An indoor positioning method, device and equipment

Abstract

The present invention relates to the field of indoor positioning technology, and discloses an indoor positioning method, device and equipment, the method comprising: collecting target tag information in a target environment, wherein an antenna array is arranged in the target environment, and the antenna array is used to collect the signal strength and signal phase of the target tag; inputting the target tag information into a pre-trained first dual-stream fusion model to obtain the location information of the target tag, wherein the first dual-stream fusion model comprises a first convolutional neural network and a second convolutional neural network, the first convolutional neural network is used to extract the signal strength feature of the target tag, and the second convolutional neural network is used to extract the signal phase feature of the target tag. The present invention can improve the robustness and accuracy of indoor positioning.

Description

Technical Field

[0001] The present invention relates to the field of indoor positioning technology, and in particular to an indoor positioning method, device and equipment. Background Art

[0002] With the development of science and technology, indoor positioning has a wide range of application scenarios. RFID (Radio Frequency Identification) is a non-contact automatic data collection technology with the advantages of small size, fast speed, low power consumption, strong anti-interference ability and long service life. It has great potential in the field of indoor equipment positioning for large-scale deployment of the Internet of Things.

[0003] In the related art, there are two indoor positioning methods: contact method and non-contact method. The contact method has high positioning accuracy and strong reliability, but requires the positioning target to carry a tag and is highly dependent on the location of the tag itself. The non-contact method has low dependence on the location of the tag itself and does not require the positioning target to carry a tag, but has poor robustness and is prone to positioning deviations.

[0004] In view of this, an indoor positioning method with good robustness and high positioning accuracy is needed. Summary of the invention

[0005] In view of this, the present invention provides an indoor positioning method to improve the robustness and accuracy of indoor positioning.

[0006] In a first aspect, the present invention provides an indoor positioning method, the method comprising: collecting target tag information in a target environment, wherein an antenna array is provided in the target environment, and the antenna array is used to collect the signal strength and signal phase of the target tag; inputting the target tag information into a pre-trained first dual-stream fusion model to obtain the location information of the target tag, wherein the first dual-stream fusion model comprises a first convolutional neural network and a second convolutional neural network, the first convolutional neural network is used to extract the signal strength characteristics of the target tag, and the second convolutional neural network is used to extract the signal phase characteristics of the target tag.

[0007] In this embodiment, the target tag's signal strength and signal phase are obtained, and the target tag information is input into the pre-trained first dual-stream fusion model to obtain the target tag's location information. Positioning based on the two signal characteristics of the target tag can improve the flexibility and positioning accuracy of the method.

[0008] In an optional embodiment, the target tag information is input into a pre-trained first dual-stream fusion model to obtain the location information of the target tag, including: inputting the signal strength in the target tag information into a first convolutional neural network to obtain a signal strength feature of the target tag; inputting the signal phase in the target tag information into a second convolutional neural network to obtain a signal phase feature of the target tag; fusing the signal strength feature and the signal phase feature, and obtaining the location information of the target tag based on the fusion result.

[0009] In this embodiment, the first convolutional neural network and the second convolutional neural network respectively extract the signal strength characteristics and signal phase characteristics of the target tag information, which can make full use of the target tag information, avoid the limitations of a single feature, and improve the accuracy of positioning.

[0010] In an optional embodiment, the signal strength feature and the signal phase feature are fused, and based on the fusion result, the location information of the target tag is obtained, including: fusing the signal strength feature and the signal phase feature to obtain a fused feature vector; and passing the fused feature vector through the first fully connected layer and the activation function in sequence to obtain the location information of the target tag.

[0011] In this embodiment, by fusing the two features and inputting the fusion result into the first fully connected layer, the signal features can be better learned and the positioning accuracy can be improved.

[0012] In an optional embodiment, the method also includes training a first dual-stream fusion model: constructing a first dual-stream fusion network, the first dual-stream fusion network including a first convolutional neural network and a second convolutional neural network; obtaining a training set, wherein the training set includes label information of a number of preset labels in the target environment and the actual position of the preset labels, wherein the label information includes signal strength information and signal phase information; based on the training set, obtaining first input information and second input information of the first preset label, wherein the first input information is the signal strength information of the first preset label, and the second input information is the signal phase information of the first preset label; inputting the first input information and the second input information into the first convolutional neural network and the second convolutional neural network respectively to extract a first position feature and a second position feature, wherein the first position feature corresponds to the signal strength, and the second position feature corresponds to the signal phase; performing feature fusion on the first position feature and the second position feature to obtain a first prediction result; comparing the first prediction result with the actual position of the first preset label, and based on the comparison result, adjusting and updating the weight parameters of the first convolutional neural network and / or the second convolutional neural network; determining the first dual-stream fusion model based on the updated weight parameters.

