Intelligent control system and method for boarding car for disabled people

Abstract

The invention relates to the technical field of intelligent control and rehabilitation assistance, in particular to an intelligent control system and method for a boarding vehicle for the disabled, and the system comprises a control system platform, a central processing platform, an upper computer, a lower computer, a binocular vision collection module and an intelligent boarding system. The control system platform receives voice and image information and transmits the voice and image information to the central processing platform, and the central processing platform generates a control instruction and sends the control instruction to the upper computer and the lower computer. And the upper computer performs data processing and analysis, and is connected with the cloud server to realize remote storage. The lower computer controls vehicle hardware such as seats, vehicle doors and alarms. The binocular vision collection module collects three-dimensional point cloud data to be processed by the upper computer. The intelligent boarding system is connected with the lower computer, automatically executes boarding operation and adjusts vehicle parts to adapt to the positions of people. According to the system, through innovative design and advanced technology, the boarding safety, efficiency and comfort of the disabled are remarkably improved, and a new standard of intelligent and humanized airport service is established.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control and rehabilitation assistance technology, in particular to an intelligent control system and method for a disabled person's boarding vehicle. Background Art

[0002] With the progress of society and the deepening of care for people with disabilities, the construction of barrier-free facilities in public places such as airports has received increasing attention. However, in the special scenario of disabled people boarding an airplane, the existing assistive devices and methods still have many shortcomings and cannot meet the needs of disabled people for safe, comfortable and efficient boarding.

[0003] At present, the most common way for disabled people to board is for airport staff to manually push wheelchairs and guide disabled passengers. Although this method is simple and direct, it has many problems. First, manual assistance is inefficient, especially in busy airports with large passenger flow, which often causes disabled passengers to wait too long. Secondly, due to the complex airport environment, manual operation is prone to errors, increasing safety hazards. Furthermore, the needs of different disabled people are different, and it is difficult to provide personalized care relying solely on manual assistance, which affects the quality of service.

[0004] In order to improve efficiency and safety, some airports have begun to try to use simple automated boarding car systems. Such systems usually have basic autonomous driving functions and can travel along a preset route. However, such systems still have many limitations. For example, they lack intelligent environmental perception capabilities and have difficulty coping with the complex and changing airport environment; the human-computer interaction interface is simple and cannot meet the needs of different types of disabled people; the system has poor adaptability and is prone to failure in bad weather or during peak passenger flow periods.

[0005] In addition, existing automated systems generally lack the ability to accurately recognize the postures and intentions of disabled people. This results in the system being unable to respond to changes in passengers' needs in a timely manner and making it difficult to prevent possible accidents. At the same time, due to the lack of intelligent seat adjustment functions, it is difficult to provide a comfortable riding experience for disabled people of different body sizes.

[0006] In summary, the existing methods and equipment for assisting disabled people in boarding have many problems, such as low efficiency, insufficient safety, lack of personalized services, and poor environmental adaptability. These problems seriously affect the boarding experience of disabled people and also bring great challenges to airport management. Therefore, it has become a top priority to develop an intelligent and humanized boarding vehicle control system for disabled people. Summary of the invention

[0007] The present invention is proposed to solve the above technical problems. The goal of the present invention is to develop an efficient, safe and intelligent control system and method for disabled people to board a plane, so as to comprehensively improve the disabled people's boarding experience.

[0008] The present invention proposes an intelligent control system for disabled people boarding a vehicle, comprising: Control system platform for: Receive voice commands and image information; Transmit the received instructions and information to the central processing platform; The central processing platform is in communication with the control system platform and is used for: Receiving voice commands and image information sent by the control system platform; Based on the voice command and the image information, generating a control command; Sending the control instruction to the upper computer and the lower computer; The host computer is connected to the central processing platform for: Receiving control instructions sent by the central processing platform; Based on the control instructions, perform data processing and analysis; Connect to cloud servers via the Internet to achieve remote storage and processing of data; The lower computer is connected to the central processing platform for: Receiving control instructions sent by the central processing platform; Control the operation of vehicle hardware devices, including seat adjustment, door switches and alarms; The binocular vision acquisition module is connected to the host computer for: Collect 3D point cloud data of vehicles and personnel; Sending the three-dimensional point cloud data to a host computer for processing; The intelligent boarding system is connected to the lower computer for: Execute automation operations during the boarding process; Adjust vehicle components based on occupant location information.

[0009] Preferably, the control system platform comprises: Voice control subsystem for: Collect voice commands; Extract features from the collected voice commands; Compare and match the extracted speech features with the pre-stored features; Image recognition subsystem for: Collect real-time image information; Preprocess and extract features of the collected images; The extracted image features are classified and recognized using the T-distribution hashing algorithm.

[0010] Preferably, the image recognition subsystem further includes: Image preprocessing module, used to grayscale, denoise, histogram equalize and sharpen the input image; Feature extraction module, used to extract edge features, texture features and shape features of images; Pattern recognition module, used for image clustering and modeling using first-order and second-order local features and first-order and second-order global features; The image judgment output module is used to use the T-distribution hash algorithm as a classifier to map image features into a high-dimensional Hamming space for classification.

[0011] Preferably, the host computer comprises: Data processing unit for: Generate skeletal motion features of vehicles and personnel based on 3D point cloud data; Use preset algorithms to identify vehicle parking status and personnel locations; Pose estimation unit, used to: Predict the person's facing direction in real time based on the person's facial image information; Calculate the horizontal distance between the person's face and the roof and front window; PTZ control unit for: Control the pan / tilt rotation angle according to the vehicle parking status and personnel location information; After the person's feet leave the ground, the gimbal is controlled to rotate according to the sharpness angle between the line connecting the person's eyes and the front.

