Multi-sensor positioning system in a tunnel and method for locating a person

Abstract

The application discloses a tunnel multi-sensor positioning device and a positioning method for personnel. The tunnel multi-sensor positioning device comprises a wireless network bridge fixedly arranged on each anchor point, a plurality of UWB positioning base stations arranged in wired connection with the anchor points and the network bridge, and a personnel positioning safety helmet worn by personnel moving in the tunnel. A server is in wired connection with the wireless network bridge of any anchor point in the tunnel. The wireless network bridge on each anchor point wirelessly bridges the wireless network bridge on the adjacent anchor point to cover the UWB signal of the UWB positioning base station to the whole tunnel area. The personnel positioning safety helmet comprises a UWB positioning tag and an IMU inertial sensor arranged on the safety helmet and a camera with a holder arranged at the top end of the safety helmet. The application also provides a positioning method for personnel by using the positioning device. The application adopts the positioning technology of fusion of UWB and visual sensors and compensates through the IMU, thereby improving the positioning precision of personnel in the tunnel.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tunnel positioning, in particular to a multi-sensor positioning system in a tunnel under construction and a positioning method for personnel. BACKGROUND

[0002] At present, with the improvement of transportation infrastructure, tunnel construction is rapidly increasing. Due to the influence of the poor construction environment of the tunnel, safety accidents often occur during the construction process, which poses a great threat to the life and property safety of the construction personnel. Traditional tunnel engineering is mostly carried out in a closed and narrow underground environment, and various large equipment needs to be used during the construction process, which seriously limits the activity range of the construction personnel, resulting in the dense distribution of the construction personnel at the work surface and the relatively scattered distribution in other areas. In addition, the complex geological conditions of the tunnel construction site result in various risks such as collapse and water gushing during the construction process, greatly increasing the difficulty of construction. Therefore, compared with other engineering projects, the construction technology of tunnel engineering is more complex, the risk is higher, and the management is more difficult. In addition, once an accident occurs, it is difficult for rescue personnel to timely grasp the location of the trapped personnel, which significantly increases the difficulty of rescue and easily causes greater casualties and property losses.

[0003] Due to the unique physical characteristics of the tunnel, such as closure and restriction, non-line-of-sight propagation, signal shielding, multipath effect, significant signal reflection, and complex structure, conventional positioning devices cannot achieve high-precision positioning, which brings great difficulty to the management of personnel in the tunnel by positioning.

[0004] Therefore, the implementation of personnel positioning in the tunnel construction site is very important for the safety of the construction personnel. By obtaining the position information of the construction personnel in real time and accurately, when the construction personnel enter a dangerous area or an accident occurs, an alarm can be triggered in time for emergency rescue and safety handling. SUMMARY

[0005] Therefore, the present application aims to overcome the shortcomings of the prior art and provides a multi-sensor positioning system in a tunnel and a positioning method for personnel. The positioning technology of UWB (Ultra-Wideband) and visual sensor fusion is adopted, and the IMU (Inertial Measurement Unit) is used for compensation to effectively avoid the error caused by the NLOS (Non-Line-of-Sight) environment. The convolutional neural network and the high-precision positioning algorithm of the RU are used to reduce the influence of the multipath effect on the positioning data. The motion model and the interactive filtering are fused to reduce the influence of the variable motion type of personnel on the positioning data and improve the positioning accuracy.

[0006] To achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0007] The tunnel multi-sensor positioning system comprises a server connected to a wide area network through a wired or wireless connection, a mobile terminal and a PC terminal connected to the server through the wide area network, and a tunnel multi-sensor positioning system, which comprises a wireless bridge fixedly arranged at each anchor point and a plurality of UWB positioning base stations arranged at the anchor points and wiredly connected to the wireless bridge, and a personnel positioning safety helmet worn by a mobile personnel in the tunnel;

[0008] The server is wiredly connected to the wireless bridge of any anchor point in the tunnel, and the wireless bridge at each anchor point wirelessly bridges the wireless bridges at adjacent anchor points to cover the UWB signals of the UWB positioning base stations to the entire tunnel area.

[0009] The personnel positioning safety helmet comprises a UWB positioning tag and an IMU inertial sensor arranged on the safety helmet, and a camera with a gimbal arranged at the top end of the safety helmet, the UWB positioning tag is connected to the UWB positioning base station through a UWB signal, and the IMU inertial sensor and the camera are wirelessly connected to the wireless bridge.

