Wearable intelligent obstacle avoidance system

Abstract

The invention discloses a wearable intelligent obstacle avoidance system, and belongs to the technical field of blind person auxiliary travel. The invention aims to solve the problem that the existing blind person travel auxiliary equipment is inconvenient to use or high in cost. The method includes the steps that an image collection module collects blind sidewalk obstacle images; the GPS positioning module obtains position coordinates of a user; the lightweight target detection model performs obstacle recognition on the blind sidewalk obstacle image based on the position coordinates of the user to obtain obstacle information; the monocular distance measurement module obtains a distance signal between the current position of the user and the obstacle according to the obstacle information; the control module sends the current distance signal to the prompt module and sends a danger prompt signal when the distance between the current position of the user and the obstacle is smaller than a set threshold value; the prompting module sends a distance signal to a user; an alarm is sent to a user according to the danger prompt signal; all the modules are integrated on the wearable article to form the wearable device. The method is used for assisting the blind in going out.

Description

Technical Field

[0001] This invention relates to a wearable intelligent obstacle avoidance system, belonging to the field of assistive travel technology for the blind. Background Technology

[0002] Currently, there is a significant gap between the travel desires of visually impaired individuals and the actual obstacles they face. Tactile paving, a vital route for the blind, is frequently obstructed, preventing the fulfillment of their travel needs. This restricted travel not only severely hinders their right to participate in social, economic, and cultural life but also leads to widespread social isolation and immense psychological stress. Therefore, accelerating the construction of an accessible environment is urgently needed, and practical intelligent mobility products are urgently required.

[0003] Existing assistive travel measures for the blind all have shortcomings. For example, in traditional methods, guide canes have limited sensory capabilities, and guide dogs are difficult to use widely. In intelligent assistive methods, mobile phone applications rely heavily on the phone's camera, which places high demands on blind people's operation. Intelligent products such as electronic guide dogs and guide glasses are expensive due to the use of modules such as LiDAR, making them difficult to apply universally. Summary of the Invention

[0004] To address the problems of inconvenience or high cost of existing assistive devices for blind people's travel, this invention provides a wearable intelligent obstacle avoidance system.

[0005] A wearable intelligent obstacle avoidance system of the present invention includes:

[0006] The image acquisition module is used to acquire images of obstacles on the tactile paving.

[0007] The GPS positioning module is used to obtain the user's location coordinates;

[0008] A lightweight target detection model is used to identify obstacles in images of tactile paving based on the user's location coordinates, thereby obtaining obstacle information;

[0009] The monocular ranging module is used to obtain the distance signal between the user's current position and the obstacle based on obstacle information;

[0010] The control module is used to send the current distance signal to the prompting module, and send a danger warning signal when the distance between the user's current position and the obstacle is less than a set threshold;

[0011] The alert module is used to send distance signals to the user and to send alarms to the user based on the danger alert signals.

[0012] The image acquisition module, GPS positioning module, lightweight target detection model, monocular ranging module, control module, and prompting module are integrated into the wearable item to form a wearable device.

[0013] The wearable intelligent obstacle avoidance system according to the present invention further includes:

[0014] The cloud module is used to draw an obstacle map based on obstacle information;

[0015] The control module sends obstacle zone prompts to the user based on the obstacle map and the user's location coordinates.

[0016] According to the wearable intelligent obstacle avoidance system of the present invention, the monocular ranging module calculates the distance signal based on the perspective projection algorithm. It utilizes the similar triangle relationship formed during the imaging process and calculates the distance between the user's current position and the obstacle by using the known actual size of the obstacle, the imaging focal length, and the pixel size of the obstacle in the tactile paving obstacle image.

[0017] According to the wearable intelligent obstacle avoidance system of the present invention, the imaging focal length is calculated. :

[0018] ,

[0019] In the formula The sample pixel size of the obstacle. Given the known distance between the location and the obstacle. The actual dimensions of the obstacle;

[0020] The distance signal between the user's current position and the obstacle. for:

[0021]

[0022] In the formula This represents the actual pixel size of the obstacle.