[0012] Preferably, the lower computer comprises: Seat adjustment device, used to automatically adjust seat height, fore-aft position and lumbar angle according to detected occupant information; A door control unit, used to control the door opening and closing according to a preset door opening algorithm; A voice broadcast system, used to provide voice prompts according to a preset voice broadcast algorithm; The sensor module includes an acoustic and optical alarm and an infrared sensor, which is used to detect vehicle status and personnel information.

[0013] As a preference, it also includes: User terminal, used for: For users to enter boarding time, destination and disability category information; Display matching results and navigation information; The server is connected to the user terminal and the host computer for: Receiving information sent by a user terminal; Perform passenger matching based on the received information; The matching results are sent to the user terminal.

[0014] Preferably, the server comprises: A storage module, used to store disability category, boarding time, target location and corresponding passenger information; Processing modules for: Match other passenger information based on disability category, boarding time and destination location; A multi-level matching algorithm that performs time-sequential matching and position-secondary matching; Location tracking module for: Track user location in real time; Update the ride list based on the user's real-time location.

[0015] Preferably, the binocular vision acquisition module comprises: The first camera and the second camera are placed in parallel up and down and are used to collect stereoscopic visual information; The third camera is fixed in the middle of the front window of the vehicle and is used to collect images in front of the vehicle; The first camera is used to obtain hand image information of a person, and the second camera is used to obtain facial image information of a person.

[0016] Preferably, the host computer and the cloud server are connected to each other through the Internet, and the cloud server is used to: Receive data sent by the host computer and store it; Perform big data analysis and artificial intelligence processing on stored data; The processing results are fed back to the host computer for continuous optimization of the system and personalized services.

[0017] An intelligent control method for a disabled person's boarding vehicle comprises the following steps: Acquire 3D point cloud data of vehicles and personnel through binocular vision acquisition module; Using a data processing unit of a host computer, generating skeleton motion features of vehicles and personnel according to the three-dimensional point cloud data; Based on the skeleton motion features, identifying whether the current vehicle is parked accurately and the location of the personnel; The person's facing direction is obtained in real time through the pose estimation unit; Control the pan / tilt rotation angle according to the vehicle parking status and personnel location information; The seat adjustment device is controlled by the lower computer to automatically adjust the seat according to the passenger information; Use the voice announcement system to remind people of their boarding locations; receiving boarding information input by the user through the user terminal and sending it to the server; The server executes the passenger matching algorithm based on the received information and returns the matching result to the user terminal; Based on the matching results and real-time location information, the disabled person is guided to complete the boarding process.

[0018] The intelligent control system and method for disabled boarding vehicles of the present invention have achieved significant technological breakthroughs in many aspects and brought about a series of remarkable beneficial effects through innovative system architecture design and advanced artificial intelligence technology.

[0019] From a macro perspective, the present invention builds a comprehensive intelligent boarding assistance ecosystem. The system is not just a simple autonomous driving device, but an organic combination of multiple cutting-edge technologies such as environmental perception, human-computer interaction, gesture recognition, and cloud data processing to form a highly coordinated whole. This system-level innovation has greatly improved the overall efficiency and service quality of disabled boarding, setting a new benchmark for intelligent airport management.

[0020] In terms of safety, the present invention achieves accurate perception and real-time analysis of the environment through multi-sensor fusion and advanced computer vision technology. The system can accurately identify potential dangerous factors, such as obstacles, uneven ground, etc., and respond in a timely manner. Especially in crowded airport environments, the system's high-precision obstacle avoidance capability significantly reduces the risk of collision, providing a safer boarding environment for people with disabilities.

[0021] Efficiency improvement is another highlight of the present invention. Through intelligent path planning and real-time scheduling algorithms, the system can automatically select the optimal route according to the current airport conditions, greatly shortening the boarding time. At the same time, the automated operation of the system reduces manual intervention, which not only improves efficiency but also reduces the probability of human error. Test data shows that compared with traditional methods, this system can shorten the boarding time by nearly 50%, which is of great significance for busy airport operations.

[0022] In terms of human-computer interaction, the present invention adopts multimodal interaction technology, including voice recognition, gesture recognition and touch interface. This all-round interaction mode can meet the needs of different types of disabled people. Whether they are visually impaired, hearing impaired or physically disabled, they can easily operate the system. It is particularly worth mentioning that the system's natural language processing capability enables disabled people to control the vehicle through natural dialogue, greatly improving the user experience.

[0023] The present invention has also made a breakthrough in personalized services. Through precise human posture recognition and intelligent analysis, the system can automatically adjust the seat position, backrest angle, etc., to provide each passenger with a tailored riding experience. This kind of intimate personalized service has greatly improved user satisfaction and made the disabled feel respected and cared for.

[0024] Environmental adaptability is another highlight of the invention. By adopting robust control algorithms and advanced sensor technology, the system can maintain stable operation in various complex environments. Whether in bad weather or during peak passenger flow, the system can maintain efficient and reliable performance. This powerful adaptability ensures the continuity and reliability of boarding services and provides strong support for airport operations.

[0025] From a microscopic perspective, the present invention has achieved innovations in many technical details. For example, the innovative design of the binocular vision acquisition module has greatly improved the accuracy of depth information acquisition; the improved skeleton extraction algorithm has improved the accuracy of posture recognition; and the head posture estimation network based on the attention mechanism has enhanced the system's ability to understand human intentions. The superposition effect of these microscopic innovations ultimately reflects a qualitative leap in the overall performance of the system.

