers to the data processing procedure of geometrically and logically merging two path segments into a continuous, uninterrupted path. Connection point smoothing is a curve-fitting operation performed to eliminate geometric abrupt changes at path connections, ensuring the synthesized path conforms to the actual walking experience. The complete new path is a complete routing scheme from the starting point to the end point, composed of the user's actual walking trajectory and the replanned subsequent local paths; its data structure is an ordered list of coordinates. The currently active navigation path is a path data object tracked by the navigation system's internal state machine, used to provide real-time guidance; updating this object switches the system's navigation basis.

[0178] For example, the user has actually walked from the original starting point (east entrance of Wangfujing Pedestrian Street) to the safe resource point (convenience store entrance). The system records the coordinate sequence of this historical trajectory. Let's assume the last point is... The first point of the locally updated path segment generated in step S7 is .right The first 3 points and Smooth interpolation is performed on the connection region between the last three points to generate a transition curve and eliminate... and Directly connecting to any possible bends will allow the system to smooth the historical trajectory (from the starting point to...). ) and local update path segment (from (Starting from the National Centre for the Performing Arts) splicing.

[0179] The newly assembled route is displayed as follows: From the east entrance of Wangfujing Pedestrian Street, follow the route the user actually took to the convenience store entrance, then continue "walk 100 meters east along the auxiliary road of East Chang'an Avenue, turn right onto Nanheyan Street..." until the National Centre for the Performing Arts. This route is set as the active route, the navigation interface is updated, the original flexible skeleton route B is cleared, and the new route is highlighted. The voice prompt is updated, prompting "From the current location, please continue straight east along the auxiliary road of East Chang'an Avenue."

[0180] See appendix Figure 2 This invention also proposes a navigation path selection system for the blind that integrates semantics and real-time information, comprising the following modules:

[0181] The data acquisition and preprocessing module acquires the coordinates of the starting point and destination of the route planning, and extracts road network data and street view points of interest data from the electronic map to generate a city road data package containing basic road attributes and street view points of interest.

[0182] The safety resource point identification and labeling module, based on urban road data packages, identifies and labels physical spaces along roads that conform to preset geometric and functional characteristics, and generates a list of safety resource points containing location and attribute descriptions.

[0183] The global resilient path planning module, based on urban road data packages and a list of safety resource points, performs global path planning with the distribution density of safety resource points along the path as one of the optimization objectives, generating a resilient skeleton path from the origin to the destination.

[0184] The user behavior status analysis module continuously analyzes inertial measurement unit data during navigation along the resilient skeleton path, generating user behavior status signals that reflect the user's real-time travel status.

[0185] The environmental anomaly sound analysis module continuously analyzes environmental sound data during navigation and generates qualitative markers of environmental anomalies by matching it with a preset abnormal sound pattern library.

[0186] The security guidance instruction generation module is used to receive user behavior status signals and environmental anomaly qualitative markers. When both meet the preset alarm triggering conditions, it calculates and outputs guidance instructions to guide to the nearest available security resource point based on the current location and the list of security resource points.

[0187] The local path replanning module allows users to move according to the guidance instructions and confirm that they have arrived at the target safe resource point. Using the safe resource point as the new starting point, the module replans the locally updated path segment to the destination coordinates.

[0188] The seamless path stitching and updating module seamlessly stitches together the user's completed actual walking trajectory with locally updated path segments to form a complete dynamically adjusted navigation path.

[0189] Each of the modules can be implemented in whole or in part through software, hardware, or a combination thereof. It supports hardware embedded in or independent of the processor in the computer device, and also supports software stored in the memory of the computer device, so that the processor can call and execute the operations corresponding to each of the above modules.

[0190] It should be noted that the human information (including but not limited to human device information and personal information) and data (including but not limited to data used for analysis, data stored and data displayed) involved in this invention are all information and data authorized by the human body or fully authorized by all parties. The collection, use and processing of related data require relevant legal standards.