[0023] According to the wearable intelligent obstacle avoidance system of the present invention, a GUI interface is designed to realize the interactive operation of the lightweight target detection model and the monocular ranging module.

[0024] According to the wearable intelligent obstacle avoidance system of the present invention, the control module is implemented based on Raspberry Pi.

[0025] According to the wearable intelligent obstacle avoidance system of the present invention, the prompting module includes a voice module for sending distance signals and alarms to the user.

[0026] According to the wearable intelligent obstacle avoidance system of the present invention, the prompting module further includes a vibration module, which sends an alarm to the user through changes in vibration frequency.

[0027] According to the wearable intelligent obstacle avoidance system of the present invention, the method of sending alarm by the voice module is that the smaller the distance between the user's current position and the obstacle, the louder the volume;

[0028] The vibration module sends alarms to the user by increasing the vibration frequency as the distance between the user's current position and the obstacle decreases.

[0029] According to the wearable intelligent obstacle avoidance system of the present invention, the wearable item includes a vest.

[0030] The beneficial effects of the present invention: The system of the present invention is used to assist blind people in traveling, and can create an accessible environment for blind people to travel through the collaboration between the device and the cloud.

[0031] This invention system, considering real-time accurate identification, simple operation, timely warnings, and ease of dissemination, is designed with the genuine benefit of visually impaired individuals in mind. It proactively constructs a real-time information network for obstacles on urban tactile paving, contributing to the creation of an accessible environment. Firstly, the system is designed as a wearable device, based on a target detection model and monocular ranging algorithm, with an obstacle recognition model algorithm designed for a dedicated dataset of tactile paving obstacles. The image acquisition module identifies obstacles in the captured images, and when an obstacle is detected, a warning module is invoked. The collected obstacle information can be uploaded to a server for management, and the distribution of obstacles on tactile paving can be summarized and displayed through a web platform. The device periodically retrieves information from the cloud to plan travel routes for the visually impaired, assisting them in their travel and achieving end-to-end cloud collaboration to contribute to the construction of an accessible environment.

[0032] This invention can effectively assist blind people in traveling and improve their travel environment. The system is simple and convenient to operate, with low cost, avoiding the use of expensive functional modules such as LiDAR; it is highly flexible, with a vest-integrated design that considers the psychological comfort of visually impaired individuals. The size and functional module placement can be tailored to specific user groups, freeing up the hands of visually impaired individuals and improving the device's versatility. This invention transforms the blind person's passive obstacle avoidance into active route planning. By establishing an edge-cloud collaborative system to display the location distribution of obstacles on urban tactile paving, it facilitates management by relevant personnel. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of the wearable intelligent obstacle avoidance system described in this invention.

[0034] Figure 2 This is a schematic diagram of the GUI interface;

[0035] Figure 3 This is a schematic diagram of the GUI interface detection results;

[0036] Figure 4 This is a control flowchart of the wearable intelligent obstacle avoidance system described in this invention;

[0037] Figure 5These are physical images showing the functional modules of a wearable smart obstacle avoidance system.

[0038] Figure 6 This is an integrated physical image of a wearable intelligent obstacle avoidance system;

[0039] Figure 7 This is an example image of an obstacle map;

[0040] Figure 8 This is an example screenshot of the obstacle information uploading webpage platform interface;

[0041] Figure 9 This is a screenshot of the server-side information management platform interface. Detailed Implementation

[0042] 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, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] Specific Implementation Method 1: Combination Figures 1 to 4 As shown, the present invention provides a wearable intelligent obstacle avoidance system, comprising:

[0044] The image acquisition module is used to acquire images of obstacles on the tactile paving.

[0045] The GPS positioning module is used to obtain the user's location coordinates;

[0046] A lightweight target detection model is used to identify obstacles in images of tactile paving based on the user's location coordinates, thereby obtaining obstacle information;

[0047] The monocular ranging module is used to obtain the distance signal between the user's current position and the obstacle based on obstacle information;

[0048] The control module is used to send the current distance signal to the prompting module, and send a danger warning signal when the distance between the user's current position and the obstacle is less than a set threshold;

[0049] The alert module is used to send distance signals to the user and to send alarms to the user based on the danger alert signals.