[0013] In this embodiment, by obtaining a training set and inputting the first input information and the second input information in the training set into the first convolutional neural network and the second convolutional neural network respectively, the data features in the training set can be learned. By comparing the first prediction result and the actual position, adjusting and updating the weight parameters, the accuracy and robustness of the first dual-stream fusion model can be improved.

[0014] In an optional embodiment, obtaining a training set includes: setting a label array in a target environment, and collecting a label information set of the label array, wherein the label information set includes label information of at least one label; performing data preprocessing on the label information set, wherein the data preprocessing includes at least one of missing value filling, data cleaning, and data normalization; and determining a training set based on the label information set after data preprocessing.

[0015] In this embodiment, by setting a label array in the target environment and collecting a label information set of the label array, the training effect of the subsequent model can be improved.

[0016] In an optional embodiment, when data preprocessing includes missing value filling, data preprocessing of the label information set includes: obtaining the missing label information, and determining the first missing label corresponding to the missing label information and the relative position information of the first missing label, wherein the relative position information is the position of the first missing label in the label array; based on the relative position information of the first missing label, obtaining the label information of the second label, wherein the second label is an adjacent label of the first missing label; based on the label information of the second label, determining the analog label information of the first missing label; and updating the label information set containing the analog label information.

[0017] In this embodiment, by filling missing values ​​in the data in the label information set, the integrity of the data can be improved, thereby improving the effect of subsequent model training.

[0018] In an optional embodiment, when data preprocessing includes data cleaning, data preprocessing of the tag information set includes: based on the tag information set, obtaining signal strength information of each tag in the tag array; establishing a sliding window on the tag array, and determining the median intensity within the sliding window, wherein the median intensity is the median of the signal strengths of the tags in the sliding window; obtaining the absolute deviation between the signal strength of each tag in the sliding window and the median intensity, and constructing an absolute deviation set; determining the median deviation and a first threshold, wherein the median deviation is the median of the absolute deviations in the absolute deviation set; comparing the absolute deviations of each tag in the tag array with the first threshold, and updating the signal strength information in the tag information set based on the comparison result.

[0019] In this embodiment, by establishing a sliding window and determining the median strength within the sliding window to obtain the absolute deviation, the local changes and characteristics of the signal strength data can be better preserved, the data quality of the signal strength information can be improved, and the accuracy of positioning can be improved.

[0020] In an optional embodiment, when data preprocessing includes data cleaning, data preprocessing of the tag information set also includes: based on the tag information set, obtaining a phase difference set between a label in the tag array and a third label, wherein the third label is an adjacent label of the label; determining a median phase difference and a second threshold, wherein the median phase difference is the median in the phase difference set; comparing the phase difference of each label in the tag array with the second threshold, and based on the comparison result, updating the signal phase information in the tag information set.

[0021] In this implementation, by performing data cleaning on the signal phase information in the tag information set, the data quality of the signal phase information can be improved and the accuracy of positioning can be improved.

[0022] In a second aspect, the present invention provides an indoor positioning device, the device comprising: a collection module, used to collect target tag information in a target environment, a reader array and an antenna array are arranged in the target environment, the reader array is used to collect the signal strength of the target tag, and the antenna array is used to collect the signal phase of the target tag;

[0023] The positioning module is used to input the target label information into the pre-trained first dual-stream fusion model to obtain the location information of the target label, wherein the first dual-stream fusion model includes a first convolutional neural network and a second convolutional neural network, the first convolutional neural network is used to extract the signal strength characteristics of the target label, and the second convolutional neural network is used to extract the signal phase characteristics of the target label.

[0024] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor are communicatively connected to each other, computer instructions are stored in the memory, and the processor executes the indoor positioning method of the first aspect or any corresponding embodiment thereof by executing the computer instructions. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0026] Figure 1is a schematic diagram of a flow chart of an indoor positioning method according to an embodiment of the present invention;

[0027] Figure 2 is a schematic flow chart of another indoor positioning method according to an embodiment of the present invention;

[0028] Figure 3 is a schematic flow chart of another indoor positioning method according to an embodiment of the present invention;

[0029] Figure 4 is a structural block diagram of an indoor positioning device according to an embodiment of the present invention;

[0030] Figure 5 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0031] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0032] The non-contact method does not require the target to carry an RFID tag. It can achieve non-contact positioning of the target by analyzing how the model emitted by the reference tag is affected by the target position.

[0033] This method mainly relies on RF (Radio Frequency) fingerprint technology, which collects characteristic information in the environment to build a location fingerprint database. Then, the fingerprint database and algorithm model are combined to calculate the target location, effectively reducing the interference caused by multipath effects in indoor environments.

[0034] Although the fingerprint-based contactless method shows good performance in target positioning in experimental scenarios, its positioning effect will drop significantly for new targets that are not involved in model training and are not in the fingerprint library.