[0191] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A blind navigation path screening method fusing semantics and real-time information, characterized in that, The method comprises the following steps: S1, obtaining the starting point coordinates and destination coordinates of path planning, and extracting road network data and street interest point data from an electronic map to generate a city road data package containing road basic attributes and street interest points; S2, based on the city road data package, identifying and labeling physical spaces along the road that meet the preset geometric and functional characteristics to generate a safety resource point list containing location and attribute descriptions; S3, based on the city road data package and the safety resource point list, taking the distribution density of safety resource points along the original navigation path as one of the optimization objectives for global path planning to generate a resilience skeleton path from the starting point to the destination; S4, during navigation along the resilience skeleton path, continuously analyzing inertial measurement unit data to generate a user behavior state signal reflecting the user's real-time travel state; Obtain the accelerometer data stream and gyroscope data stream in the inertial measurement unit at a fixed frequency; Process the accelerometer data stream to calculate the user's real-time walking speed and step frequency variation coefficient within a continuous time window; When the real-time walking speed continuously falls below the preset proportion of the user's historical average speed for more than a first time threshold, or when the step frequency variation coefficient exceeds a preset variation threshold accompanied by a small range of rotary motion indicated by the gyroscope data, it is determined that the user is in a blocked travel state; if both conditions are not met, it is determined that the user is in a normal travel state; Encapsulate the blocked travel state or normal travel state as a user behavior state signal; S5, during navigation, continuously analyze environmental sound data, and generate environmental anomaly qualitative labels by matching with a preset abnormal sound pattern library; Frame and feature extract continuous environmental sound data, including average volume and energy proportion of specific frequency bands; Match the extracted sound features with the preset abnormal sound pattern library, which contains feature templates of continuous mechanical booming sound, dense car horn sound, and large vehicle idling sound; When the matching degree of continuous multiple frames of sound features with a certain abnormal sound pattern exceeds the set matching degree threshold, it is determined that there is a corresponding type of anomaly in front, and an environmental anomaly qualitative label is generated; S6, receive the user behavior state signal and the environmental anomaly qualitative label, and when both satisfy the preset alarm triggering condition, calculate and output guidance instructions to the nearest available safety resource point according to the current location and the safety resource point list; S7, the user moves according to the guidance instructions and confirms that he has arrived at the target safety resource point, and then takes the safety resource point as a new starting point to re-plan a local update path segment to the destination coordinates; S8, the actual walking trajectory completed by the user is seamlessly spliced with the local update path segment to form a complete dynamically adjusted navigation path; Select a plurality of coordinate points at the end of the actual walking trajectory and a plurality of coordinate points at the beginning of the local update path segment; Generate a transition curve between the actual walking trajectory and the local update path segment using a spline interpolation algorithm; Connect the actual walking trajectory, the transition curve, and the local update path segment in time sequence to form a complete dynamically adjusted navigation path.

2. The method of claim 1, wherein the method further comprises: The city road data package containing road basic attributes and street interest points is generated, comprising the following steps: Receiving a path planning request input by a user through a navigation application, and parsing a starting point coordinate and a destination coordinate of the path planning request; Initiating a request to an electronic map service application programming interface, and obtaining data of road basic attributes in an area covering potential paths centered on a line connecting the starting point and the destination, the road basic attributes including a road name, a lane width, a sidewalk width, and a road surface material; Obtaining, by the electronic map service, data of street view points of interest in the area, the street view points of interest including names and locations of street shops, public facilities, square entrances, and bus stations; Integrating the data of the road basic attributes and the data of the street view points of interest, and generating a city road data package.

3. The method of claim 1, wherein the method further comprises: Generating a safe resource point list containing location and attribute descriptions, including the following steps: Traversing the street view points of interest in the city road data package, and combining with building contour data associated with the street view points of interest to identify physical spaces located within a preset distance range of a road and having a depth and a width of inward recess not less than first and second threshold values, respectively; Performing functional screening on the identified physical spaces, and retaining spaces of a preset safe place type; Creating a record for each screened physical space, the record containing a geographic location coordinate and an attribute description composed of a space type and an expected number of accommodated persons, and generating the safe resource point list.