[0010] Further, a time synchronization server is further included, which is connected to the server.

[0011] Further, the anchor points in the tunnel are arranged in a Z-shaped line to ensure that the personnel wearing the personnel positioning safety helmet is connected to at least three UWB positioning base stations of the anchor points at any position in the tunnel.

[0012] Further, the UWB positioning base station comprises a UWB positioning module and a UWB directional antenna, the UWB positioning module is wiredly connected to the UWB directional antenna, and the UWB positioning module comprises a positioning base station DW1000 positioning module.

[0013] Further, the UWB positioning tag of the personnel positioning safety helmet comprises a positioning tag DW1000 positioning module.

[0014] Further, a construction trolley positioning device fixed on a construction trolley in the tunnel and a vehicle positioning device of a vehicle in the tunnel are further included, and the construction trolley positioning device and the vehicle positioning device both comprise a UWB positioning module.

[0015] Further, the server is connected to the mobile terminal or the PC terminal through the wide area network, the mobile terminal is provided with a related APP, and the PC terminal is provided with a related software, which can display real-time or historical personnel position information.

[0016] According to the personnel positioning method of the foregoing tunnel multi-sensor positioning system, the method comprises the following steps:

[0017] S1, the data collected by each anchor UWB base station and the data collected by the wireless gateway, including the RSSI value of the UWB positioning tag, the IMU inertial sensor data and the photo taken by the camera on the anchor, are converged to the server according to the time stamp;

[0018] S2, the server determines whether to use the Y-R-OPA algorithm model first, then use the ACF-DSA algorithm model, or directly use the ACF-DSA algorithm model for calculation according to whether the RSSI value of the UWB positioning tag on the personnel positioning safety helmet is lower than a certain value and whether the proportion of the UWB positioning base station being blocked in the photo taken by the camera on the anchor is greater than a certain percentage.

[0019] S3, the server uses the AIM-EKF algorithm model again on the basis of the calculation in S2 to obtain the real-time positioning data of the personnel positioning safety helmet and stores the data in the message storage queue.

[0020] S4, the mobile terminal or PC terminal connects the server to obtain and consume the message storage queue, and the personnel in the tunnel can display the real-time or historical data in the mobile terminal or PC terminal.

[0021] Further, in S2, the Y-R-OPA algorithm model processes the photo of the anchor collected by the camera using the YOLO11 algorithm, integrates the attention mechanism to quickly determine the position of the UWB positioning base station of the anchor, identifies the blocking condition between the UWB positioning tag on the personnel positioning safety helmet and the UWB positioning base station on the anchor, uses the IOU-RSS fusion position selection method, combines the game theory, compensates the data during the blocking through the IMU, and guarantees the integrity of the data through the comprehensive model algorithm.

[0022] Further, in S2, the ACF-DSA algorithm model is a fusion algorithm that generalizes the ACIR data set through time convolution GAN and identifies the ACIR data through CNN-GRU, and provides the accuracy of positioning.

[0023] Further, in S3, the AIM-EKF algorithm model is an interactive multiple model extended Kalman filter fusion algorithm under the adaptive transition probability matrix, combines multiple models to describe the possible behavior of the system, and the linear model corresponds to the Kalman filter and the nonlinear model corresponds to the extended Kalman filter. Through the interaction between the models and the adjustment of the adaptive transition probability matrix, the AIM-EKF algorithm can effectively process the state estimation problem of the mixed linear and nonlinear dynamic system.

[0024] Compared with the prior art, the beneficial effects of the present application are:

[0025] 1、The present application can realize high-precision positioning of personnel in the tunnel by the fusion of multiple sensors and the application of multiple advanced algorithm models, and meet the strict requirements of tunnel construction on positioning accuracy.

[0026] 2、The present application can still maintain high positioning stability and reduce the occurrence of positioning errors and loss under complex tunnel environments such as signal shielding and multipath effect.

[0027] 3、The present application can acquire and display the position information of personnel in the tunnel in real time, facilitate the management personnel to timely grasp the personnel dynamics, and ensure construction safety.