[0050] The image acquisition module, GPS positioning module, lightweight target detection model, monocular ranging module, control module, and prompting module are integrated into the wearable item to form a wearable device.

[0051] The lightweight object detection model employs an LW-DETR-based model. The image acquisition module can be a camera. The model is trained and adjusted using a dedicated dataset of images of tactile paving obstacles. Fine-tuning of the LW-DETR series models is performed, ultimately selecting the LW-DETR large fine-tuned model. Compared to traditional YOLO models, this model requires no post-processing, has a moderate size, good real-time performance, high accuracy, and is easy to deploy at the edge with limited computing resources. It can achieve obstacle category recognition. Figure 2 and Figure 3 As shown, the model algorithm can be software tested and verified, and deployed on a Raspberry Pi to design wearable smart devices. Device functions include using a camera to detect obstacles in front in real time and promptly issuing voice and vibration alerts to blind users. Then, the detected obstacle data is uploaded to a cloud module for management, and a web platform is built to draw a "problem map" of obstacles on the tactile paving, such as... Figure 7 and Figure 8 As shown; the obstruction of tactile paving is visualized and summarized, and the summarized information from the cloud is transmitted back to the device, such as... Figure 9 As shown, this serves as a reminder for blind people to avoid areas with many obstacles on the tactile paving, further assisting them in their travel.

[0052] Furthermore, this embodiment also includes:

[0053] The cloud module is used to draw an obstacle map based on obstacle information;

[0054] The control module sends obstacle zone prompts to the user based on the obstacle map and the user's location coordinates.

[0055] In this embodiment, the monocular ranging module calculates the distance signal based on the perspective projection algorithm. It utilizes the similar triangle relationship formed during the imaging process and calculates the distance between the user's current position and the obstacle by taking into account the actual size of the obstacle, the imaging focal length, and the pixel size of the obstacle in the tactile paving obstacle image.

[0056] Furthermore, calculate the imaging focal length. :

[0057] ,

[0058] In the formula The sample pixel size of the obstacle. Given the known distance between the location and the obstacle. The actual dimensions of the obstacle;

[0059] The distance signal between the user's current position and the obstacle. for:

[0060]

[0061] In the formula This represents the actual pixel size of the obstacle.

[0062] In practical use, the imaging focal length is first determined through actual measurement. Set an actual width (or height) as The object is placed at a distance from the camera. Take a picture of the object's location and measure its pixel width (or height) in the image. Based on the property that corresponding sides of similar triangles are proportional, the camera focal length can be derived. Then, when the relative position of the camera and the object changes, the pixel size of the object in the image also changes. At this point, the focal length at which the image is captured... Given the current distance, calculate the current distance. The imaging focal length. This is not the physical focal length in the traditional photographic sense, but rather the camera's intrinsic focal length. Because the image focal length is affected by factors such as camera sensor size and lens mounting errors, it cannot be directly obtained from lens markings and must be accurately measured using experiments or professional calibration tools. After calibration, the distance between the device and the obstacle is estimated by combining the obstacle size information obtained from the diversified strategy.

[0063] Design a GUI interface to enable interactive operation between the lightweight target detection model and the monocular ranging module.

[0064] A GUI interface was designed on a computer to perform software testing on the trained target detection model and ranging algorithm. The GUI interface visualizes and interacts with the results of obstacle recognition on tactile paving. Through detection and distance calculation, it outputs intelligent real-time voice prompts to plan travel routes for the blind, thus verifying the effectiveness of the algorithm model.

[0065] As an example, the control module is implemented based on a Raspberry Pi.

[0066] In this embodiment, each module is deployed on a hardware device, and the Raspberry Pi hardware motherboard can be controlled by a program to perform obstacle recognition on the image captured by the camera. The wearable device uses a Raspberry Pi as the control motherboard. The specific hardware deployment process of the algorithm model is as follows: first, the operating system is burned into the Raspberry Pi; then, it is connected to a computer via wired or wireless means, and remote desktop software is used to access the Raspberry Pi for code writing and device control. To avoid dependency conflicts, maintain a clean system environment, and enhance security, a virtual environment is created on the Raspberry Pi for environment configuration and installation of necessary libraries. Subsequent inference and recognition are performed in the virtual environment.