[0035] Therefore, the embodiments of the present invention provide a novel indoor target positioning method that can overcome the defects of related technologies and improve the positioning accuracy and robustness of new targets to meet a wider range of indoor positioning requirements.

[0036] According to an embodiment of the present invention, an indoor positioning method embodiment is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0037] In this embodiment, an indoor positioning method is provided. Figure 1 is a flow chart of an indoor positioning method according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0038] Step S101 : collecting target tag information in a target environment. An antenna array is provided in the target environment. The antenna array is used to collect signal strength and signal phase of the target tag.

[0039] The target tag information includes the signal strength and signal phase of the target tag. By acquiring the target tag information, the RFID system can realize non-contact automatic identification of the target tag through radio waves.

[0040] The antenna array is not limited in its location. When it is set at different locations, the target tag information collected is different. It can be set at the boundary of the target environment so that the signal strength and signal phase at each location in the target environment are different, so as to improve the accuracy of the target tag positioning.

[0041] The antennas in the antenna array can activate the target tag and read the signal strength data and signal phase data of the target tag after detecting the target tag.

[0042] In an actual application, the signal strength data of the target tag may be collected continuously for multiple times, and an average value of the collected signal strength data may be obtained as a basis for subsequent positioning.

[0043] In another practical application, the signal phase data of the target tag may be collected continuously for multiple times, and the phase difference of the collected signal phase data may be obtained as a basis for subsequent positioning.

[0044] The type and quantity of antennas are not limited. In a practical application, four Laird S9028 square antennas are set up to form a 2*2 antenna array to receive target tag information.

[0045] Step S102: input the target tag information into a pre-trained first dual-stream fusion model to obtain the location information of the target tag, wherein the first dual-stream fusion model includes a first convolutional neural network and a second convolutional neural network, the first convolutional neural network is used to extract the signal strength characteristics of the target tag, and the second convolutional neural network is used to extract the signal phase characteristics of the target tag.

[0046] The structures of the first convolutional neural network and the second convolutional neural network are not restricted. The first convolutional neural network is used to extract the signal strength characteristics of the target tag. It is possible to process the signal strength in the target tag information by setting a plurality of convolutional neural networks to extract the first positioning information.

[0047] The second convolutional neural network is used to extract the signal phase characteristics of the target tag. You can choose to set up several convolutional neural networks to process the signal phase in the target tag information and extract the second positioning information.

[0048] After obtaining the output results of the first convolutional neural network and the second convolutional neural network, the outputs of the two convolutional neural networks need to be fused to integrate information of different signal features. The choice of fusion strategy is not restricted, and can be achieved through weighted averaging, concatenation, or the introduction of an attention mechanism. Based on the fusion results, the positioning information of the target label is obtained.

[0049] The indoor positioning method provided in this embodiment obtains the signal strength and signal phase of the target tag and inputs the target tag information into the pre-trained first dual-stream fusion model to obtain the location information of the target tag. Positioning based on the two signal characteristics of the target tag can improve the flexibility and positioning accuracy of the method.

[0050] In this embodiment, an indoor positioning method is provided. Figure 2 is a flow chart of an indoor positioning method according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0051] Step S201, collecting target tag information in a target environment, wherein an antenna array is provided in the target environment, and the antenna array is used to collect signal strength and signal phase of the target tag.

[0052] For details, please see Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0053] Step S202: input the target tag information into a pre-trained first dual-stream fusion model to obtain the location information of the target tag, wherein the first dual-stream fusion model includes a first convolutional neural network and a second convolutional neural network, the first convolutional neural network is used to extract the signal strength characteristics of the target tag, and the second convolutional neural network is used to extract the signal phase characteristics of the target tag.

[0054] Specifically, the above step S202 includes:

[0055] Step S2021, input the signal strength in the target tag information into the first convolutional neural network to obtain the signal strength feature of the target tag.

[0056] Among them, the architecture selection of the first convolutional neural network is not restricted, and common architectures such as VGG-16 and ResNet can be used to extract features from the signal strength in the target label information and identify the relationship between the signal strength and the location information.

[0057] Step S2022: input the signal phase in the target tag information into the second convolutional neural network to obtain the signal phase feature of the target tag.

[0058] Among them, the architecture selection of the second convolutional neural network is not restricted, and the characteristics of the signal phase can be extracted through multiple convolutional layers.

[0059] In a practical application, both the first convolutional neural network and the second convolutional neural network adopt a one-dimensional convolutional neural network structure, and extract the signal strength characteristics and signal phase characteristics of the target label respectively through residual blocks.

[0060] Step S2023, fuse the signal strength feature and the signal phase feature, and obtain the location information of the target tag based on the fusion result.

