Short-time startup satellite positioning prediction method for smart helmets

Abstract

Embodiments of this application provide a method, apparatus, device, and computer-readable storage medium for predicting satellite positioning during short-term power-up of a smart helmet. The method includes acquiring positioning sequence data; mapping the positioning sequence data to a constructed environmental feature matrix to obtain local environmental features; inputting the positioning sequence data and local environmental features into a trained prediction model to obtain a positioning prediction value; and, based on the positioning prediction value, completing a prediction of the current position. This method significantly improves positioning prediction accuracy while reducing computational costs.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of data processing, and in particular to a short-time power-on satellite positioning prediction method, apparatus, device, and computer-readable storage medium for a smart helmet. Background Art

[0002] With the development of satellite positioning technology and personal protective equipment, companies have begun to try to install satellite positioning modules on smart helmets to achieve intelligent management of the positioning and attendance of construction workers.

[0003] However, positioning modules generally consume high power. If left on for extended periods, the smart helmet's battery life would be dramatically reduced. Therefore, the module can only be periodically activated for short periods of time to collect and collect positioning data, then enters sleep mode. Furthermore, due to the multipath effect of the building environment and satellite signal obstruction, positioning errors during short periods of activation in actual applications often exceed 20 meters, severely limiting positioning reliability.

[0004] Current methods for improving satellite positioning accuracy include:

[0005] Through built-in inertial navigation hardware;

[0006] An LSTM neural network integrated with a self-attention mechanism is used to predict the satellite clock bias;

[0007] Satellite positioning based on the assistance of long short-term memory network;

[0008] In the above method, the positioning accuracy is improved by built-in inertial navigation hardware, which greatly increases the hardware cost;

[0009] The LSTM neural network integrated with the self-attention mechanism predicts the satellite clock bias. This requires obtaining satellite clock bias data, which is costly and consumes a lot of power, making it unsuitable for mass installation and daily use on helmet terminals.

[0010] Satellite positioning assisted by long short-term memory networks requires the terminal device to have a large amount of data storage space and model training capabilities, and is not suitable for helmet positioning terminals with limited storage and computing capabilities.

[0011] In summary, how to build a satellite positioning method that is fast, small in size, requires low computing power, and has high accuracy is a problem that needs to be solved urgently. Summary of the Invention

[0012] According to the embodiments of the present application, a short-time power-on satellite positioning prediction solution for smart helmets is provided. Through the environmental basic deviation correction mechanism, the systematic impact of environmental factors such as weather, ionospheric error, and on-site electromagnetic interference on the positioning of the same area is corrected; matrix modeling of environmental characteristics is introduced to improve the accuracy of positioning prediction; data processing is performed by full connection and parameter quantization in time steps, which greatly reduces the demand for computing resources (requiring less than 5KB of RAM), making the construction of the overall solution more lightweight.

[0013] In a first aspect of the present application, a method for predicting satellite positioning during short-term power-on of a smart helmet is provided. The method comprises:

[0014] Obtain positioning sequence data;

[0015] Mapping the positioning sequence data to the constructed environmental feature matrix to obtain local environmental features;

[0016] The positioning sequence data and local environmental features are input into a trained prediction model to obtain a positioning prediction value; based on the positioning prediction value, the current position is predicted.

[0017] Furthermore, before inputting the positioning sequence data into the trained prediction model, the method further includes:

[0018] Cleaning the positioning sequence data to remove abnormal values in the positioning sequence data to obtain first positioning sequence data;

[0019] Correcting the longitude and latitude in the first positioning sequence data using the acquired environmental deviation value to obtain second positioning sequence data;

[0020] performing differential processing on the second positioning sequence data to obtain third positioning sequence data;

[0021] The third positioning sequence data is standardized, and the numerical range of the third positioning sequence data is compressed to a preset range to obtain target positioning sequence data input into the prediction model.