[0067] To improve the real-time performance and recognition rate of obstacle recognition, the initial model training and algorithm design are carried out on the computer. After training the dataset using the model, a PT model that meets the actual needs of obstacle recognition on tactile paving is obtained. Then, it is converted into an ONNX model and the model file is transferred to a Raspberry Pi for deployment. The trained model is used to perform inference and recognition on real-time video streams to identify whether there are obstacles and their categories in the video stream, resulting in higher real-time performance.

[0068] The prompting module includes a voice module, used to send distance signals and alarms to the user.

[0069] The alert module also includes a vibration module, which sends an alarm to the user by changing the vibration frequency.

[0070] When an obstacle is detected, the vibration and voice modules are activated to provide an alert. The obstacle information collected by the device is then uploaded to the cloud module for management. The distribution of obstacles on the tactile paving is summarized and displayed through a web platform. The device periodically retrieves cloud information to plan travel routes for the blind, assisting them in their travel and achieving end-to-end cloud collaboration to help build an accessible environment.

[0071] As an example, the method by which the voice module sends an alarm is that the smaller the distance between the user's current location and the obstacle, the louder the volume;

[0072] The vibration module sends alarms to the user by increasing the vibration frequency as the distance between the user's current position and the obstacle decreases.

[0073] As an example, the wearable items include vests.

[0074] The system of this invention can be used to help blind people plan their travel routes, and from passive to active, build an end-to-cloud collaborative system to help create an accessible environment for blind people in cities and improve their travel environment.

[0075] Based on the deployed model algorithms, a wearable device for assisting blind people in their travel is designed, using a Raspberry Pi as the main control module and connecting other functional modules. The Raspberry Pi serves as the main control unit, connected to an external camera module, GPS positioning module, vibration module, and voice module. By deploying algorithms and models, the camera can capture real-time images, identify and detect obstacle types ahead, calculate distances, and monitor for anomalies in the camera feed. It plays corresponding obstacle warning audio and provides vibration alerts based on distance for identified obstacles. The device also uploads obstacle information from the cloud module (server) to build a database and retrieves aggregated obstacle information from the cloud to remind blind people to avoid areas with many obstacles, specifically prompting them to replan their route when they reach medium- or high-risk areas. Specifically, the camera module captures the image ahead; the GPS positioning module obtains the device's current location, facilitating the uploading of obstacle location information and assessment of the area's danger level; the vibration module provides vibration alerts based on obstacle distance, with closer objects vibrating more intensely; and the voice module can play obstacle warning audio through a speaker or be replaced with headphones. Design control code in the Raspberry Pi to control the operation of each module according to its function.

[0076] Considering the psychological comfort of visually impaired individuals and the ease of operation of the devices, the smart device is integrated with a regular vest. All hardware modules are integrated into a wearable device. After the user turns on the device, the camera identifies obstacles in the central view in front. When an obstacle is detected, the vibration module and voice broadcast module are invoked to provide a reminder. The data on obstacles on the tactile paving is also uploaded to the server for management. The system also periodically retrieves data on the locations of medium- and high-risk areas with a high distribution of obstacles on the tactile paving from the cloud to remind blind people to avoid obstacles and plan their travel routes.

[0077] The system's image acquisition module, GPS positioning module, lightweight target detection model, monocular ranging module, control module, and prompting module form the terminal for identifying obstacle data on tactile paving. This data is then uploaded to the server for storage and management. Based on the data uploaded from the device, a web platform is built to display the distribution of obstructed tactile paving within a city. Others can access the webpage to view the distribution of obstacles on tactile paving, check its accessibility, receive feedback, and address unreasonable obstructions, ensuring the safety and convenience of blind people. Furthermore, the webpage allows for manual uploads; anyone encountering malicious obstruction of tactile paving can take photos and upload the information to the server for display on the webpage, making the obstacle distribution information more comprehensive and accurate. The server-side design includes a management interface with filtering functions. Data maintenance personnel can filter obstacle information by type and time, perform verification, deletion, and modification operations, and set obstacle handling status for convenient management.