[0028] 4、The present application is not only suitable for personnel positioning, but also can be used for positioning construction trolleys and construction vehicles, realizing comprehensive monitoring and management of various equipment and personnel in the tunnel. DETAILED DESCRIPTION

[0029] Figure 1 It is a structural schematic diagram of the tunnel multi-sensor positioning system of the present application.

[0030] Figure 2 It is a schematic diagram of the anchor point position in the tunnel of the present application.

[0031] Figure 3 It is a structural block diagram of the UWB positioning base station of the present application.

[0032] Figure 4 It is a structural schematic diagram of the personnel positioning safety helmet of the present application.

[0033] Figure 5 It is a structural block diagram of the UWB positioning tag in the personnel positioning safety helmet of the present application.

[0034] Figure 6 It is a flowchart of the personnel positioning method of the present application.

[0035] In the figure: 1, tunnel; 2, anchor point; 3, wireless network bridge; 4, UWB positioning base station; 41, UWB positioning module; 411, positioning base station power module; 412, positioning base station microprocessor; 413, positioning base station communication module; 414, positioning base station DW1000 positioning module; 42, UWB directional antenna; 5, personnel; 6, personnel positioning safety helmet; 61, safety helmet body; 62, UWB positioning tag; 621, positioning tag power module; 622, positioning tag microprocessor; 623, positioning tag DW1000 positioning chip; 63, IMU inertial sensor; 64, camera with gimbal; 7, construction trolley; 8, construction trolley positioning device; 9, vehicle; 10, vehicle positioning device. DETAILED DESCRIPTION

[0036] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort belong to the scope of protection of the present application.

[0037] As shown in Figure 1 , Figure 2 , Figure 3 , Figure 4 and Figure 5 , the tunnel multi-sensor positioning system comprises a server connected to a wide area network through a wired or wireless connection, a mobile terminal and a PC terminal connected to the server through the wide area network, and a tunnel multi-sensor positioning system, wherein the tunnel multi-sensor positioning system comprises a wireless bridge 3 fixedly arranged at each anchor point 2 in the tunnel 1, a plurality of UWB positioning base stations 4 arranged in wired connection with the anchor point 2 and the wireless bridge 3, and a personnel positioning safety helmet 6 worn by the personnel 5 moving in the tunnel.

[0038] The anchor point 2 is a spatial point for installing the wireless bridge 3 and the UWB positioning base station 4 in the tunnel 1. The anchor point 2 can be arranged on the side wall or top wall of the tunnel 1, or can be arranged by means of a hanger, a support rod, etc. so that the wireless bridge 3 and the UWB positioning base station 4 are at the spatial point. The anchor point 2 can also be arranged on a construction trolley 7, a support, etc.

[0039] At the anchor point 2, at least two wireless bridges 3 and two UWB positioning base stations 4 are arranged back to back to realize wireless connection of the wireless bridge 3 and bidirectional coverage of the UWB signal.

[0040] The anchor points 2 in the tunnel 1 are arranged in a Z-shaped line, i.e. the line of the nearest anchor points 2 is in a Z shape, so as to ensure that the personnel 5 wearing the personnel positioning safety helmet 6 is connected to the UWB positioning base stations 4 of at least three anchor points 2 when the personnel 5 is at any position in the tunnel 1.

[0041] The wireless bridge 3 can be a commercial wireless bridge, and there are many mature products on the market that can be directly selected.

[0042] The UWB positioning base station 4 comprises a UWB positioning module 41 and a UWB directional antenna 42, the UWB positioning module 41 and the UWB directional antenna 42 are separately arranged, the UWB directional antenna 42 can be flexibly arranged to meet different scenarios of different anchor points 2 in the tunnel 1, the UWB positioning module 41 is connected with the UWB directional antenna 42 by wire, the UWB positioning module 41 comprises a positioning base station power module 411, a positioning base station microprocessor 412, a positioning base station network module 413 and a positioning base station DW1000 positioning module 414, the positioning base station microprocessor 412 is connected with the positioning base station power module 411, the positioning base station network module 413 and the positioning base station DW1000 positioning module 414, the positioning base station power module 411 is connected with the positioning base station network module 413 and the positioning base station DW1000 positioning module 414, the positioning base station DW1000 positioning module 414 is connected with the UWB directional antenna by wire, the model of the positioning base station microprocessor 412 is STM32F103C8T6. The UWB directional antenna 42 comprises an antenna array composed of several wideband antennas units and a shielding plate arranged at the back of the antenna array, so that the UWB directional antenna 42 The directivity of the antenna and the focusing performance of the signal are enhanced, and the accuracy of the antenna beam pointing is enhanced. Since the wide-band antenna unit is used, the entire system can have a wide frequency bandwidth. The propagation range and quality of the signal are significantly enhanced.