[0022] Furthermore, the step of correcting the longitude and latitude in the first positioning sequence data by using the acquired environmental deviation value to obtain the second positioning sequence data includes:

[0023] Subtracting the first positioning sequence data from the collected target data to obtain a plurality of difference values;

[0024] Calculating an average of the multiple differences to obtain an environmental deviation;

[0025] Based on the environmental deviation, the first positioning sequence data is corrected to obtain second positioning sequence data.

[0026] Furthermore, the prediction model can be constructed in the following way:

[0027] Generate a training sample set, wherein the training sample includes positioning sequence data and local environment features with annotation information; the annotation information is a positioning prediction value;

[0028] The prediction model is trained using samples in the training sample set, with sample files as input and positioning prediction values as output. When the unification rate between the output positioning prediction value and the labeled positioning prediction value meets a preset threshold, the training of the prediction model is completed.

[0029] Furthermore, the environmental feature matrix includes:

[0030] With the collection point as the center, an environmental feature matrix is constructed based on the construction project drawings;

[0031] The value of each grid in the environmental feature matrix is the normalized value of the building height at that location:

[0032] ;

[0033] in, For the building height.

[0034] Furthermore, the loss function of the prediction model includes:

[0035] Calculate the mean absolute error of longitude and latitude separately;

[0036] The longitude mean absolute error and latitude mean absolute error are summed to obtain the loss function of the prediction model.

[0037] Furthermore, the mean absolute error of longitude and / or latitude can be calculated as follows:

[0038] ;

[0039] in, is the predicted value of longitude and latitude;

[0040] is the true value of longitude and latitude.

[0041] In a second aspect of the present application, a short-time startup satellite positioning prediction device for a smart helmet is provided. The device comprises:

[0042] An acquisition module, used for acquiring positioning sequence data;

[0043] A processing module, configured to map the positioning sequence data to a constructed environmental feature matrix to obtain local environmental features;

[0044] The positioning module is used to input the positioning sequence data and local environmental characteristics into the trained prediction model to obtain a positioning prediction value; based on the positioning prediction value, complete the prediction of the current position.

[0045] In a third aspect of the present application, an electronic device is provided, comprising: a memory and a processor, wherein the memory stores a computer program, and the processor implements the above method when executing the program.

[0046] In a fourth aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the method according to the first aspect of the present application is implemented.

[0047] The short-time power-on satellite positioning prediction method for a smart helmet provided in an embodiment of the present application obtains positioning sequence data; maps the positioning sequence data to a constructed environmental feature matrix to obtain local environmental features; inputs the positioning sequence data and local environmental features into a trained prediction model to obtain positioning prediction values; and completes the prediction of the current position based on the positioning prediction values, thereby significantly improving the accuracy of positioning predictions while reducing computational costs.

[0048] It should be understood that the contents described in the Summary of the Invention are not intended to limit the key or important features of the embodiments of the present application, nor are they intended to limit the scope of the present application. Other features of the present application will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The above and other features, advantages and aspects of the embodiments of the present application will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, wherein:

[0050] Figure 1 Flowchart of a short-time power-on satellite positioning prediction method for a smart helmet according to an embodiment of the present application;

[0051] Figure 2 Schematic diagram of environment matrix modeling according to an embodiment of the present application;

[0052] Figure 3 Schematic diagram of environment matching according to an embodiment of the present application;

[0053] Figure 4 Schematic diagram of the neural network model structure according to an embodiment of the present application;

[0054] Figure 5 1 is a block diagram of a short-time power-on satellite positioning prediction device applied to a smart helmet according to an embodiment of the present application;

[0055] Figure 6 A schematic diagram of the structure of a terminal device or server suitable for implementing an embodiment of the present application. DETAILED DESCRIPTION

[0056] To make the purpose, technical solutions, and advantages of the embodiments of the present disclosure more clear, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present disclosure.

[0057] In this document, the term "and / or" simply describes a relationship between related objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document generally indicates that the related objects are in an "or" relationship.