[0061] Among them, the method of feature fusion is not restricted. The fused feature vector can be obtained by feature summation or feature concatenation, and then the fused feature vector is input into the fully connected layer, and the output result of the fully connected layer is used as the position of the target label.

[0062] In some optional implementations, the above step S2023 includes:

[0063] Step a1, fuse the signal strength feature and the signal phase feature to obtain a fused feature vector.

[0064] Among them, the method of feature fusion is not restricted, and the fused feature vector can be obtained by weighted averaging or direct concatenation.

[0065] Step a2: Pass the fused feature vector through the first fully connected layer and the activation function in sequence to obtain the location information of the target label.

[0066] The fused feature vector is input into the first fully connected layer, which will learn the features in the fused feature vector and integrate and transform them to obtain the location information of the final target label.

[0067] The choice of activation function is not restricted, and you can choose ReLU (Rectified Linear Unit) to increase the robustness of the model.

[0068] By fusing the two features and inputting the fusion result into the first fully connected layer, the signal features can be better learned and the positioning accuracy can be improved.

[0069] In the indoor positioning method provided in this embodiment, the first convolutional neural network and the second convolutional neural network respectively extract the signal strength characteristics and signal phase characteristics of the target tag information, which can make full use of the target tag information, avoid the limitations of a single feature, and improve the accuracy of positioning.

[0070] In this embodiment, an indoor positioning method is provided. Figure 3 is a flow chart of an indoor positioning method according to an embodiment of the present invention. Figure 3 As shown, the process includes the following steps:

[0071] Step S301, collecting target tag information in a target environment, wherein an antenna array is provided in the target environment, and the antenna array is used to collect signal strength and signal phase of the target tag.

[0072] For details, please see Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0073] Step S302: training a first model.

[0074] Specifically, the above step S302 includes:

[0075] Step S3021, construct a first dual-stream fusion network, the first dual-stream fusion network includes a first convolutional neural network and a second convolutional neural network.

[0076] Among them, the construction method of the first dual-stream fusion network is not restricted. According to the actual situation, the network structure of the first dual-stream fusion network can be determined first, where the network structure includes the number of convolutional layers, convolution kernels, etc. Then the corresponding activation function is determined to improve the nonlinear processing ability of the network. Then a one-dimensional pooling layer is used to reduce the feature dimension and extract more important features to improve the accuracy of positioning.

[0077] Step S3022: Acquire a training set, wherein the training set includes tag information of a plurality of preset tags in the target environment and actual positions of the preset tags, wherein the tag information includes signal strength information and signal phase information.

[0078] There is no restriction on the method of obtaining the training set. Several preset labels can be set in the target environment, and the label information of the preset labels can be directly collected.

[0079] In some optional implementations, the above step S3022 includes:

[0080] Step a1: setting a tag array in a target environment and collecting a tag information set of the tag array, wherein the tag information set includes tag information of at least one tag.

[0081] There are no restrictions on the arrangement of the tag array and the number of tags. The number of tags can be determined according to the size of the target environment. The larger the area of ​​the target environment, the more tags there are.

[0082] The model of the tag is also not restricted. In a practical application, using the Impinj UHF omnidirectional tag H47, signals can be received and sent in all directions, avoiding the problem of location information loss caused by general directional tags that can only receive and send signals in a fixed direction.

[0083] In a practical application, the interval between each tag is set to 30 cm to avoid coupling effects between tags.

[0084] Step a2, performing data preprocessing on the label information set, wherein the data preprocessing includes at least one of missing value filling, data cleaning and data normalization.

[0085] The data preprocessing method is not limited. The collected data can be processed by data normalization, data filtering, etc. to ensure data quality and improve the accuracy of model positioning.

[0086] In order to improve data quality, the data in the label information set can be cleaned and normalized. And when there are missing values ​​in the information in the label information set, the missing values ​​can be filled.

[0087] In a practical application, data preprocessing includes data normalization. The data normalization method is not limited, and all data can be directly scaled to a specified range. Or the data can be preprocessed by Z-score normalization or other methods.

[0088] In some optional implementations, the data preprocessing includes missing value filling, and the above step a2 includes:

[0089] Step b1, obtaining missing tag information, and determining the first missing tag corresponding to the missing tag information and the relative position information of the first missing tag, wherein the relative position information is the position of the first missing tag in the tag array.

[0090] Among them, the tag information set can be identified first to confirm whether there is data missing. If not, there is no need to fill the missing value of the tag information set. If it exists, the position of the missing tag information is first determined, such as the row and column of the array.

[0091] Step b2: based on the relative position information of the first missing tag, obtain the tag information of the second tag, wherein the second tag is an adjacent tag of the first missing tag.

[0092] According to the location of the first missing tag, the tag information of the second tag adjacent to the first missing tag is determined, wherein the number of the second tags is not limited and can be selected according to actual conditions.