[0078] Figure 1 The overall design of the present invention is presented, including the design of a lightweight target detection model and ranging algorithm, algorithm software verification and hardware deployment, wearable device terminal, server interface and web page display platform interface. Figure 2 The GUI interface shown can recognize obstacles in images, videos, and real-time camera feeds from a computer. Figure 3 It enables the identification and distance measurement of obstacles in images, and calls the Llama 3.2 language model to generate intelligent voice, which combines obstacle information and distance information to provide voice broadcast. Figure 4 In the process, after the control is activated, the camera is called to capture frames and detect them at intervals to obtain the time and GPS latitude and longitude. The multi-threaded control auxiliary module realizes the function of reminding blind people and information interaction. Figure 5 The device includes a Raspberry Pi control board, a camera module, a GPS module, a vibration module, and a power module. Figure 6 The design integrates the functional modules with the vest. The camera module is separately equipped on the belt and can be flexibly adjusted in position. The vibration module is embedded inside the vest, close to the skin to easily alert blind people. The voice broadcast has both speaker and headphone options. The power switch is located on the right side of the waist. Figure 7 The system summarizes and displays the status of urban tactile paving obstructed by obstacles. Green indicates obstacles that have been cleared, while red indicates those that have not, facilitating viewing and management by relevant personnel. Clicking the view switch button displays the summary results of urban tactile paving obstacle information. By measuring various indicators, the system calculates the danger level of each area and transmits the summary results back to the device in real time from the cloud, reminding blind people to avoid areas with many obstacles on tactile paving. Figure 8 The web platform interface is designed to allow users to manually upload obstacle information, enriching the sources of obstacle information. Figure 9 After the information on detected obstacles to the tactile paving is uploaded and displayed, managers can screen them as needed to verify the occupancy status and add a "processed" mark to obstacles that have been removed.

[0079] To better understand the technical solution of this invention, a detailed description is provided below with reference to the accompanying drawings:

[0080] 1. Select and design an obstacle recognition and detection model to recognize images of obstacles on tactile paving.

[0081] We collected and extracted datasets of 15 common tactile paving obstacles from multiple authoritative websites and professional datasets, encompassing 5066 training images and 1266 validation images. Each image is annotated with bounding boxes. The data primarily comes from the Obstacle-dataset (OD), with reference to the TP-Dataset as well. The annotations in txt and VOC formats provided by the OD dataset were converted to COCO's JSON annotation format. For this converted dataset, we trained three models: LW-DETR-tiny, LW-DETR-large, and LW-DETR-xlarge. By comparing detection accuracy and training time, we selected the LW-DETR large model, which offers moderate size, good real-time performance, and high accuracy, for fine-tuning on the obstacle dataset. Compared to traditional YOLO models, this model requires no post-processing steps, offering high real-time performance and accuracy.

[0082] 2. Design a ranging algorithm to obtain distance data between the recognition device and the obstacle while detecting the obstacle.

[0083] After calibrating specific camera modules, standard size databases are used to directly reference industry standard sizes for standardized obstacles (such as traffic signs and fire-fighting facilities), and statistical learning methods are used to statistically analyze the size distribution of similar obstacles from large-scale labeled datasets. Through these two methods, 15 types of obstacle size information are obtained to estimate the distance between the device and the obstacle.

[0084] 3. Design a GUI interface on the computer to perform software testing on the trained target detection model and ranging algorithm.

[0085] Design a GUI interface to implement the target detection algorithm, and visualize and interactively display the detection results of obstacles on tactile paving. Figure 2 As shown, the designed interface supports obstacle recognition in three input modes: image, video, and camera. After clicking the "Detection Run" button, obstacle detection and distance calculation are performed on the target. The detection results are as follows. Figure 3 As shown; then the detection data and images are transmitted to a multimodal large model to generate semantic descriptions, output intelligent real-time voice broadcast prompts, plan travel routes for blind people, and guide blind people to walk in safe areas with fewer obstacles.