[0058] Figure 1 A flow chart of a method for predicting satellite positioning for a short-time power-on of a smart helmet according to an embodiment of the present disclosure is shown. The method includes:

[0059] S110, obtaining positioning sequence data.

[0060] In some embodiments, positioning sequence data can be acquired through a positioning module pre-installed in the helmet. For example, after the positioning module is activated, initialized, and successfully calculated for longitude and latitude, 300 seconds of continuous data reading at a frequency of 1 Hz is considered complete, resulting in positioning sequence data. Specifically, the positioning module is activated, actively searches for and captures satellite signals, demodulates and decodes the captured satellite signals, extracts key parameters such as time and orbital parameters, and calculates them to obtain NMEA standard positioning data. The data is then output as a sequence of raw positioning data with a time interval of 1 second.

[0061] S120, mapping the positioning sequence data to the constructed environmental feature matrix to obtain local environmental features.

[0062] In some embodiments, the environment surrounding each acquisition point can be modeled and an environmental matrix can be output to allow the model to capture the impact of environmental characteristics on positioning error. In this disclosure, a square area with a side length of 200 meters is modeled with the acquisition point as the center (the maximum error after power-on does not exceed 100 meters). During modeling, the grid is divided into 10-meter granularity, and the 200-meter-by-200-meter square area is divided into 20-by-20 grids. The value of each grid represents the normalized value of the building height at that location:

[0063] ;

[0064] in, For the building height.

[0065] The environment matrix modeled around a certain collection point is as follows: Figure 2 As shown, a blank grid indicates that the building height at that location is 0. Since the center point is the data collection point and its actual longitude and latitude are known, the longitude and latitude range of each grid can be calculated, that is, the specific location of any given longitude and latitude in the grid can be calculated.

[0066] Furthermore, when the environment is more complex, that is, the construction project is not a standard square and the scale is usually larger than 200 meters, actual modeling can be carried out based on the construction project drawings, and the grid is divided into 10-meter granularity, and the building heights of each grid are normalized. In this case, the longitude and latitude of the four vertices of the grid in the lower left corner must be measured to subsequently calculate the longitude and latitude range of each grid, and thus the specific location of any given longitude and latitude in the grid.

[0067] It should be noted that the above side length range and grid granularity are only examples, and the specific values can be adjusted according to the actual application scenario.

[0068] In some embodiments, in order to obtain more accurate positioning prediction data and solve problems such as the large absolute value of the longitude and latitude data in the positioning sequence data and the small effective change (for example, the longitude and latitude are 121.456789°E, 31.234567°N, and the change is approximately on the order of 0.00001), the positioning sequence data obtained in step S110 can be cleaned first.

[0069] Specifically, the positioning sequence data is preliminarily cleaned to remove abnormal values in the positioning sequence data to obtain the first positioning sequence data:

[0070] For example, a total of n valid longitude and latitude values for the previous n seconds (e.g., the previous 20 seconds) are continuously read at a frequency of 1 Hz. Data is cleaned while being collected, discarding any data with a distance threshold between the previous x seconds and the nth second that exceeds a threshold, or data with completely identical longitude and latitude for the previous x seconds. A preset segment of data is then retrieved from the nth second onward, ensuring that the data size is n and that the data is usable, thereby obtaining the first positioning sequence data. The X, distance threshold, and / or preset segment can be pre-set based on the actual application scenario.

[0071] Furthermore, a reference device can be deployed at the construction project site. The reference device's positioning is always on, and the environmental deviation (the difference between the precise longitude and latitude and the positioning test longitude and latitude) is reported to the server every minute. After the hard hat completes the positioning data collection, it obtains the latest environmental deviation value of the current construction project from the server. Based on the obtained environmental deviation value, the longitude and latitude in the first positioning sequence data are corrected to obtain the second positioning sequence data. Compared with the existing technology, the present disclosure corrects the systematic impact of environmental factors such as weather, ionospheric errors, and on-site electromagnetic interference on the positioning of the same area:

[0072] The environmental deviation value can be determined using the reference value measured by the RTK device. Specifically, the environmental deviation is obtained by subtracting and averaging the reference values measured by the reference device and the RTK device. The reference device is a helmet positioning module that is fixed and continuously operational, and its actual position can be measured using RTK.