[0026] It is worth emphasizing that this invention not only solves technical problems, but also makes important contributions to social value. It creates a more equal and friendly travel environment for people with disabilities, and reflects the care of society for vulnerable groups. At the same time, the intelligent and automated characteristics of the system also reduce labor costs for airports, improve management efficiency, and have significant economic benefits.

[0027] In conclusion, the present invention has comprehensively improved the safety, efficiency and comfort of disabled people boarding aircraft through innovative system design and advanced technology application, and set a new standard for intelligent and humanized airport services. It not only solves many problems existing in the existing technology, but also creates a new era of assistive devices for the disabled, which has great practical value and social significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is the overall structural diagram of the system of the present invention.

[0029] Figure 2 It is the internal structure diagram of the control system platform of the present invention.

[0030] Figure 3 It is the internal structure diagram of the host computer of the present invention.

[0031] Figure 4 It is the internal structure diagram of the lower computer of the present invention.

[0032] Figure 5 It is a structural diagram of the binocular vision acquisition module of the present invention.

[0033] Figure 6 It is a structural diagram of the server of the present invention. DETAILED DESCRIPTION

[0034] Please refer to the attached Figure 1-Figure 6The present invention provides an intelligent control system and method for boarding vehicles for the disabled. The system is intended to provide a safer, more comfortable and more efficient boarding experience for the disabled. The present invention will be described in detail below in conjunction with specific implementation methods.

[0035] like Figure 1 As shown, the intelligent control system for disabled boarding vehicles of the present invention comprises a control system platform 1, a central processing platform 2, a host computer 3, a slave computer 4, a binocular vision acquisition module 5 and an intelligent boarding system 6. These modules are interconnected by wired or wireless communication to form a complete intelligent control system.

[0036] The control system platform 1 is the core of the entire system, responsible for receiving voice commands and image information, and transmitting this information to the central processing platform 2. Preferably, the control system platform 1 uses a high-performance embedded processor with powerful real-time processing capabilities and low power consumption. In one embodiment of the present invention, the processor of the control system platform 1 has a main frequency of up to 2.5 GHz and a memory capacity of 8 GB to ensure that the system can smoothly process complex voice and image data.

[0037] The central processing platform 2 is connected to the control system platform 1 in communication. Its main function is to receive voice commands and image information sent by the control system platform 1, and generate control commands based on this information. The generated control commands will then be sent to the host computer 3 and the slave computer 4. The central processing platform 2 adopts a distributed architecture and can process multiple data streams in parallel, which greatly improves the response speed and processing efficiency of the system. In practical applications, the average response time of the central processing platform 2 does not exceed 50ms, which is crucial for achieving smooth human-computer interaction.

[0038] The host computer 3 maintains a real-time communication connection with the central processing platform 2. Its main task is to receive control instructions sent by the central processing platform 2 and perform data processing and analysis based on these instructions. In addition, the host computer 3 is also connected to the cloud server via the Internet to achieve remote storage and processing of data. This cloud architecture not only improves the scalability of the system, but also lays the foundation for future big data analysis and artificial intelligence applications. In a preferred embodiment of the present invention, the host computer 3 uses a high-performance industrial computer equipped with an Intel i7 processor and 32GB RAM to ensure the rapid execution of complex algorithms.

[0039] The lower computer 4 also maintains a communication connection with the central processing platform 2. Its main responsibility is to receive control instructions from the central processing platform 2 and control various hardware devices of the vehicle accordingly, such as seat adjustment, door switches and alarms, etc. The lower computer 4 adopts a real-time operating system to ensure that control instructions can be executed within milliseconds, thereby ensuring the real-time performance and security of the system.

[0040] The binocular vision acquisition module 5 is directly connected to the host computer 3 and is responsible for collecting the three-dimensional point cloud data of vehicles and personnel. These data are then sent to the host computer 3 for processing. The binocular vision acquisition module 5 uses a high-resolution stereo camera with a resolution of up to 4K and a frame rate of 60fps, which can capture subtle human movements and facial expressions.

[0041] The intelligent boarding system 6 is connected to the lower computer 4 and is responsible for executing various automated operations during the boarding process and adjusting vehicle components according to the personnel position information. The system integrates a variety of sensors and actuators to accurately control various components of the vehicle and improve the safety and comfort of the boarding process.

[0042] like Figure 2 As shown, in one embodiment of the present invention, the control system platform 1 includes a voice control subsystem 11 and an image recognition subsystem 12. The voice control subsystem 11 is responsible for collecting voice commands, extracting features from the collected voice commands, and comparing and matching the extracted voice features with pre-stored features. In order to improve the accuracy of voice recognition, the voice control subsystem 11 uses a deep learning algorithm, and its recognition accuracy can still reach more than 95% in a noisy environment.

[0043] The image recognition subsystem 12 is responsible for collecting real-time image information, preprocessing and extracting features from the collected images, and classifying and identifying the extracted image features using the T-distribution hash algorithm. The core idea of ​​the T-distribution hash algorithm is to map image features to a high-dimensional Hamming space and determine the similarity of images by calculating the Hamming distance. The mathematical expression of the algorithm is as follows: , in, Representing images and The similarity of and Represent two images respectively. and is the corresponding hash code, is the degree of freedom parameter. By adjusting The value of can control the robustness and accuracy of the algorithm. In this system, The value range is 5 to 10, and the specific value is dynamically adjusted according to the actual application scenario.

[0044] The image recognition subsystem 12 also includes an image preprocessing module 121, a feature extraction module 122, a pattern recognition module 123 and an image judgment output module 124. The image preprocessing module 121 performs grayscale, denoising, histogram equalization and sharpening processing on the input image. In the denoising process, an improved bilateral filtering algorithm is used, and its mathematical expression is: , in, is the filtered image, is the original image, is the range kernel function, is the spatial kernel function, is the normalization factor. This improved bilateral filtering algorithm can effectively remove noise while retaining edge details, laying the foundation for subsequent feature extraction.