[0086] 4. Deploy the trained target detection model and ranging algorithm on the Raspberry Pi hardware device.

[0087] The wearable device uses a Raspberry Pi 4B (4G) development board as the main control unit, equipped with a Broadcom BCM2711 chip, a quad-core ARM Cortex-A72 processor with a main frequency of up to 1.5GHz, 4GB of LPDDR4 memory, and integrates the OpenCV 4G computer vision library, which can realize functions such as image recognition. In addition, the development board has a variety of peripheral interfaces, which can realize a variety of auxiliary functions.

[0088] The hardware deployment process for the algorithm model is based on the selected Raspberry Pi model. The operating system is burned into the Raspberry Pi, and then the computer is wirelessly connected via SSH. The remote desktop VNC software is used to access the Raspberry Pi for environment configuration, file transfer, code writing, and device control. The model weight parameters trained on the computer are transferred to the Raspberry Pi. The camera module is called to perform inference and recognition of obstacles in the picture. The recognition has good real-time performance and high accuracy.

[0089] 5. Based on the deployment of the model algorithm, use Raspberry Pi as the main control module and connect other functional modules to design a wearable device that is comprehensive in function and easy to operate to assist blind people in their travel.

[0090] Using a Raspberry Pi as the main control unit, and connecting external camera, GPS positioning, vibration, and voice modules, the system deploys algorithms and models to capture real-time video from the camera, identify and detect obstacles ahead, calculate distances, and monitor for anomalies in the camera feed. It plays corresponding obstacle warning audio and provides vibration alerts based on distance for identified obstacles. The system also uploads obstacle information from the tactile paving to a server to build a database and retrieves cloud-aggregated obstacle information to remind blind users to avoid areas with many obstacles, specifically prompting them to replan their route when they reach medium- or high-risk areas. Control code is designed within the Raspberry Pi to control the functionality of each module, with several auxiliary modules using multi-threaded control. The control flow is as follows: Figure 4 As shown.

[0091] The functional modules of the intelligent obstacle avoidance device are shown below. Figure 5As shown, the camera module uses a wide-angle HD CSI camera adapted for Raspberry Pi, employing an OV5647 chip to capture real-time images and video. The GPS positioning module uses an L76K GNSS module; when connected to the Raspberry Pi, VCC is connected to the 5V power supply pin, GND to the ground pin, TX to GPIO15 (serial receive), and RX to GPIO14 (serial transmit). The vibration module uses a Vibration miniature vibration module, operating at 5V, controlled by high / low level / PWM signals. The signal pin is connected to Raspberry Pi GPIO27 to transmit vibration control signals, VCC is connected to the 5V power supply pin, and GND is connected to the ground pin. The voice playback module uses Python's pydub library to play audio, connecting to speakers or headphones via USB or a 3.5mm interface.

[0092] The functional modules are integrated with a regular vest, forming a wearable device. Figure 6 As shown. The camera module is flexibly placed by embedding it into the waistband, while the other functional modules are integrated into the vest design. The vibration module is located close to the skin on the chest and can be equipped with headphones for voice prompts. The control motherboard, GPS module, and power module are enclosed inside the vest, and the power switch is located on the user's right waist. After the user turns on the device, the camera identifies obstacles in the central image in front. When an obstacle is detected, the vibration module and voice broadcast module are activated to provide a warning. When there is an anomaly in the camera image, an abnormality voice prompt will be played to warn the user. The device also uploads obstacle data on the tactile paving to the server for management. It periodically retrieves data on the locations of medium- and high-risk areas with a high distribution of obstacles on the tactile paving from the cloud and provides warnings when the user enters a medium- or high-risk area.

[0093] 6. Build an end-to-cloud collaborative system by designing the information interaction between the device and the cloud.