[0073] ;

[0074] ;

[0075] in, and The latitude and longitude information collected by the PTK equipment respectively;

[0076] and The first latitude and longitude information respectively;

[0077] and Second latitude and longitude information respectively;

[0078] The first positioning sequence data is corrected based on the environmental deviation to obtain the second positioning sequence data:

[0079] ;

[0080] ;

[0081] in, and They are the latitude and longitude collected by the experimental equipment (the positioning module in the helmet).

[0082] Furthermore, the second positioning sequence data is differentially processed to obtain the third positioning sequence data. That is, the coordinates collected in the first n seconds (e.g., 20 seconds) of each data set can be added to the real coordinates, and the longitude and latitude of a total of n+1 data sets can be subtracted in sequence to convert them into the longitude and latitude changes between adjacent points. , , used to eliminate the influence of absolute coordinates and enhance the sensitivity of the model.

[0083] Furthermore, the third positioning sequence data is normalized to compress the numerical range of the third positioning sequence data to a preset range, thereby obtaining target positioning sequence data input to the prediction model. For example, the third positioning sequence data is subjected to Min-Max normalization to compress the numerical range to the interval [-1, 1] to accelerate model convergence:

[0084] ;

[0085] ;

[0086] ;

[0087] ;

[0088] ;

[0089] ;

[0090] In some embodiments, as Figure 3 As shown, the longitude and latitude of each time step in the target positioning sequence data are mapped to the corresponding grid of the environment matrix, and the values of the upper, lower, left, and right adjacent grids of the grid are read as the four environmental feature vectors (North_i, South_i, West_i, East_i) of the time step.

[0091] S130, inputting the positioning sequence data and local environmental features into a trained prediction model to obtain a positioning prediction value; and completing the prediction of the current position based on the positioning prediction value.

[0092] In some embodiments, the prediction model may be trained as follows:

[0093] Generate a training sample set, wherein the training sample includes positioning sequence data and local environment features with annotation information; the annotation information is a positioning prediction value;

[0094] The prediction model is trained using samples in the training sample set, with sample files as input and positioning prediction values as output. When the unification rate between the output positioning prediction value and the labeled positioning prediction value meets a preset threshold, the training of the prediction model is completed.

[0095] Furthermore, when generating the training sample set, to improve the accuracy of the neural network model's positioning and prevent the inherent errors of a single device from affecting the dataset, seven positioning modules of the same model were used for data collection in this disclosure. Five of these modules were used for data collection, while the remaining two modules were kept on continuously as environmental reference devices for measuring environmental deviations. The equipment was placed on a 1.5-meter-high wooden frame for positioning data collection, simulating the height of the positioning equipment when a construction worker wearing a hard hat actually positions the equipment. During data collection, the positioning module was activated and placed on the frame. After the module initialized and successfully calculated the longitude and latitude, a continuous reading of 300 seconds at a 1 Hz frequency was considered complete. After data collection at a point was completed, the longitude and latitude of the device's location were measured using an RTK device. The RTK error was less than 0.1 m, two orders of magnitude lower than the positioning module's error (10 m), making it negligible. Therefore, the RTK measurement value was directly used as the true value of the device's position. In order to cover various environments during construction, this disclosure selects data collection points between 120 and 122 degrees east longitude and 31 and 32 degrees north latitude, with no less than 10 collection points, covering a variety of environments such as open areas, close buildings on one side, close buildings on both sides, buildings surrounded by buildings, and the edge of dense woods. In order to collect as much data as possible for model training, 5 rounds of data collection were carried out at 10 collection points, 10 times per round, and each time 5 devices collected a set of positioning data 300 seconds after startup. In the end, a total of 2,500 sets of data were collected for neural network model training. The normally open module reads the measured longitude and latitude after 1 hour of opening for each round of collection at each collection point, and uses RTK to collect its accurate longitude and latitude.