[0045] The feature extraction module 122 is responsible for extracting edge features, texture features and shape features of the image. In edge feature extraction, the Canny algorithm is used, and its main steps include Gaussian filtering, calculating gradient amplitude and direction, non-maximum suppression and double threshold detection. The core formula of the Canny algorithm is as follows: , , in, is the gradient magnitude, and are the gradients in the x and y directions respectively, is the gradient direction.

[0046] The pattern recognition module 123 uses the first-order and second-order local features and the first-order and second-order global features to perform image clustering and modeling. In the clustering process, an improved K-means algorithm is used, and its objective function is: , in, is the number of clusters, is the number of samples, It is Class samples, It is The center of the class. By minimizing , the optimal clustering result can be obtained.

[0047] The image judgment output module 124 uses the T-distribution hash algorithm as a classifier to map image features to a high-dimensional Hamming space for classification. The output of this module will serve as an important basis for the system to judge the current scene and personnel status.

[0048] From the above detailed description, it can be seen that the intelligent control system for disabled boarding vehicles of the present invention integrates a variety of advanced technologies, including computer vision, speech recognition, machine learning, etc. The organic combination of these technologies enables the system to accurately perceive the environment and personnel status, providing disabled people with a safe, comfortable and efficient boarding experience. At the same time, the modular design and cloud architecture of the system also provide a good foundation for future functional expansion and performance optimization.

[0049] In the intelligent control system for disabled people boarding vehicles of the present invention, the host computer 3 plays a vital role. Figure 3 As shown, the host computer 3 includes a data processing unit 31, a posture estimation unit 32 and a pan / tilt control unit 33. These units work together to ensure that the system can accurately perceive the environment and personnel status and make corresponding control decisions.

[0050] The data processing unit 31 is a core component of the host computer 3. The unit receives the three-dimensional point cloud data from the binocular vision acquisition module 5 and generates the skeletal motion features of vehicles and personnel using advanced algorithms. In a preferred embodiment of the present invention, the data processing unit 31 adopts an improved bone extraction algorithm, which combines deep learning and geometric constraints to accurately extract the human skeletal structure under complex backgrounds. The core idea of ​​the algorithm can be expressed as: , in, represents the extracted bone structure, is a nonlinear mapping function, is the depth image, is a color image, It is the geometric constraint of the human body. Through this multimodal fusion method, the accuracy of skeleton extraction can reach more than 98%, laying a solid foundation for subsequent posture analysis.

[0051] In addition, the data processing unit 31 is also responsible for identifying the parking status of the vehicle and the location of the personnel. In order to achieve high-precision parking status recognition of the vehicle, the present invention adopts a method based on point cloud registration. The method first constructs a standard point cloud model of the vehicle, and then registers the point cloud data collected in real time with the standard model. The registration process uses the iterative closest point (ICP) algorithm, and its objective function is: , in, is the rotation matrix, is the translation vector, and are the corresponding points in the real-time point cloud and the standard model respectively. By minimizing , the precise position and posture of the vehicle can be obtained. When the position and posture deviation is less than the preset threshold (usually ±5cm and ±2°), the system determines that the vehicle is parked accurately.

[0052] The pose estimation unit 32 is another key component of the present invention. This unit is mainly responsible for predicting the person's facing direction in real time based on the person's facial image information, and calculating the horizontal distance between the person's face and the roof and front window. In the prediction of the facing direction, the present invention adopts a multi-task learning framework based on deep learning to simultaneously predict the key points of the face and the head posture. The loss function of the framework can be expressed as: , in, is the key point prediction loss, is the pose prediction loss, is the regularization term, , and is the weight coefficient. By adjusting these weights, the importance of different tasks can be balanced and the overall accuracy of the prediction can be improved. In the practical application of the present invention, the average error of the facing direction prediction does not exceed 5°, which provides accurate input for subsequent gimbal control.

[0053] The pan-tilt control unit 33 is the executive part of the host computer 3, and is responsible for controlling the pan-tilt rotation angle according to the vehicle parking state and the personnel position information. In the preferred embodiment of the present invention, the pan-tilt adopts a three-axis servo motor drive, which can achieve precise translation and rotation control. The motion control of the pan-tilt adopts an improved PID algorithm, and its control equation is: , in, is the control output, is the error signal, , and are the proportional, integral and differential coefficients respectively, is the feed-forward coefficient, is the feedforward compensation term. By introducing feedforward compensation, the system can respond to human actions more quickly and improve the smoothness and accuracy of tracking.

[0054] In particular, after the person's feet leave the ground, the pan / tilt control unit 33 controls the pan / tilt to rotate along the angle between the person's eyes and the front. This intelligent tracking strategy ensures that the camera always captures the best viewing angle of the person, providing reliable image input for subsequent posture analysis and safety monitoring.

[0055] The lower computer 4 is another important component of the system of the present invention, which is mainly responsible for executing specific hardware control tasks. Figure 4As shown, the lower computer 4 includes a seat adjustment device 41, a door control unit 42, a voice broadcast system 43 and a sensor module 44. These modules are interconnected via an industrial bus (such as a CAN bus) to ensure that control instructions can be quickly and reliably transmitted and executed.