[0094] The device acts as a terminal to identify obstacle data on the tactile paving and uploads the data to the server. The device primarily uses the `requests` library in Python to send HTTP requests and communicate with the server, uploading detection results and obstacle images. The uploaded data includes form data and file data. Form data is encoded in dictionary format and contains information such as obstacle type, GPS location information, upload time, and device ID. File data uses tuples to specify the filename and file content, and the detected images are uploaded as files. The server's URL and API endpoint are configured initially to specify the target address for data uploads.

[0095] A web platform is built based on data uploaded from the device to the server to display the distribution of tactile paving obstructed by obstacles in a city, specifically as follows: Figure 7As shown, the website displays the distribution of tactile paving obstacles that have been addressed and those that have not, and summarizes the distribution to show the risk level of each area. The website also allows for manual uploads, such as... Figure 8 As shown, anyone encountering malicious obstruction of tactile paving can take a photo and upload the information to the server, which will then be displayed on a webpage, enriching the information sources. The server is built using Flask, listening for POST requests from hardware terminals and GET requests from the web platform. Data uploaded from hardware terminals is verified by the server, stored in a JSON file, and the results are returned. The web platform initiates GET requests through different routes to retrieve different types of obstacle reports. A server information management interface is also designed, such as... Figure 9 As shown, on this interface, data maintenance personnel can filter obstacle information by type and time, verify the status of obstacles on the tactile paving, and set the obstacle handling status for convenient management by relevant personnel.

[0096] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. A wearable intelligent obstacle avoidance system, characterized in that, include: The image acquisition module is used to acquire images of obstacles on the tactile paving. The GPS positioning module is used to obtain the user's location coordinates; A lightweight target detection model is used to identify obstacles in images of tactile paving based on the user's location coordinates, thereby obtaining obstacle information; The monocular ranging module is used to obtain the distance signal between the user's current position and the obstacle based on obstacle information; The control module is used to send the current distance signal to the prompting module, and send a danger warning signal when the distance between the user's current position and the obstacle is less than a set threshold; The notification module is used to send distance signals to the user; It also sends alarms to users based on danger warning signals; The image acquisition module, GPS positioning module, lightweight target detection model, monocular ranging module, control module, and prompting module are integrated into the wearable item to form a wearable device.

2. The wearable intelligent obstacle avoidance system according to claim 1, characterized in that, Also includes: The cloud module is used to draw an obstacle map based on obstacle information; The control module sends obstacle zone prompts to the user based on the obstacle map and the user's location coordinates.

3. The wearable intelligent obstacle avoidance system according to claim 1, characterized in that, The monocular ranging module calculates the distance signal based on the perspective projection algorithm. It utilizes the similar triangle relationship formed during the imaging process and, through the known actual size of the obstacle, the imaging focal length, and the pixel size of the obstacle in the tactile paving obstacle image, calculates the distance between the user's current position and the obstacle.

4. The wearable intelligent obstacle avoidance system according to claim 3, characterized in that, Calculate the imaging focal length : ， In the formula The sample pixel size of the obstacle. Given the known distance between the location and the obstacle. The actual dimensions of the obstacle; The distance signal between the user's current position and the obstacle. for: In the formula This represents the actual pixel size of the obstacle.

5. The wearable intelligent obstacle avoidance system according to claim 4, characterized in that, Design a GUI interface to enable interactive operation between the lightweight target detection model and the monocular ranging module.

6. The wearable intelligent obstacle avoidance system according to claim 5, characterized in that, The control module is implemented based on a Raspberry Pi.

7. The wearable intelligent obstacle avoidance system according to claim 5, characterized in that, The prompting module includes a voice module, used to send distance signals and alarms to the user.

8. The wearable intelligent obstacle avoidance system according to claim 7, characterized in that, The alert module also includes a vibration module, which sends an alarm to the user by changing the vibration frequency.

9. The wearable intelligent obstacle avoidance system according to claim 8, characterized in that, The method by which the voice module sends the alarm is that the smaller the distance between the user's current location and the obstacle, the louder the volume; The vibration module sends alarms to the user by increasing the vibration frequency as the distance between the user's current position and the obstacle decreases.

10. The wearable intelligent obstacle avoidance system according to claim 1, characterized in that, The wearable items include vests.

Images