[0043] The server is connected with the wireless bridge 3 of any anchor point 2 in the tunnel 1 by wire, such as the wireless bridge 3 of the anchor point 2 closest to the mouth of the tunnel 1, the wireless bridge 3 of each anchor point 2 wirelessly bridges the wireless bridge 3 of the adjacent anchor point 2 to cover the UWB signal of the UWB positioning base station 4 to the entire tunnel 1 area.

[0044] The personnel positioning safety helmet 6 comprises a UWB positioning tag 62 and an IMU inertial sensor 63 arranged on the safety helmet 61, and a camera 64 with a gimbal arranged at the top end of the safety helmet 61, wherein the UWB positioning tag 62 and the IMU inertial sensor 63 are arranged on the side surface of the safety helmet 61, and a lithium battery is further arranged on the safety helmet 61 to supply power to the UWB positioning tag 62, the IMU inertial sensor 63 and the camera 64 with the gimbal, the pixel of the camera 64 with the gimbal is at least 4 million, the UWB positioning tag 62 comprises a positioning tag power module 621, a positioning tag microprocessor 622 and a positioning tag DW1000 positioning module 623, the positioning tag microprocessor 622 is connected with the positioning tag power module 621 and the positioning tag DW1000 positioning module 623, the positioning tag DW1000 positioning module 623 is provided with an antenna for transmitting and receiving UWB signals, and the model of the positioning tag microprocessor 622 is the same as that of the UWB positioning base station 4. STM32F103C8T6.

[0045] The UWB positioning tag 62 is connected with the UWB positioning base station 4 through a UWB signal, and the IMU inertial sensor 63 and the camera with a holder 64 are wirelessly connected with the wireless network bridge 3.

[0046] Due to the use of their own time by the wireless network bridge 3, the UWB positioning base station 4, the UWB positioning tag 62, the IMU inertial sensor 63, and the camera with a holder 64 for various reasons, the time of each device is not synchronized, and therefore, a time synchronization server is also needed to unify the time, which is connected with the server, so as to unify the time of each device of the multi-sensor positioning system in the tunnel 1, facilitate the collection of various data in the later period, and improve the positioning accuracy. The time synchronization server can be a commercial time synchronization server, such as an Internet time synchronization server or a Beidou time synchronization server.

[0047] In order to be more consistent with the actual situation of the construction in the tunnel 1, the construction trolley positioning device 8 fixed on the construction trolley 7 in the tunnel 1 and the vehicle positioning device 10 of the construction vehicle 9 in the tunnel 1 are also included, and the construction trolley positioning device 8 and the vehicle positioning device 10 both contain a UWB positioning module, which also uses a DW1000 positioning module. Since the construction trolley 7 is a work trolley, and the vehicle 9 is a construction vehicle, the construction trolley 7 positioning device and the vehicle 9 positioning device both need a shell with good impact resistance.

[0048] Specifically, the construction trolley 7, such as an excavation trolley, an inverted arch, a waterproof trolley, and a secondary lining trolley, displays the position data of the working face, the inverted arch working face, and the secondary lining working face on a mobile terminal or a PC terminal through the positioning of these construction trolleys 7. The specific vehicle 9 positioning device can display the position and movement trajectory of the vehicle 9.

[0049] The method for positioning personnel using the aforementioned multi-sensor positioning system in the tunnel includes the following steps:

[0050] S1, the data collected by each anchor point 2 UWB positioning base station 4 and the data collected by the wireless gateway 3, including the RSSI value of the UWB positioning tag 62, the IMU inertial sensor 63 data, and the photo taken by the camera with a holder 64 against the anchor point 2, are converged to the server according to the time stamp.