[0093] Step b3: Determine the simulated label information of the first missing label based on the label information of the second label.

[0094] The simulated label information of the first missing label can be determined based on the average value of the signal strength and signal phase of several second labels. The prediction model can also be trained, and the label information of the second label is input into the prediction model to obtain the simulated label information of the first missing label.

[0095] Step b4: updating the tag information set including the simulated tag information.

[0096] By filling missing values ​​in the data in the label information set, the integrity of the data can be improved, thereby improving the effect of subsequent model training.

[0097] In some optional implementations, when data preprocessing includes data cleaning, the above step a2 includes:

[0098] Step c1, based on the tag information set, obtaining the signal strength information of each tag in the tag array.

[0099] If there are missing values ​​in the tag information set, the missing values ​​need to be processed first. The missing values ​​can be directly removed or filled according to the average value of the signal strength adjacent to the missing values.

[0100] Step c2, establishing a sliding window on the tag array, and determining the median intensity within the sliding window, wherein the median intensity is the median of the signal strengths of the tags within the sliding window.

[0101] The signal strength information on the tag array can be converted into a one-dimensional array, and a sliding window can be established on the one-dimensional array, wherein the size of the sliding window can be set according to the number of tags contained in the tag array. In an actual application, the size of the sliding window is 5, that is, the sliding window contains the signal strengths of five tags, and the median strength is the median of the signal strengths of these five tags.

[0102] The step size of the sliding window can be set according to actual conditions. In a practical application, the step size of the sliding window is 1.

[0103] Step c3, obtaining the absolute deviation between the signal strength of each tag in the sliding window and the median strength, and constructing an absolute deviation set.

[0104] The absolute deviation is the absolute value of the difference between the signal intensity of the tag and the median intensity.

[0105] Step c4, determining a median deviation and a first threshold, wherein the median deviation is the median of the absolute deviations in the absolute deviation set.

[0106] The setting of the first threshold is not limited, and the first threshold can be set according to actual conditions, or can be directly set to two or three times the median of the deviation.

[0107] Step c5: comparing the absolute deviation of each tag in the tag array with the first threshold, and updating the signal strength information in the tag information set based on the comparison result.

[0108] If the absolute deviation is less than the first threshold, the corresponding signal strength information is retained. If the absolute deviation is greater than or equal to the first threshold, the signal strength information of the corresponding tag is replaced with the intensity median.

[0109] In a practical application, the size of the sliding window is 5 and the step size of the sliding window is 1. The signal strength information of the tag array is obtained and converted into a one-dimensional array, which stores the signal strength information of each tag on the tag array. Starting from the first element of the data, slide the sliding window to the starting position of the array.

[0110] Get the median of the signal strength of the five tags in the sliding window as the median strength. Then get the absolute deviation between the signal strength of the five tags in the sliding window and the median strength to construct an absolute deviation set. Then determine the median deviation and the first threshold, and compare the relationship between the absolute deviation of the tags in the sliding window and the first threshold, and update the signal strength information in the sliding window based on the comparison result. Then move the sliding window backward by one step, that is, move the signal strength sample position of one tag to the right. Then repeat the above steps until the sliding window passes through the entire data to update the signal strength information in the tag information set.

[0111] In some optional implementations, the above step a2 further includes:

[0112] Step c6: based on the tag information set, obtaining a phase difference set between the tag in the tag array and a third tag, wherein the third tag is an adjacent tag of the tag.

[0113] The third tag is determined according to the position information of the tag in the tag array. If the adjacent tags are not unique, the tag with the largest phase difference with the adjacent tag can be selected as the third tag.

[0114] Step c7, determining the median phase difference and the second threshold, wherein the median phase difference is the median in the phase difference set.

[0115] The setting of the second threshold is not limited, and the second threshold can be set according to actual conditions. The second threshold is directly set to two or three times the median of the phase difference.

[0116] Step c8, comparing the phase difference of each tag in the tag array with the second threshold, and updating the signal phase information in the tag information set based on the comparison result.

[0117] If the phase difference in the phase difference set is less than the second threshold, the signal phase information of the tag corresponding to the phase difference is retained. If the phase difference in the phase difference set is greater than the second threshold, the signal phase information of the tag corresponding to the phase difference is adjusted. The adjustment method is not limited and can be directly removed.

[0118] By performing data cleaning on the signal strength information in the tag information set, the data quality of the signal strength information can be improved and the accuracy of positioning can be improved.

[0119] Step a3: determine the training set based on the label information set after data preprocessing.

[0120] By setting a label array in the target environment and collecting the label information set of the label array, the training effect of the subsequent model can be improved.