[0056] The seat adjustment device 41 is a key device to ensure the comfort of the disabled. The device can automatically adjust the seat height, front and rear position and waist angle according to the detected personnel information. In one embodiment of the present invention, the seat adjustment adopts a fuzzy control algorithm, and its fuzzy rule can be expressed as: IF (height is tall) AND (weight is heavy) THEN (seat height is low) AND (fore / aft position is rear); IF (height is low) AND (weight is light) THEN (seat height is high) AND (anteroposterior position is front); IF (waist circumference is large) THEN (lumbar support angle is large); Through this fuzzy control strategy, the system can adaptively adjust the seats according to the physical characteristics of different disabled people to provide the best riding experience.

[0057] The door control unit 42 is responsible for controlling the door opening and closing according to the preset door opening algorithm. Considering the special needs of disabled people getting on and off the vehicle, the present invention adopts a segmented control strategy. In the initial stage of door opening, the door speed is slow (about 0.1m / s) to prevent collision; the door speed is accelerated (about 0.3m / s) in the middle stage to improve efficiency; and the door speed is reduced again in the final stage (about 0.05m / s) to achieve soft landing. This strategy not only ensures safety, but also improves the efficiency of door opening and closing.

[0058] The voice announcement system 43 is one of the human-computer interaction interfaces of the present invention, and is responsible for providing voice prompts according to a preset voice announcement algorithm. The system uses natural language processing technology and can generate natural and friendly voice prompts according to the current situation. For example, when a disabled person is about to board the bus, the system will announce: Please be careful of the height difference of the floor and hold the handrail with your left hand. This personalized voice prompt greatly improves the ease of use and safety of the system.

[0059] The sensor module 44 includes an acoustic and optical alarm, an infrared sensor, etc., which are used to detect the vehicle status and personnel information. Among them, the infrared sensor adopts an array design, which can accurately detect the position and posture of the personnel. When an abnormal situation (such as a person falling) is detected, the acoustic and optical alarm will immediately sound an alarm, and the system will start the emergency processing process.

[0060] The intelligent control system for boarding vehicles for the disabled of the present invention also includes a user terminal and a server, which together constitute the remote interaction and data processing platform of the system. The user terminal can be a smart phone or a tablet computer, which interacts with the system through a specially developed application. The user can input information such as boarding time, destination and disability category through the terminal, and the system will perform intelligent matching and route planning based on this information.

[0061] The server serves as the central data processing center of the system and maintains real-time communication connection with the user terminal and the host computer. The server adopts a distributed architecture, including a storage module, a processing module, and a location tracking module. The storage module uses distributed database technology to efficiently store and retrieve a large amount of user information and vehicle data. The processing module implements a complex matching algorithm, including time sequence matching and position secondary matching. This multi-level matching algorithm can further optimize the spatial distance and improve the accuracy and efficiency of matching based on time constraints.

[0062] The location tracking module uses GPS and indoor positioning technology to track the user's location in real time. In outdoor environments, the system mainly relies on GPS positioning, with a positioning accuracy of up to 5 meters. In indoor environments (such as airport terminals), the system uses hybrid positioning technology based on Wi-Fi and Bluetooth, and the positioning accuracy can be improved to within 1 meter. Based on real-time location information, the system can dynamically update the ride list and optimize vehicle scheduling, thereby improving overall service efficiency.

[0063] From the above detailed description, it can be seen that the intelligent control system for disabled boarding vehicles of the present invention integrates a number of advanced technologies, including computer vision, artificial intelligence, sensor fusion, cloud computing, etc. The organic combination of these technologies enables the system to provide disabled people with a safe, comfortable and efficient boarding experience. At the same time, the modular design and scalable architecture of the system also lay a solid foundation for the continuous optimization and upgrading of future functions.

[0064] In the intelligent control system for disabled people boarding vehicles of the present invention, the binocular vision acquisition module 5 is a key component for achieving accurate environmental perception. Figure 5 As shown, the module includes a first camera 51, a second camera 52 and a third camera 53. This multi-camera configuration can capture environmental information around the vehicle in all directions, providing a reliable data basis for the intelligent decision-making of the system.

[0065] The first camera 51 and the second camera 52 are placed in parallel up and down, and are mainly used to collect stereoscopic visual information. This layout imitates the structure of the human eye and can effectively perceive depth information. In a preferred embodiment of the present invention, the baseline distance between the two cameras is 10 cm, which is obtained through repeated experiments and optimization. A baseline distance that is too small will result in a decrease in the accuracy of depth estimation, while a baseline distance that is too large will increase the difficulty of parallax matching of close objects. A baseline distance of 10 cm strikes a good balance between accuracy and usability.

[0066] The core of the stereo vision system is disparity calculation and depth estimation. The present invention adopts an improved semi-global matching algorithm (SGM), and its objective function can be expressed as: , in, is the disparity map, It's pixels The matching cost, and is the smoothing term coefficient, is the indicator function. By minimizing this energy function, the system can obtain a high-quality disparity map and then calculate accurate depth information. In practical applications, the algorithm can process a pair of 1080p resolution stereo images in 0.1 seconds, and the average error of depth estimation is less than 1cm.

[0067] The third camera 53 is fixed in the middle of the front window of the vehicle and is mainly used to collect the plane image in front of the vehicle. The introduction of this camera greatly enhances the environmental perception capability of the system. For example, it can be used to detect obstacles in front, identify road signs, etc. In one embodiment of the present invention, the third camera 53 uses a wide-angle lens with a horizontal field of view of 120°, which ensures that the system can capture a sufficiently wide field of view.

[0068] It is worth noting that the first camera 51 is specifically used to obtain the image information of the person's hand, while the second camera 52 is responsible for obtaining the image information of the person's face. This division of labor enables the system to simultaneously capture different parts of the human body with high quality, providing rich visual information for subsequent posture analysis and intention recognition. In order to cope with different lighting conditions, the first camera 51 is also equipped with an infrared fill light to ensure that clear hand images can be obtained even in low-light environments.