[0051] For example, the UWB positioning tag 62 on the personnel positioning safety hat 7 worn by a person 5 in the tunnel 1 is always connected with the UWB positioning base station 4 of several anchor points 2, the UWB positioning base station 4 collects the RSSI value of the personnel positioning safety hat 6, at the same time, the camera 64 with a pan-tilt head on the personnel positioning safety hat 6 rotates and shoots the photo with the anchor point 2. According to the time stamp, that is, a certain point in time, at this time point, the RSSI value of the personnel positioning safety hat 6 collected on each anchor point 2, the photo and the IMU value are converged into the server to form the time stamp positioning data.

[0052] S2, the server determines whether to use the Y-R-OPA algorithm model first, then the ACF-DSA algorithm model, or directly use the ACF-DSA algorithm model for calculation according to whether the RSSI value of the UWB positioning tag 62 on the personnel positioning safety hat 6 to a certain anchor point 2 is lower than a certain value, and whether the proportion of the UWB positioning base station 4 being blocked in the photo of the anchor point 2 shot by the camera 64 with a pan-tilt head on the personnel positioning safety hat 6 is greater than a certain percentage.

[0053] The Y-R-OPA algorithm model processes the photo of the anchor point 2 collected by the camera 64 with a pan-tilt head using the YOLO11 algorithm, quickly determines the position of the UWB positioning base station 4 of the anchor point 2 by integrating the attention mechanism, identifies the blocking condition between the UWB positioning tag 62 on the personnel positioning safety hat 6 and the UWB positioning base station 4 on the anchor point 2, and when the blocking proportion is greater than 40%, uses the IOU-RSS fusion position selection method, combines the game theory, compensates the data during the blocking through the IMU inertial sensor 63, and ensures the integrity of the comprehensive model algorithm of the data.

[0054] Y-R-OPA not only comprehensively considers the factors of blocking and signal strength, but also adopts the game theory strategy

[0055] to realize the optimal selection of the stopping point. By using the IMU inertial sensor for data compensation processing, the overall performance and robustness of the system are significantly improved.

[0056] The network architecture of YOLO11 consists of an improved backbone network, an improved neck network

[0057] Neck, a head network Head. The backbone network obtains image features, and the neck network further fuses and extracts features, and finally the head network generates a prediction result. The embodiment improves the YOLO11 algorithm model by introducing a lightweight attention mechanism, helps to focus on important information of the model, improves detection accuracy, and reduces false positive rate. The lightweight convolutional attention mechanism (Convolutional Block Attention Module, CBAM) is composed of a channel attention module (Channel Attention Module, CAM) and a spatial attention module (Spatial Attention Module, SAM).

[0058] The channel attention module captures channel global information to obtain pooled features by global average pooling and global maximum pooling of the channel feature map, and obtains channel attention weights by a shared multilayer perceptron. The channel weighted feature map is obtained by multiplying the original feature map by the channel attention weights. The spatial attention module performs average pooling and maximum pooling processing on the channel weighted feature map, and a convolutional layer connects the pooled features to generate spatial attention weights. The final feature map is obtained by multiplying the channel weighted feature map by the spatial attention weights. The introduction of the convolutional attention module can make the model pay more attention to important features of the base station, thereby improving the perception and discrimination ability and speed of the model. By continuously training the channel and spatial information of the model, the importance of the base station is strengthened, and the detection accuracy and robustness are improved. The embodiment integrates the channel attention module and the spatial attention module on the neck network.

[0059] Let the output of a layer of the improved YOLO11 neck network be X, and the output after the CBAM module be:

[0060]

[0061] Equation One

[0062] For each prediction box, the model outputs its position confidence and class probability. Let P be the final prediction result: Equation Two

[0063] The improved YOLO11 discriminates target objects and potential occlusions, and outputs core data such as the position of the detection box and its confidence. The detection result of the improved YOLO11 clearly labels the identified target and occlusion. According to the calculated and evaluated occlusion, the pan-tilt rotation angle is initially determined, and the IOU and RSSI signals at different positions are obtained by rotation.

[0064] ​The severity of the occlusion is evaluated according to the proportion of the area of the target object covered by the occluder or the degree of overlap between the occluder and the target object bounding box. The interaction ratio is a key indicator in the field of computer vision and target detection for measuring the degree of overlap between the predicted bounding box and the real bounding box.