[0121] Step S3023: Based on the training set, first input information and second input information of the first preset label are obtained, wherein the first input information is signal strength information of the first preset label, and the second input information is signal phase information of the first preset label.

[0122] Among them, according to the label information of the first preset label in the training set, it can be divided into signal strength information and signal phase information.

[0123] Step S3024, inputting the first input information and the second input information into the first convolutional neural network and the second convolutional neural network respectively to extract the first position feature and the second position feature, wherein the first position feature corresponds to the signal strength and the second position feature corresponds to the signal phase.

[0124] The structures of the first convolutional neural network and the second convolutional neural network are not limited. The first convolutional neural network is used to learn the first input information, extract the first position feature, and identify the relationship between the signal strength and the position; the second convolutional neural network is used to learn the second input information, extract the second position feature, and identify the relationship between the signal phase and the position.

[0125] Step S3025: Fusing the first position feature and the second position feature to obtain a first prediction result.

[0126] There is no restriction on the feature fusion method, and the fused feature vector can be obtained by feature summation or feature concatenation, and then the fused feature vector is input into the fully connected layer, and the output result of the fully connected layer is used as the first prediction result.

[0127] Step S3026, comparing the first prediction result with the actual position of the first preset label, and based on the comparison result, adjusting and updating the weight parameters of the first convolutional neural network and / or the second convolutional neural network.

[0128] After obtaining the updated initial weight parameters, the test set can be used to evaluate and train the performance of the network model. Based on the test results, the initial weight parameters can also be adjusted and updated for a second time.

[0129] Step S3027: Determine a first dual-stream fusion model based on the updated weight parameters.

[0130] After the first dual-stream fusion model training is determined, the model can be verified through the test set, and the first dual-stream fusion model can be evaluated based on the model results. The acquisition of the test set is not restricted, and a part of the training set can be directly divided as the test set, and the test set does not participate in the model training. The evaluation indicators are not restricted, and the confusion matrix, accuracy, recall rate, etc. of the first dual-stream fusion model can be obtained.

[0131] Step S303: Input the target tag information into the pre-trained first dual-stream fusion model to obtain the location information of the target tag, wherein the first dual-stream fusion model includes a first convolutional neural network and a second convolutional neural network, the first convolutional neural network is used to extract the signal strength characteristics of the target tag, and the second convolutional neural network is used to extract the signal phase characteristics of the target tag.

[0132] For details, please see Figure 1 Step S102 of the illustrated embodiment will not be described in detail here.

[0133] The indoor positioning method provided in this embodiment can learn the data features in the training set by acquiring a training set and inputting the first input information and the second input information in the training set into the first convolutional neural network and the second convolutional neural network respectively. By comparing the first prediction result and the actual position, adjusting and updating the weight parameters, the accuracy and robustness of the first dual-stream fusion model can be improved.

[0134] In this embodiment, an indoor positioning device is also provided, which is used to implement the above embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0135] This embodiment provides an indoor positioning device, such as Figure 4 As shown, including:

[0136] The acquisition module 401 is used to acquire target tag information in the target environment. An antenna array is provided in the target environment. The antenna array is used to acquire the signal strength and signal phase of the target tag.

[0137] The positioning module 402 is used to input the target tag information into the pre-trained first dual-stream fusion model to obtain the location information of the target tag, wherein the first dual-stream fusion model includes a first convolutional neural network and a second convolutional neural network, the first convolutional neural network is used to extract the signal strength characteristics of the target tag, and the second convolutional neural network is used to extract the signal phase characteristics of the target tag.

[0138] In some optional implementations, the positioning module 402 includes:

[0139] The signal strength feature extraction unit is used to input the signal strength in the target tag information into the first convolutional neural network to obtain the signal strength feature of the target tag.

[0140] The signal phase feature extraction unit is used to input the signal phase in the target label information into the second convolutional neural network to obtain the signal phase feature of the target label.

[0141] The fusion unit is used to fuse the signal strength feature and the signal phase feature, and obtain the location information of the target tag based on the fusion result.

[0142] In some optional embodiments, the fusion unit comprises:

[0143] The feature vector acquisition subunit is used to fuse the signal strength feature and the signal phase feature to obtain a fused feature vector.

[0144] The position information acquisition subunit is used to pass the fused feature vector through the first fully connected layer and the activation function in sequence to obtain the position information of the target label.

[0145] In some optional implementations, the indoor positioning device further includes:

[0146] The model training module 403 is used to train the first dual-stream fusion model.