[0069] The system of the present invention makes full use of cloud computing technology to connect the host computer 3 with the cloud server through the Internet. This architecture design greatly enhances the computing power and data processing capabilities of the system. The cloud server is not only used to receive and store data sent by the host computer 3, but also undertakes the tasks of big data analysis and artificial intelligence processing.

[0070] In terms of data storage, the cloud server uses distributed storage technology. Specifically, the system uses Apache Hadoop Distributed File System (HDFS) to store massive amounts of structured and unstructured data. HDFS's block replication mechanism (the default replication factor is 3) ensures high data availability and fault tolerance. For hot data that needs to be accessed quickly, the system also introduces the Redis memory database as a cache layer, which greatly improves the data reading speed.

[0071] In terms of big data analysis, the cloud server deploys the Apache Spark distributed computing framework. Spark's in-memory computing feature makes it an order of magnitude faster than traditional MapReduce when processing iterative algorithms. For example, when analyzing user riding patterns, the system uses the collaborative filtering algorithm in the Spark MLlib machine learning library. Its core idea can be expressed as: , in, is the predicted score, is the global average rating, and are the biases of the user and item respectively, and are the latent vectors of items and users respectively. In this way, the system can predict users’ riding preferences and provide more personalized services.

[0072] In terms of AI processing, the cloud server deploys a deep learning platform based on TensorFlow. The platform is capable of performing complex tasks such as image recognition and natural language processing. For example, when identifying the special needs of people with disabilities, the system uses a multimodal deep learning model to process image, voice, and text data simultaneously. The loss function of the model can be expressed as: , in, , and are the losses of image, speech and text modalities respectively, is the loss of multimodal fusion, , , and is the weight coefficient. Through this multimodal fusion approach, the system can more comprehensively understand the needs of people with disabilities and provide more accurate services.

[0073] After processing the data, the cloud server will feed back the results to the host computer 3 for continuous optimization and personalized services of the system. For example, based on the analysis results of historical data, the system can predict the demand for rides in a certain period of time and dispatch vehicles in advance; or automatically adjust the default configuration of in-vehicle equipment according to the user's usage habits.

[0074] The present invention also provides an intelligent control method for disabled people boarding vehicles matching the above system. The method comprises the following steps: First, the three-dimensional point cloud data of vehicles and people is acquired through the binocular vision acquisition module 5. This step is the basis of the entire method and provides raw data for subsequent processing. The system adopts a high frame rate (60fps) data acquisition strategy to ensure that the rapid movements of the human body can be captured.

[0075] Then, the data processing unit 31 of the host computer 3 is used to generate the skeletal motion features of the vehicle and the person according to the three-dimensional point cloud data. This step uses a 3D human posture estimation algorithm based on deep learning. Specifically, the system uses an improved VoxelNet network, and its loss function can be expressed as: , in, is the classification loss, is the regression loss, is the joint-related loss, , and is the corresponding weight. In this way, the system can accurately extract the 17 key points of the human body and build a complete skeleton model.

[0076] Next, based on the skeleton motion features, the system identifies whether the current vehicle is parked correctly and the location of the person. The vehicle parking status is determined using a method based on point cloud registration, as described above. For the determination of the person's location, the system combines the skeleton model and depth information to accurately locate the human body in 3D space.

[0077] At the same time, the person's facing direction is obtained in real time through the posture estimation unit 32. This step is crucial for the system to understand the person's intention. The present invention adopts a head posture estimation network based on the attention mechanism, and its core idea is to let the network automatically learn the key areas of the face. The attention mechanism can be expressed as: , in, , and are query, key, and value matrices respectively, is the key dimension. In this way, the system can accurately estimate the head pose in complex backgrounds, with the average error of the facing direction within 5°.

[0078] According to the parking status of the vehicle and the location of the personnel, the system controls the pan / tilt rotation angle. This step adopts a feedforward-feedback control strategy, which ensures both rapid response and precise control. The output of the controller can be expressed as: , Among them, the first three constitute the classic PID controller, Based on the current state By carefully adjusting the control parameters, the system can achieve millisecond-level response and sub-degree-level control accuracy.

[0079] At the same time, the lower computer 4 controls the seat adjustment device 41 to automatically adjust the seat according to the personnel information. The adjustment process adopts the fuzzy control algorithm, as described above. This adaptive adjustment greatly improves the riding comfort of the disabled.

[0080] Finally, the system uses the voice announcement system 43 to remind the person of the boarding location. The generation of voice prompts uses a Transformer-based text generation model, which can generate natural and friendly voice content according to the current situation. For example, when it is detected that the user is visually impaired, the system will generate more detailed voice guidance.

[0081] During the entire process, users can enter boarding information through the terminal, and the system will send this information to the server for processing. The server executes a complex matching algorithm, considering multiple factors such as time, space, and disability type, to find the most suitable vehicle and route for the user. The matching results will be returned to the user terminal in real time to guide the disabled person to complete the entire boarding process.

[0082] Through the above steps, the method of the present invention can provide all-round intelligent boarding services for disabled people, greatly improving the safety, comfort and efficiency of the boarding process. Each step of the method incorporates advanced artificial intelligence technology, reflecting the innovation and advancement of the present invention.

[0083] In order to fully verify the superiority of the intelligent control system and method of the disabled boarding vehicle of the present invention, we designed a series of simulation experiments. These experiments simulated the real airport environment, covering different types of disabled people, various weather conditions and complex boarding scenarios.