[0096] In some embodiments, the prediction models of the present disclosure are as follows Figure 4 As shown in the figure, it is a multi-layer neural network structure. The input layer contains 19 time steps between the 1st and 20th seconds, with 6 nodes in each time step, totaling 114 nodes. The first hidden layer is a feature cross layer, which is divided into 19 time steps, with 8 nodes in each time step. The 8 nodes in each time step are fully connected to the 6 nodes in the corresponding time step of the input layer, and there is no connection between different time steps. The second hidden layer is a fully connected layer with a total of 16 nodes, which is fully connected to the 8×19=152 nodes in the feature cross layer. The output layer is a fully connected layer with a total of 2 nodes, representing the true position longitude after difference and normalization. and latitude The loss function of the model uses the mean absolute error (MAE), and MAE is calculated for longitude and latitude separately:

[0097] ;

[0098] in, is the predicted value of longitude and latitude;

[0099] is the true value of longitude and latitude;

[0100] The sum of longitude MAE and latitude MAE is used as the loss function of the model.

[0101] The algorithm disclosed in this paper adopts a time-stepped full-connection and parameter quantization method, requiring less than 5KB of RAM, making it lighter, requiring lower computing resources, and adaptable to embedded hardware.

[0102] After model training and validation are complete, the model can be deployed and run on the Hezhou Air820UG communication module, which integrates Beidou, GPS, and GLONASS multi-mode satellite positioning capabilities. First, the model needs to be dynamically quantized using methods provided by PyTorch. Then, it is exported as an ONNX file containing quantization information. Finally, the model is converted to LUA using the ONNX2Lua toolchain and burned into the embedded hardware, completing the model download. Furthermore, the environmental matrix of the construction project model must be stored on the terminal. After the model is downloaded, the predict method is implemented in LUA on the hardhat side to perform positioning prediction. This method receives the positioning sequence input from the positioning module, performs differential and normalization preprocessing, then performs environmental matching and feature concatenation. The concatenated vector serves as the model input. A lightweight forward propagation calculation is then performed using the downloaded model parameters to obtain the model prediction value. Finally, denormalization and de-differentiation are performed to obtain the predicted longitude and latitude values, completing the model operation. Specifically, the generated positioning data sequence and environmental matrix are input into the predict algorithm implemented on the hardhat side, which outputs the predicted longitude and latitude.

[0103] Furthermore, the ROM and RAM usage of the prediction model can be evaluated as follows. In the feature cross-layer, each time step has 6 × 8 = 48 weight parameters and 8 bias parameters, totaling 56. For 19 time steps, this adds up to 56 × 19 = 1064 parameters. The fully connected layer has 152 × 16 + 16 = 2448 parameters. The output layer has 16 × 2 + 2 = 34 parameters. The model has 3564 parameters, which are stored in 8-bit fixed-point format after quantization, occupying approximately 3.5KB of ROM. Taking a large 1km*1km construction site as an example, the environment matrix has 10,000 grid parameters, which are stored in 8-bit fixed-point format after quantization, occupying approximately 9.8KB of ROM. A total of 13.3KB of ROM is required. When running the algorithm in memory, the 3.5KB model parameters must first be read into memory. Then, the input layer has 6×19=114 variables, the feature cross layer has 8×19=152 variables, the fully connected layer has 16 variables, and the output layer has 2 variables, totaling 282 temporary variables. After quantization, they are stored in 32-bit fixed-point format, occupying approximately 1.1KB of RAM, and a total of 4.6KB of RAM. Both ROM and RAM usage are within the hardware parameter range of the Hezhou Air820UG. It can be seen that the prediction model constructed through this disclosure can run smoothly on the helmet terminal.