[0147] Among them, the model training module 403 includes:

[0148] A construction unit, used to construct a first dual-stream fusion network, where the first dual-stream fusion network includes a first convolutional neural network and a second convolutional neural network;

[0149] A training set acquisition unit, used to acquire a training set, wherein the training set includes label information of a plurality of preset labels in a target environment and actual positions of the preset labels, wherein the label information includes signal strength information and signal phase information;

[0150] An information acquisition unit, configured to acquire first input information and second input information of a first preset label based on a training set, wherein the first input information is signal strength information of the first preset label, and the second input information is signal phase information of the first preset label;

[0151] A feature extraction unit, used to input the first input information and the second input information into the first convolutional neural network and the second convolutional neural network respectively, so as to extract a first position feature and a second position feature, wherein the first position feature corresponds to the signal strength, and the second position feature corresponds to the signal phase;

[0152] A first prediction result acquisition unit, used for performing feature fusion on the first position feature and the second position feature to acquire a first prediction result;

[0153] An adjustment unit, used to compare the first prediction result with the actual position of the first preset label, and based on the comparison result, adjust and update the weight parameters of the first convolutional neural network and / or the second convolutional neural network;

[0154] The model determination unit is used to determine a first dual-stream fusion model based on the updated weight parameters.

[0155] In some optional implementations, the training set acquisition unit includes:

[0156] The tag information set collecting subunit is used to set a tag array in the target environment and collect a tag information set of the tag array, wherein the tag information set includes tag information of at least one tag.

[0157] The data preprocessing subunit is used to perform data preprocessing on the label information set, wherein the data preprocessing includes at least one of missing value filling, data cleaning and data normalization.

[0158] The training set determination subunit is used to determine the training set based on the label information set after data preprocessing.

[0159] In some optional implementations, the data preprocessing subunit includes:

[0160] The missing tag information acquisition component is used to acquire the missing tag information and determine the first missing tag corresponding to the missing tag information and the relative position information of the first missing tag, wherein the relative position information is the position of the first missing tag in the tag array.

[0161] The second tag information acquisition component is used to acquire tag information of a second tag based on the relative position information of the first missing tag, wherein the second tag is an adjacent tag of the first missing tag.

[0162] The simulated tag information determining component is used to determine the simulated tag information of the first missing tag based on the tag information of the second tag.

[0163] The tag information collection updating component is used to update the tag information collection containing the simulation tag information.

[0164] In some optional implementations, the data preprocessing subunit includes:

[0165] A signal strength acquisition component, used to acquire signal strength information of each tag in the tag array based on the tag information set;

[0166] A sliding window establishment component is used to establish a sliding window on the tag array and determine the median intensity within the sliding window, wherein the median intensity is the median of the signal strength of the tags within the sliding window;

[0167] An absolute deviation set building component is used to obtain the absolute deviation between the signal strength of each tag in the sliding window and the median strength, and build an absolute deviation set;

[0168] A first determination component is used to determine a median deviation and a first threshold, wherein the median deviation is the median of absolute deviations in the absolute deviation set;

[0169] The signal strength information updating component is used to compare the absolute deviation of each tag in the tag array with the first threshold value, and based on the comparison result, update the signal strength information in the tag information set.

[0170] In some optional implementations, the data preprocessing subunit includes:

[0171] A phase difference set acquisition component, used to acquire a phase difference set between a tag in the tag set and a third tag based on the tag information set, wherein the third tag is an adjacent tag of the tag;

[0172] A second determination component is used to determine a phase difference median and a second threshold, wherein the phase difference median is a median in a phase difference set;

[0173] The signal phase information updating component is used to compare the phase difference of each tag in the tag array with the second threshold value, and based on the comparison result, update the signal phase information in the tag information set.

[0174] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0175] The indoor positioning device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0176] The embodiment of the present invention also provides a computer device having the above Figure 4 The indoor positioning device shown.

[0177] See also Figure 5 , Figure 5 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 5As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 5 A processor 10 is taken as an example.

[0178] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0179] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0180] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0181] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0182] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 5 The example of connecting through bus is taken in the following.

[0183] The input device 30 can receive input digital or character information, and generate key signal input related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator bar, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display and a plasma display. In some optional embodiments, the display device may be a touch screen.

[0184] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0185] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0186] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. An indoor positioning method, characterized in that: The method comprises: Collecting target tag information in a target environment, wherein an antenna array is arranged in the target environment, and the antenna array is used to collect signal strength and signal phase of the target tag; the antenna array is arranged at a boundary of the target environment; Inputting the target tag information into a pre-trained first dual-stream fusion model to obtain the location information of the target tag, wherein the first dual-stream fusion model includes a first convolutional neural network and a second convolutional neural network, the first convolutional neural network is used to extract the signal strength feature of the target tag, and the second convolutional neural network is used to extract the signal phase feature of the target tag; The step of inputting the target label information into a pre-trained first dual-stream fusion model to obtain the position information of the target label includes: Inputting the signal strength in the target tag information into the first convolutional neural network to obtain the signal strength feature of the target tag; Inputting the signal phase in the target tag information into the second convolutional neural network to obtain the signal phase feature of the target tag; The signal strength feature and the signal phase feature are fused, and based on the fusion result, the location information of the target tag is obtained.