[0084] The simulation conditions are set as follows:

[0085] 1. Environment: Simulate the T3 terminal of an international airport, with an area of ​​about 500,000 square meters and an average daily passenger flow of 100,000.

[0086] 2. Weather conditions: Sunny, cloudy and rainy days each account for one-third of the test time.

[0087] 3. Types of disabled people: Visually impaired, hearing impaired, and physically disabled (wheelchair users) each account for one third.

[0088] 4. Test periods: morning peak (7:00-9:00), off-peak period (11:00-13:00), evening peak (17:00-19:00).

[0089] 5. Test period: 30 consecutive days, 18 hours per day.

[0090] We compared the embodiments of the present invention with two comparative examples:

[0091] Embodiment 1: The intelligent control system and method for disabled boarding vehicles of the present invention.

[0092] Comparative Example 1: Traditional manually assisted boarding method, where airport staff manually push the wheelchair and guide the disabled person to board the aircraft.

[0093] Comparative Example 2: A simple automated boarding car system with basic autonomous driving capabilities but lacks intelligent human-computer interaction and precise environmental perception capabilities.

[0094] The test indicators and their detection methods are as follows:

[0095] 1. Boarding efficiency: the average time required for disabled people to arrive at the designated assembly point and successfully board the aircraft.

[0096] Detection method: Use a high-precision timer to record the time of the entire process.

[0097] 2. Safety: The number of safety accidents (such as collisions, falls, etc.) that occurred during the boarding process.

[0098] Detection method: All security-related events are recorded through airport surveillance systems and on-board sensors.

[0099] 3. User satisfaction: The rating given by people with disabilities on their boarding experience (on a scale of 1 to 10).

[0100] Detection method: Questionnaire survey and face-to-face interview after boarding.

[0101] 4. Environmental adaptability: The normal operation rate of the system under different weather conditions and passenger flow density.

[0102] Detection method: Record the system's operating status and interruption times under various conditions.

[0103] 5. Human-computer interaction accuracy: the rate at which the system correctly understands and executes user instructions.

[0104] Detection method: Record the interaction success rate through log analysis and manual observation.

[0105] 6. Posture recognition accuracy: The rate at which the system correctly recognizes the posture and intention of disabled people.

[0106] Detection method: Compare the system recognition results with the manual labeling results.

[0107] The test results are shown in the following table: ,

[0108] Analysis and discussion:

[0109] 1. Boarding efficiency: The average boarding time of the system of the present invention is 12.5 minutes, which is 50.6% faster than the traditional manual method and 33.2% faster than the simple automated system. This is mainly due to the intelligent path planning and precise environmental perception capabilities of the system, which can find the optimal path in a complex airport environment.

[0110] 2. Safety: The system of the present invention has only 0.3 safety accidents per thousand boardings, which is much lower than the other two methods. This reflects the system's powerful obstacle detection and avoidance capabilities, as well as accurate human posture recognition functions, which can prevent potential safety risks in a timely manner.

[0111] 3. User satisfaction: This method received a high score of 9.2, significantly higher than the other two methods. User feedback indicated that the system's intelligent voice interaction, automatic seat adjustment and other functions greatly improved the boarding experience.

[0112] 4. Environmental adaptability: The present invention maintains a normal operation rate of 98.5% under various weather conditions and passenger flow densities, showing excellent environmental adaptability. This is due to the multi-sensor fusion technology and robust control algorithm adopted by the system.

[0113] 5. Human-computer interaction accuracy: The present invention has achieved a high accuracy of 97.8%, far exceeding that of simple automated systems. This is mainly due to the deep learning speech recognition and natural language processing technology used by the system.

[0114] 6. Posture recognition accuracy: The posture recognition accuracy of the present invention reaches 96.5%, which is 17.6 percentage points higher than that of the simple automation system. This fully demonstrates the advancement of the present system in computer vision and artificial intelligence algorithms.

[0115] In summary, the intelligent control system for disabled boarding vehicles of the present invention is significantly superior to traditional methods and simple automation systems in all key indicators. It not only greatly improves boarding efficiency and safety, but also greatly enhances user experience. The high environmental adaptability of the system means that it can operate stably under various complex conditions, which is particularly important for highly dynamic environments such as airports. In addition, the system's outstanding performance in human-computer interaction and gesture recognition reflects its strong level of intelligence and its ability to accurately understand and respond to the needs of people with disabilities.

[0116] Best Mode

[0117] Based on the above test results, we have determined the best embodiment of the present invention. In this embodiment, the system adopts the following optimized configuration: 1. The binocular vision acquisition module uses a high-speed camera with a resolution of 1920x1080 and a frame rate of 120fps, which improves the accuracy and real-time performance of image acquisition.

[0118] 2. The host computer uses the NVIDIA Jetson AGX Xavier module, which has 32TOPS of AI computing power and can process complex visual and voice data in real time.

[0119] 3. The gimbal is made of carbon fiber material, which is light in weight and high in strength. It is equipped with an encoder with an accuracy of 0.01°, achieving millisecond-level response and sub-degree control accuracy.

[0120] 4. The seat adjustment device adopts a silent stepper motor with an adjustment accuracy of 1mm. At the same time, the noise is controlled below 35 decibels, providing a comfortable and quiet riding experience.

[0121] 5. The human-computer interaction interface uses a large 12.3-inch touch screen and is equipped with anti-glare and anti-fingerprint coating to ensure clear display under all lighting conditions.

[0122] 6. The system software adopts the real-time operating system RTX, and the interrupt response time is less than 1 microsecond, which ensures the real-time performance and reliability of the system.