[0104] According to the embodiments of the present disclosure, the following technical effects are achieved:

[0105] Through neural networks, the relationship between positioning change trends and actual positions during short-term power-on periods in historical data is captured, ultimately enabling positioning prediction and improving the positioning accuracy of satellite positioning modules during short-term power-on periods.

[0106] The model is trained using environmental features, which can fully capture the relationship between positioning changes and real coordinates during short-term power-on under different environmental characteristics, significantly improving the model's prediction results.

[0107] The prediction model is lightweight, has fast computing speed, small size, low computing power requirements, and is suitable for embedded hardware.

[0108] It should be noted that for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that this application is not limited by the order of the actions described, because according to this application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all optional embodiments, and the actions and modules involved are not necessarily required by this application.

[0109] The above is an introduction to the method embodiment. The following is a device embodiment to further illustrate the solution described in this application.

[0110] Figure 5A block diagram 500 of a short-time power-on satellite positioning prediction device for a smart helmet according to an embodiment of the present application is shown. Figure 5 Shown include:

[0111] An acquisition module 510 is used to acquire positioning sequence data;

[0112] A processing module 520 is configured to map the positioning sequence data to a constructed environmental feature matrix to obtain local environmental features;

[0113] The positioning module 530 is used to input the positioning sequence data and local environmental features into the trained prediction model to obtain a positioning prediction value; and complete the prediction of the current position based on the positioning prediction value.

[0114] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the described module can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0115] Figure 6 A schematic diagram of the structure of a terminal device or server suitable for implementing an embodiment of the present application is shown.

[0116] like Figure 6 As shown, the terminal device or server includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 602 or the program loaded from the storage part 608 into the random access memory (RAM) 603. Various programs and data required for the operation of the terminal device or server are also stored in the RAM 603. The CPU 601, ROM 602 and RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0117] The following components are connected to the I / O interface 605: an input section 606 including a keyboard, a mouse, and the like; an output section 607 including devices such as a cathode ray tube (CRT), a liquid crystal display (LCD), and a speaker; a storage section 608 including a hard disk; and a communication section 609 including a network interface card such as a LAN card or a modem. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I / O interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 610 as needed, so that computer programs read therefrom can be installed into the storage section 608 as needed.

[0118] In particular, according to an embodiment of the present application, the above method flow steps can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a machine-readable medium, and the computer program includes program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 609, and / or installed from the removable medium 611. When the computer program is executed by the central processing unit (CPU) 601, the above-mentioned functions defined in the system of the present application are executed.

[0119] It should be noted that the computer-readable medium shown in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination of the above. More specific examples of computer-readable storage media can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this application, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, device, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. This propagated data signal can take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device. Program code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wireline, optical fiber cable, RF, or any suitable combination thereof.

[0120] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the aforementioned module, program segment, or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0121] The units or modules involved in the embodiments described in this application may be implemented by software or hardware. The units or modules described may also be provided in a processor. The names of these units or modules do not, in certain circumstances, constitute limitations on the units or modules themselves.

[0122] As another aspect, the present application further provides a computer-readable storage medium, which may be included in the electronic device described in the above embodiments, or may exist independently and not be incorporated into the electronic device. The computer-readable storage medium stores one or more programs, which, when used by one or more processors, execute the method described in the present application.

[0123] The above description is merely a preferred embodiment of the present application and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of application involved in this application is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the aforementioned application concept. For example, a technical solution formed by replacing the above-mentioned features with (but not limited to) technical features with similar functions applied for in this application.