4. The method of claim 1, wherein the method further comprises: Generating a resilience skeleton path from the starting point to the destination, including the following steps: Based on the road basic attributes and the street view points of interest of the city road data package, using a graph search algorithm to calculate a plurality of initial candidate paths from the starting point to the destination; For each initial candidate path, setting sampling points at equal intervals along the path line, and counting the number of safe resource points existing in the safe resource point list within a preset radius range around each sampling point; Based on the statistical results of the sampling points, calculating the average number of safe resource points per unit length of each path as a safe resource point density indicator; Performing a weighted comprehensive evaluation of the safe resource point density indicator, a path length indicator, and a travel time indicator, and selecting a path with the optimal comprehensive evaluation as the resilience skeleton path.

5. The method of claim 1, wherein the method further comprises: Calculating and outputting guidance instructions to a nearest available safe resource point, including the following steps: When an alarm triggering condition is met, obtaining a current location coordinate of the user; Taking the current location coordinate as a reference point, searching for a safe resource point closest to the user and located in front of the user in a forward direction in the safe resource point list as a nearest available safe resource point; Calculating a straight line distance and a direction from the current location to the nearest available safe resource point, and generating guidance instructions in combination with an attribute description of the safe resource point.

6. The method of claim 5, wherein the method further comprises: The alarm triggering condition includes the following steps: The user behavior state signal is in a blocked state, and the environmental anomaly qualitative label is not empty.

7. A blind navigation path screening system fusing semantics and real-time information, characterized by, The following modules are included: A data acquisition and preprocessing module that obtains a path planning starting point coordinate and a destination coordinate, extracts road network data and street view point of interest data from an electronic map, and generates a city road data package containing road basic attributes and street view points of interest; A safe resource point identification and labeling module that identifies and labels physical spaces along a road that meet preset geometric and functional characteristics based on a city road data package, and generates a safe resource point list containing location and attribute descriptions; and A global resilience path planning module, based on the city road data packet and the safety resource point list, performs global path planning with the safety resource point distribution density along the path as one of the optimization objectives, and generates a resilience skeleton path from the starting point to the destination; A user behavior state analysis module, during the navigation process along the resilience skeleton path, continuously analyzes the inertial measurement unit data to generate a user behavior state signal reflecting the user's real-time travel state; The accelerometer data stream and the gyroscope data stream in the inertial measurement unit are acquired at a fixed frequency; The accelerometer data stream is processed to calculate the user's real-time walking speed and step frequency variation coefficient within a continuous time window; When the real-time walking speed continuously falls below the preset proportion of the user's historical average speed for more than a first time threshold, or when the step frequency variation coefficient exceeds a preset variation threshold accompanied by a small range of rotary motion indicated by the gyroscope data, it is determined that the user is in a travel blocked state; if both conditions are not met, it is determined that the user is in a normal travel state; The travel blocked state or the normal travel state is encapsulated as a user behavior state signal; An environmental abnormal sound analysis module, during the navigation process, continuously analyzes environmental sound data, and generates an environmental abnormal qualitative label by matching with a preset abnormal sound pattern library; The continuous environmental sound data is framed and features are extracted, including average volume and energy proportion of specific frequency bands; The extracted sound features are matched with the preset abnormal sound pattern library, which includes feature templates of continuous mechanical booming sound, dense car horn sound, and large vehicle idling sound; When the matching degree of continuous multiple frames of sound features with a certain abnormal sound pattern exceeds a set matching degree threshold, it is determined that there is a corresponding type of abnormality in front, and an environmental abnormal qualitative label is generated; A safety guidance instruction generation module, for receiving user behavior state signals and environmental abnormal qualitative labels, when both satisfy a preset alarm triggering condition, calculates and outputs guidance instructions to guide to the nearest available safety resource point according to the current location and the safety resource point list; A local path re-planning module, after the user moves and confirms arrival at the target safety resource point, re-plans a local update path segment to the destination coordinates with the safety resource point as the new starting point; A path seamless splicing and updating module, the actual walking trajectory completed by the user is seamlessly spliced with the local update path segment to form a complete dynamically adjusted navigation path; A plurality of coordinate points at the end of the actual walking trajectory and a plurality of coordinate points at the beginning of the local update path segment are selected; A transition curve is generated between the actual walking trajectory and the local update path segment using a spline interpolation algorithm; The actual walking trajectory, the transition curve, and the local update path segment are connected in time sequence to form a complete dynamically adjusted navigation path.

Images