[0123] This optimal implementation gives full play to the technical advantages of the present invention, achieves excellent results in practical application, and is highly praised by users and airport management. It not only improves the efficiency and safety of disabled people boarding, but also sets a new benchmark for intelligent and humanized airport services.

[0124] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. Intelligent control system for disabled people boarding vehicles, characterized in that: include: Control system platform for: Receive voice commands and image information; Transmit the received instructions and information to the central processing platform; The central processing platform is in communication with the control system platform and is used for: Receiving voice commands and image information sent by the control system platform; Based on the voice command and the image information, generating a control command; Sending the control instruction to the upper computer and the lower computer; The host computer is connected to the central processing platform for: Receiving control instructions sent by the central processing platform; Based on the control instructions, perform data processing and analysis; Connect to cloud servers via the Internet to achieve remote storage and processing of data; The lower computer is connected to the central processing platform for: Receiving control instructions sent by the central processing platform; Control the operation of vehicle hardware devices, including seat adjustment, door switches and alarms; The binocular vision acquisition module is connected to the host computer for: Collect 3D point cloud data of vehicles and personnel; Sending the three-dimensional point cloud data to a host computer for processing; The intelligent boarding system is connected to the lower computer for: Execute automation operations during the boarding process; Adjust vehicle components based on occupant location information.

2. The intelligent control system for disabled people boarding vehicles according to claim 1 is characterized in that: The control system platform includes: Voice control subsystem for: Collect voice commands; Extract features from the collected voice commands; Compare and match the extracted speech features with the pre-stored features; Image recognition subsystem for: Collect real-time image information; Preprocess and extract features of the collected images; The extracted image features are classified and recognized using the T-distribution hashing algorithm.

3. The intelligent control system for disabled people boarding vehicles according to claim 2 is characterized in that: The image recognition subsystem also includes: Image preprocessing module, used to grayscale, denoise, histogram equalize and sharpen the input image; Feature extraction module, used to extract edge features, texture features and shape features of images; Pattern recognition module, used for image clustering and modeling using first-order and second-order local features and first-order and second-order global features; The image judgment output module is used to use the T-distribution hash algorithm as a classifier to map image features into a high-dimensional Hamming space for classification.

4. The intelligent control system for disabled people boarding vehicles according to claim 1 is characterized in that: The host computer comprises: Data processing unit for: Generate skeletal motion features of vehicles and personnel based on 3D point cloud data; Use preset algorithms to identify vehicle parking status and personnel locations; Pose estimation unit, used to: Predict the person's facing direction in real time based on the person's facial image information; Calculate the horizontal distance between the person's face and the roof and front window; PTZ control unit for: Control the pan / tilt rotation angle according to the vehicle parking status and personnel location information; After the person's feet leave the ground, the gimbal is controlled to rotate according to the sharpness angle between the line connecting the person's eyes and the front.

5. The intelligent control system for disabled people boarding vehicles according to claim 1 is characterized in that: The lower computer comprises: Seat adjustment device, used to automatically adjust seat height, fore-aft position and lumbar angle according to detected occupant information; A door control unit, used to control the door opening and closing according to a preset door opening algorithm; A voice broadcast system, used to provide voice prompts according to a preset voice broadcast algorithm; The sensor module includes an acoustic and optical alarm and an infrared sensor, which is used to detect vehicle status and personnel information.

6. The intelligent control system for disabled people boarding vehicles according to claim 1 is characterized in that: Also includes: User terminal, used for: For users to enter boarding time, destination and disability category information; Display matching results and navigation information; The server is connected to the user terminal and the host computer for: Receiving information sent by a user terminal; Matching passengers based on the received information; The matching results are sent to the user terminal.

7. The intelligent control system for disabled people boarding vehicles according to claim 6 is characterized in that: The server comprises: A storage module, used to store disability category, boarding time, target location and corresponding passenger information; Processing modules for: Match other passenger information based on disability category, boarding time and destination location; A multi-level matching algorithm that performs time-sequential matching and position-secondary matching; Location tracking module for: Track user location in real time; Update the ride list based on the user's real-time location.

8. The intelligent control system for disabled people boarding vehicles according to claim 1 is characterized in that: The binocular vision acquisition module includes: The first camera and the second camera are placed in parallel up and down and are used to collect stereoscopic visual information; The third camera is fixed in the middle of the front window of the vehicle and is used to collect images in front of the vehicle; The first camera is used to obtain hand image information of a person, and the second camera is used to obtain facial image information of a person.

9. The intelligent control system for disabled people boarding vehicles according to claim 1 is characterized in that: The host computer and the cloud server are connected to each other through the Internet, and the cloud server is used to: Receive data sent by the host computer and store it; Perform big data analysis and artificial intelligence processing on stored data; The processing results are fed back to the host computer for continuous optimization of the system and personalized services.

10. An intelligent control method for disabled people boarding a vehicle, characterized in that: The following steps are involved: Acquire 3D point cloud data of vehicles and personnel through binocular vision acquisition module; Using a data processing unit of a host computer, generating skeleton motion features of vehicles and personnel according to the three-dimensional point cloud data; Based on the skeleton motion features, identifying whether the current vehicle is parked accurately and the location of the personnel; The person's facing direction is obtained in real time through the pose estimation unit; Control the pan / tilt rotation angle according to the vehicle parking status and personnel location information; The seat adjustment device is controlled by the lower computer to automatically adjust the seat according to the passenger information; Use the voice announcement system to remind people of their boarding locations; receiving boarding information input by the user through the user terminal and sending it to the server; The server executes the passenger matching algorithm based on the received information and returns the matching result to the user terminal; Based on the matching results and real-time location information, the disabled person is guided to complete the boarding process.

Images