2. The method according to claim 1, characterized in that The fusing the signal strength feature and the signal phase feature, and acquiring the location information of the target tag based on the fusion result includes: Fusing the signal strength feature and the signal phase feature to obtain a fused feature vector; The fused feature vector is sequentially passed through the first fully connected layer and the activation function to obtain the location information of the target label.

3. The method according to claim 1, characterized in that The method also includes training a first two-stream fusion model: Constructing a first dual-stream fusion network, wherein the first dual-stream fusion network includes a first convolutional neural network and a second convolutional neural network; Acquire a training set, wherein the training set includes label information of a plurality of preset labels in a target environment and actual positions of the preset labels, wherein the label information includes signal strength information and signal phase information; Based on the training set, obtaining first input information and second input information of a first preset label, wherein the first input information is signal strength information of the first preset label, and the second input information is signal phase information of the first preset label; Inputting the first input information and the second input information into a first convolutional neural network and a second convolutional neural network respectively to extract a first position feature and a second position feature, wherein the first position feature corresponds to the signal strength, and the second position feature corresponds to the signal phase; Performing feature fusion on the first position feature and the second position feature to obtain a first prediction result; Comparing the first prediction result with the actual position of the first preset label, and adjusting and updating the weight parameters of the first convolutional neural network and / or the second convolutional neural network based on the comparison result; Based on the updated weight parameters, a first dual-stream fusion model is determined.

4. The method according to claim 3, characterized in that The obtaining of the training set comprises: Setting a tag array in a target environment, and collecting a tag information set of the tag array, wherein the tag information set includes tag information of at least one tag; Performing data preprocessing on the label information set, wherein the data preprocessing includes at least one of missing value filling, data cleaning and data normalization; Based on the label information set after data preprocessing, a training set is determined.

5. The method according to claim 4, characterized in that In the case where the data preprocessing includes missing value filling, the data preprocessing of the label information set includes: Acquire missing tag information, and determine the first missing tag corresponding to the missing tag information and the relative position information of the first missing tag, wherein the relative position information is the position of the first missing tag in the tag array; Based on the relative position information of the first missing tag, acquiring tag information of a second tag, wherein the second tag is an adjacent tag of the first missing tag; Determining the simulated label information of the first missing label based on the label information of the second label; The tag information set including the simulation tag information is updated.

6. The method according to claim 4, characterized in that In the case where the data preprocessing includes data cleaning, the data preprocessing of the tag information set includes: Based on the tag information set, acquiring signal strength information of each tag in the tag array; Establishing a sliding window on the tag array, and determining the median intensity within the sliding window, wherein the median intensity is the median of the signal intensities of the tags within the sliding window; Obtaining the absolute deviation between the signal strength of each tag in the sliding window and the median of the strength, and constructing an absolute deviation set; Determine a median deviation and a first threshold, wherein the median deviation is a median of absolute deviations in a set of absolute deviations; The absolute deviation of each tag in the tag array is compared with the first threshold, and based on the comparison result, the signal strength information in the tag information set is updated.

7. The method according to claim 4, characterized in that In the case where the data preprocessing includes data cleaning, the data preprocessing of the tag information set further includes: Based on the tag information set, acquiring a phase difference set between a tag in the tag array and a third tag, wherein the third tag is an adjacent tag of the tag; Determine a phase difference median and a second threshold, wherein the phase difference median is the median in the phase difference set; The phase difference of each tag in the tag array is compared with the second threshold, and based on the comparison result, the signal phase information in the tag information set is updated.

8. An indoor positioning device, characterized in that: The device comprises: A collection module, used to collect target tag information in a target environment, wherein an antenna array is provided in the target environment, and the antenna array is used to collect the signal strength and signal phase of the target tag; the antenna array is provided at the boundary of the target environment; A positioning module, used to input the target tag information into a pre-trained first dual-stream fusion model to obtain the location information of the target tag, wherein the first dual-stream fusion model includes a first convolutional neural network and a second convolutional neural network, the first convolutional neural network is used to extract the signal strength feature of the target tag, and the second convolutional neural network is used to extract the signal phase feature of the target tag; Among them, the positioning module includes: A signal strength feature extraction unit, used to input the signal strength in the target tag information into the first convolutional neural network to obtain the signal strength feature of the target tag; A signal phase feature extraction unit is used to input the signal phase in the target label information into the second convolutional neural network to obtain the signal phase feature of the target label; The fusion unit is used to fuse the signal strength feature and the signal phase feature, and obtain the location information of the target tag based on the fusion result.

9. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the indoor positioning method according to any one of claims 1 to 7 by executing the computer instructions.

Images