[0065] IOU is the ratio of the intersection area of the predicted bounding box and the real bounding box to the union area, i.e. IOU = (intersection area) / (union area). The calculation formula of IOU can be specifically expressed as:

[0066] Formula three

[0067] Where A is the predicted bounding box, B is the real bounding box, A∩B is the intersection area of the two, and A∪B is the union area of the two. In two-dimensional target detection, these areas are usually the areas of rectangles.

[0068] The IOU data and RSSI data are normalized and time-synchronized to ensure that they are compared on the same scale. Traverse all candidate positions and calculate the comprehensive score of each position. Select the position with the highest comprehensive score as the optimal position and output the coordinates of the optimal position, the corresponding IOU value, RSSI value and comprehensive score. Each candidate position is regarded as a participant in the game, and IOU and RSSI are regarded as components of the participant's payoff function or payment function. The core idea of game theory is to find a strategy combination to ensure that all participants can reach a balanced state, i.e. no one can get better returns by adjusting their own strategy.

[0069] Define the participant, each candidate position i, where i ∈ {1,2,..., n}, the strategy of each position n can be represented as The payoff function u i of each position i is defined as the comprehensive score S i , i.e.

[0070] Formula four

[0071] Here, the payoff depends on the strategy choices of all positions.

[0072] Introduce the concept of Nash equilibrium, if for each position i, there is:

[0073] Formula five

[0074] Where represents the strategy combination of all other positions except position .

[0075] Each position Update its own strategy according to the current IOU and RSSI and other location strategies. The algorithm determines whether a Nash equilibrium is reached by comparing the benefit change of the current iteration and the last iteration. When the benefit change is less than a certain threshold, it is considered in this embodiment that the equilibrium is reached. That is

[0076] Formula six

[0077] For all , the output optimal position and its corresponding IOU, RSSI and comprehensive score, i.e. (i * ,IOU i ,RSSI i ,S i ), confirm the final resting point of the gimbal, the gimbal rotates, and waits for the next cycle calculation.

[0078] Due to non-line-of-sight obstruction, gimbal rotation and other reasons, the UWB positioning signal may be disturbed, and then the positioning data is inaccurate or completely lost. The IMU inertial sensor provides real-time acceleration and angular velocity information to correct the measurement error and drift of the UWB.

[0079] Sparse matrix refers to a matrix in which most elements are zero. In Kalman filtering, the state transition matrix, the observation matrix and the covariance matrix may be sparse. Using the characteristics of sparse matrix can significantly reduce the amount of calculation. Sparse matrix compresses sparse rows (Compressed Sparse Row, CSR) to store data in a compressed storage manner. The CSR data expression mainly includes: values, column indices, row offsets, etc. The non-zero elements in the matrix are saved as the values array. The column index of the non-zero element in the corresponding position of the values array is saved as the column indices array. The index of the row offsets array represents the row index of the first non-zero element of each row, and the value is the index of the values array. The last element value is the number of non-zero elements. The space occupied by the CSR storage format under double-precision floating-point number is

[0080] Formula seven

[0081] Wherein, is the number of rows of the sparse matrix, is the number of non-zero elements in the sparse matrix.

[0082] IMU fusion UWB positioning under improved Kalman filtering, the position , velocity , quaternion​ acceleration bias velocity bias Modeling as a state variable, the differential equation can be expressed as

[0083] Equation eight

[0084] Equation nine

[0085] wherein represents the position prediction value of the state variable by the IMU, and △t is the sampling time of the IMU.

[0086] The position prediction value of the state variable and the distance l between each base station of the UWB are calculated i (i is the number of base stations), the position prediction of the state variable is obtained by equation ten, and the distance l is obtained by the distance calculation formula between two points i The pseudo-range d output by the UWB system i is subtracted from the predicted distance l i as the measurement vector of the filtering equation Optimize the state variable, and update the measurement matrix at the same time. The state variable update formula can be expressed as:

[0087] Equation ten

[0088] wherein,

[0089] Equation eleven

[0090] The ACF-DSA algorithm model generalizes the ACIR data set through time convolution GAN, and is a kind of fusion algorithm for identifying ACIR data through CNN-GRU, and provides the accuracy of positioning.