Claims

1. A short-time startup satellite positioning prediction method for a smart helmet, characterized in that: include: Obtain positioning sequence data; Mapping the positioning sequence data to the constructed environmental feature matrix to obtain local environmental features, including: Mapping the longitude and latitude of each time step in the positioning sequence data to the corresponding grid of the environmental feature matrix; reading the values of the upper, lower, left, and right adjacent grids of the grid as the four environmental feature vectors of the time step to obtain the local environmental features; The environmental feature matrix includes: With the collection point as the center, an environmental feature matrix is constructed based on the construction project drawings; The value of each grid in the environmental feature matrix is the normalized value of the building height at that grid location: ； in, is the building height; Inputting the positioning sequence data and local environmental features into the trained prediction model to obtain the true position longitude and latitude after representing the difference and normalization; Performing inverse normalization and inverse difference processing on the longitude and latitude of the actual position to obtain predicted values of longitude and latitude, thereby completing the prediction of the current position; Before inputting the positioning sequence data into the trained prediction model, the method further includes: Cleaning the positioning sequence data to remove abnormal values in the positioning sequence data to obtain first positioning sequence data; Correcting the longitude and latitude in the first positioning sequence data using the acquired environmental deviation value to obtain second positioning sequence data; performing differential processing on the second positioning sequence data to obtain third positioning sequence data; The third positioning sequence data is standardized, and the numerical range of the third positioning sequence data is compressed to a preset range to obtain target positioning sequence data input into the prediction model.

2. The method according to claim 1, characterized in that The step of correcting the longitude and latitude in the first positioning sequence data using the acquired environmental deviation value to obtain the second positioning sequence data includes: Obtain RTK equipment and reference equipment data; Subtracting the RTK device data from the reference device data to obtain a plurality of latitude and longitude difference values; Calculating an average of the plurality of longitude and latitude differences to obtain an environmental deviation; Based on the environmental deviation, the first positioning sequence data is corrected to obtain second positioning sequence data.

3. The method according to claim 2, characterized in that The prediction model can be trained as follows: Generate a training sample set, wherein the training sample includes positioning sequence data and local environment features with annotation information; the annotation information is a positioning prediction value; The prediction model is trained using samples in the training sample set, with sample files as input and positioning prediction values as output. When the unification rate between the output positioning prediction value and the labeled positioning prediction value meets a preset threshold, the training of the prediction model is completed.

4. The method according to claim 3, characterized in that The loss function of the prediction model includes: Calculate the mean absolute error of longitude and latitude separately; The longitude mean absolute error and latitude mean absolute error are summed to obtain the loss function of the prediction model.

5. The method according to claim 4, characterized in that The mean absolute error of longitude and / or latitude can be calculated as follows: ； in, is the predicted value of longitude and latitude; is the true value of longitude and latitude.

6. A short-time startup satellite positioning prediction device for a smart helmet, characterized in that: include: An acquisition module, used for acquiring positioning sequence data; A processing module is used to map the positioning sequence data to the constructed environmental feature matrix to obtain local environmental features, including: Mapping the longitude and latitude of each time step in the positioning sequence data to the corresponding grid of the environmental feature matrix; reading the values of the upper, lower, left, and right adjacent grids of the grid as the four environmental feature vectors of the time step to obtain the local environmental features; The environmental feature matrix includes: With the collection point as the center, an environmental feature matrix is constructed based on the construction project drawings; The value of each grid in the environmental feature matrix is the normalized value of the building height at that grid location: ； in, is the building height; A positioning module, configured to input the positioning sequence data and local environmental features into a trained prediction model to obtain the true position longitude and latitude after representing the difference and normalization; Performing inverse normalization and inverse difference processing on the longitude and latitude of the actual position to obtain predicted values of longitude and latitude, thereby completing the prediction of the current position; Before inputting the positioning sequence data into the trained prediction model, the method further includes: Cleaning the positioning sequence data to remove abnormal values in the positioning sequence data to obtain first positioning sequence data; Correcting the longitude and latitude in the first positioning sequence data using the acquired environmental deviation value to obtain second positioning sequence data; performing differential processing on the second positioning sequence data to obtain third positioning sequence data; The third positioning sequence data is standardized, and the numerical range of the third positioning sequence data is compressed to a preset range to obtain target positioning sequence data input into the prediction model.

7. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images