[0091] The algorithm model of ACF-DSA improves the accuracy of positioning. In the data processing process, mainly pay attention to the wave peak and the state before and after it, so as to distinguish normal signals and abnormal signals. By taking the wave peak as the focus, the feature extraction technology of expanding 200ms window is adopted to obtain the core data before and after the wave peak. Combined with manual and automatic marking methods, abnormal data is identified. In this embodiment, the abnormal signals and normal signals are deeply cleaned and extracted, and the generator of time convolution GAN is used to generate complex simulation data, and the discriminator is used to enhance the identification ability of abnormal data. The abnormal signal is identified by the algorithm model combining convolutional neural network and gated recurrent unit. In this embodiment, through the ACF-DSA algorithm, the ACIR detection accuracy is as high as 99%, and the positioning precision is improved.

[0092] S3, the server calculates again on the basis of S2, adopts the AIM-EKF algorithm model to calculate, obtains the real-time positioning data of the personnel positioning safety helmet 6, and stores according to the message storage queue.

[0093] The AIM-EKF algorithm model is an interactive multi-model extended Kalman filter fusion algorithm under an adaptive transition probability matrix, combines multiple models to describe the possible behavior of the system, a linear model corresponds to a Kalman filter, and a nonlinear model corresponds to an extended Kalman filter. Through the interaction between the models and the adjustment of the adaptive transition probability matrix, the AIM-EKF algorithm can effectively handle the state estimation problem of mixed linear and nonlinear dynamic systems.

[0094] For the AIM-EKF algorithm, define models, each of which can be linear or nonlinear, for each model initialize its state estimate and covariance matrix , initialize the model probability , indicating the probability that the current model is the true model.

[0095] For each model , the linear model uses Kalman filter state prediction and the nonlinear equation uses extended Kalman filter state prediction as follows, and the extended Kalman filter needs to perform nonlinear transformation first.

[0096] Equation twelve

[0097] Equation thirteen

[0098] where, is the state transition matrix of the linear model , and is the state transition function of the nonlinear model , used to predict the next state.

[0099] The covariance prediction of the linear model and the covariance prediction of the nonlinear model are as follows:

[0100] Equation fourteen

[0101] Equation fifteen

[0102] where, is the process noise covariance matrix of the model .

[0103] Calculate the mixing probability , which represents the joint effect of the model transition probability and the previous model probability:

[0104] Equation Sixteen

[0105] where, is the transition probability from model to model

[0106] Compute the overall state prediction and covariance matrix:

[0107] Equation Seventeen

[0108] For each model use the corresponding Kalman filter or extended Kalman filter for measurement update:

[0109] Equation Eighteen

[0110] Equation Nineteen

[0111] Update the state estimate:

[0112] Equation Twenty

[0113] Update the covariance matrix:

[0114] Equation Twenty-One

[0115] where, is the observation matrix of model is the observation noise covariance matrix of model is the observation value, is the identity matrix. Update the model probability

[0116]

[0117] Equation Twenty-Two

[0118] Equation Twenty-Three

[0119] Compute the final state estimate and covariance matrix:

[0120] Equation Twenty-Four

[0121] Equation Twenty-Five

[0122] ​​​​The above steps are repeated to perform state estimation and data fusion of the next moment. The AIM-EKF algorithm can switch between different motion models and combine the observation data of UWB and IMU to realize accurate estimation of the target state. The use of the adaptive transition probability matrix further improves the flexibility and robustness of the algorithm.

[0123] S4, the mobile terminal or PC terminal connects the server, acquires and consumes the message storage queue, and personnel 5 in tunnel 1 can display real-time or historical data in the mobile terminal or PC terminal.

[0124] The positioning accuracy of the UWB is improved, the practical scene application significance brought is extraordinary, a safe and civilized construction environment is created, and not only the safety of the construction personnel is guaranteed, but also the engineering progress is accelerated.

[0125] Finally, it should be explained that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit it, and other modifications or equivalent replacements of the technical solutions of the present application made by those skilled in the art should be covered in the scope of the claims of the present application as long as they do not deviate from the spirit and scope of the technical solutions of the present application.

Claims

1. A multi-sensor positioning system in a tunnel, comprising a server connected to a wide area network through a wired or wireless connection, a mobile terminal and a PC terminal connected to the server through the wide area network, characterized in that: The tunnel multi-sensor positioning system comprises a wireless network bridge fixedly arranged on each anchor point, a plurality of UWB positioning base stations arranged in wired connection with the wireless network bridge at the anchor point, and a personnel positioning safety helmet worn by a person moving in the tunnel, wherein the anchor point is a spatial point for installing the wireless network bridge and the UWB positioning base station in the tunnel; the server is in wired connection with the wireless network bridge arranged at any anchor point in the tunnel, and the wireless network bridge at each anchor point wirelessly bridges the wireless network bridge at the adjacent anchor point to cover the UWB signal of the UWB positioning base station to the entire tunnel area; the personnel positioning safety helmet comprises a UWB positioning tag and an IMU inertial sensor arranged on the positioning safety helmet, and a camera with a holder arranged at the top end of the positioning safety helmet, wherein the UWB positioning tag is connected with the UWB positioning base station through the UWB signal, and the IMU inertial sensor and the camera are wirelessly connected with the wireless network bridge; further comprising a personnel positioning method, comprising the following steps: S1, the data collected by the UWB base station arranged at each anchor point and the data collected by the wireless gateway, including the RSSI value of the UWB positioning tag, the IMU inertial sensor data and the photo taken by the camera facing the anchor point, are converged to the server according to the time stamp; S2, the server determines whether to first use the Y-R-OPA algorithm model, then use the ACF-DSA algorithm model, or directly use the ACF-DSA algorithm model for calculation according to whether the RSSI value of the UWB positioning tag on the personnel positioning safety helmet at a certain anchor point is lower than a certain value, and whether the proportion of the UWB positioning base station being blocked in the photo taken by the camera on the personnel positioning safety helmet facing the anchor point is greater than a certain percentage; S3, the server further uses the AIM-EKF algorithm model for calculation on the basis of the calculation in S2 to obtain the real-time positioning data of the personnel positioning safety helmet, and stores the data in a message storage queue; S4, a mobile terminal or a PC terminal connects the server to obtain and consume the message storage queue, and the personnel in the tunnel can display the real-time or historical data in the mobile terminal or the PC terminal; In S2, the Y-R-OPA algorithm model processes the photo of the anchor point collected by the camera using the YOLO11 algorithm, quickly determines the position of the UWB positioning base station of the anchor point by fusing the attention mechanism, identifies the blocking condition between the UWB positioning tag on the personnel positioning safety helmet and the UWB positioning base station on the anchor point, and uses the IOU-RSS fusion position selection method combined with the game theory to compensate the data during the blocking period through the IMU to ensure the integrity of the data. In S2, the ACF-DSA algorithm model is a fusion algorithm for generalizing the ACIR data set through time convolution GAN and identifying the ACIR data through CNN-GRU to provide the accuracy of positioning.

2. The in-tunnel multi-sensor positioning system of claim 1, wherein: Further comprising a time synchronization server connected with the server.

3. The in-tunnel multi-sensor positioning system of claim 1, wherein: The anchor points in the tunnel are arranged in a Z-shaped line to ensure that the personnel wearing the personnel positioning safety helmet are connected with the UWB positioning base stations of at least three anchor points at any position in the tunnel.

4. The in-tunnel multi-sensor positioning system of claim 1, wherein: The UWB positioning base station comprises a UWB positioning module and a UWB directional antenna, the UWB positioning module is wiredly connected to the UWB directional antenna, and the UWB positioning module comprises a positioning module DW1000 positioning module.

5. The in-tunnel multi-sensor positioning system of claim 1, wherein: The UWB positioning tag of the personnel positioning safety helmet comprises a positioning tag DW1000 positioning module.

6. The in-tunnel multi-sensor positioning system of claim 1, wherein: The construction trolley positioning device fixed on the construction trolley in the tunnel and the vehicle positioning device of the vehicle in the tunnel are also included, and the construction trolley positioning device and the vehicle positioning device both comprise a UWB positioning module.

7. The in-tunnel multi-sensor positioning system of claim 6, wherein: In S3, the AIM-EKF algorithm model is an interactive multiple model extended Kalman filter fusion algorithm under a self-adaptive transition probability matrix, a plurality of models are combined to describe the possible behavior of the system, a linear model corresponds to a Kalman filter, a nonlinear model corresponds to an extended Kalman filter, and through the interaction between the models and the adjustment of the self-adaptive transition probability matrix, the AIM-EKF algorithm can effectively solve the state estimation problem of a mixed linear and nonlinear dynamic system.

Images