阅读说明：本技术 一种目标不携带设备的室内定位方法 (Indoor positioning method for target without carrying equipment ) 是由 丁数学 张康康 谭本英 李玉洁 李广伟 于 2021-06-22 设计创作，主要内容包括：本发明公开了一种目标不携带设备的室内精确定位方法,该方法涉及获取目标在室内不同位置对无线电波的遮挡作用的量化数据与目标所在位置的标签,并对数据集进行划分得到训练样本集与测试样本集；然后将所有位置上的量化数据进行处理,最后将测试样本集使用字典进行线性表示,通过在目标方程中加入正则化来限制解的范围,求解可以得到一个包含目标位置信息的稀疏向量,稀疏向量中极大值的位置进行加权联合计算得到目标位置。本发明能够对目标室内位置进行有效预测,对智能家居控制,目标监控,银行防盗和智能医疗等提供技术基础。(The invention discloses an indoor accurate positioning method of a target without carrying equipment, which relates to the steps of obtaining quantitative data of shielding effect of the target on radio waves at different indoor positions and a label of the position of the target, and dividing a data set to obtain a training sample set and a test sample set; and then processing the quantized data at all positions, finally linearly representing the test sample set by using a dictionary, limiting the range of a solution by adding regularization into an object equation, solving to obtain a sparse vector containing object position information, and performing weighted joint calculation on the position of a maximum value in the sparse vector to obtain the object position. The invention can effectively predict the indoor position of the target and provide a technical basis for intelligent home control, target monitoring, bank anti-theft, intelligent medical treatment and the like.)

1. An indoor accurate positioning method without carrying equipment on a target is characterized in that: the method comprises the following steps:

(1) acquiring quantitative data of shielding effects of the target on radio waves at different indoor positions and label data of the target position, and dividing the data to obtain a training sample set and a test sample set; wherein the training sample set comprises received radio data of the target at different positions and a position label of the target;

(2) processing the quantized data on all positions, processing the data on each position into a column vector, and forming a dictionary by all the obtained column vectors in a column splicing mode;

(3) constructing a target equation, simultaneously considering a linear fitting error and a regularization term of a test signal, and performing iterative solution to obtain a final sparse vector;

(4) and performing joint analysis calculation on a plurality of maximum values in the sparse vector to obtain a target prediction position.

2. The method of claim 1, wherein: the method for acquiring the label data and dividing the label data in the step (1) comprises the following steps:

1) dividing an indoor positioning scene into square grids with the same size, wherein each grid represents a position where a target possibly exists and corresponds to a unique digital label;

2) the method comprises the following steps that L radio receiving sensors which are uniformly distributed are arranged around a monitoring area, the positions of the L radio receiving sensors are fixed, a radio transmitting sensor is configured, the radio transmitting sensor is provided with K variable transmitting positions, wireless communication between one transmitting sensor and one receiving sensor is called a group of links, and N-K-L groups of links are shared in the whole experimental scene;

3) the target is present at a certain position in the room, the radio transmitting sensor transmits a radio signal, and all receiving sensors receive the signal; the transmitting sensor moves all the selectable positions and transmits radio signals, and all the receiving nodes receive arriving signals at the same time; then, extracting signal strength from the frequency spectrum of each link as effective information, and obtaining an information matrix at each position of a target, wherein the dimension size is L rows and K columns; meanwhile, the digital label information of the target at each grid position is stored and can be used for corresponding to a real two-dimensional position coordinate;

4) for the collected data set, the ratio of 5: the scale of 1 is divided into a training sample set and a test sample set.

3. The method of claim 1, wherein: the column splicing method of the column vector in the step (2) comprises the following steps:

1) firstly, converting an information matrix obtained by a target at a first position into a column vector, wherein the dimension size is N, and the label information is 1;

2) repeating the measurement tau times at the first position, respectively converting into column vectors to obtain tau column vectors in total, wherein the label information is Lab₁＝{1,1,…,1}；

3) Repeating the steps 1) and 2) until the data of the target on all the optional positions are collected; and respectively carrying out column-column splicing on the column vectors at each position in sequence to form a final dictionary, carrying out two-norm normalization on each column of the dictionary, and simultaneously carrying out column-splicing on corresponding label information to form labeling information.

4. The method of claim 3, wherein: in the step 3), the specific way of normalizing the two norms of the column vector x is as follows:

wherein

5. The method of claim 1, wherein the step (3) comprises the steps of:

1) firstly, an objective equation is constructed, wherein the objective equation should include two parts, namely fitting error and regularization term, and the objective equation is as follows:

where y represents a test signal, W is a dictionary matrix, x is the sparse vector sought,represents the introduced non-convex GMC regularization; both α and λ are positive parameters; z is an auxiliary vector;

2) since the number of targets is always far less than the total number of the network grid points, a sparse coding method is adopted to solve the sparse vector.

6. The method of claim 5, wherein: solving of the sparse codes in the step 2) adopts a forward and backward splitting algorithm.

7. The method according to claim 1, wherein the step (4) comprises the steps of:

1) firstly, a sparse vector x is obtained as x ═ x_1,1,x_1,2,…,x_1,τ,x_2,1,...x_S,τAfter that, x is equal to Rⁿ(ii) a By summing the x coefficients of each sparse vectorWherein 1. ltoreq. p. ltoreq.S, thus obtaining x^*＝{x₁,...,x_p,...x_S}，x^*In order to obtain the weight vectors corresponding to different predicted labels, the larger the weight is, the greater the probability of predicting the current position is, and x is calculated^*The elements in (a) are arranged in descending order;

2) extracting a plurality of maximum values in the weight vector, wherein each weight corresponds to a different position label, and the position labels can correspond to the two-dimensional coordinates one by one;

3) and normalizing the weights, and performing combined calculation to obtain the predicted target position.

8. The method of claim 7, wherein: the normalization processing mode and the target position predicting method in the step 3) are as follows:

normalization treatment: first, several maximum value vectors are arranged according toNormalization is carried out, wherein num is the number of the maximum values, and normalized weights can be obtained;

predicting the target position: calculated by the following formula:where loc denotes the position of the final target, a_iIndicating the normalized weight corresponding to location i, and loc (i) indicating the true two-dimensional coordinate corresponding to location i.

Technical Field

The invention belongs to the field of indoor positioning, and particularly relates to an indoor accurate positioning method for a target without carrying equipment.

Background

Localization has always played an indispensable role in the construction of the intelligent world. According to different positioning principles, positioning is divided into two types from a large direction, one is positioning of a target carrying device, and the other is positioning of a non-carrying device. The positioning of the target carrying equipment means that the detected target carries detectable electronic equipment and actively participates in positioning activities. The positioning of target carrying devices has become well established in some fields. For example, GPS-based positioning technology has been used successfully for vehicle and personal navigation; the positioning technology based on the RFID is also developed comprehensively in the field of indoor vehicle positioning and the like. However, the positioning of the target carrying device is not suitable in some application scenarios, for example, in the anti-theft detection of banks, it cannot be expected that an intruder carries an electronic device for detection. In the life of intelligence house, through the location personage position, can control intelligent desk lamp or other intelligent electrical apparatus, like my can not require the guest to carry unified "identity card" that can supply discernment. Therefore, the concept of device-less positioning has emerged and is of interest for research.

The location of the portable device is that the tracked entity neither needs to carry the device nor actively participate in the location algorithm. There are many mature technologies that are used in the field of location without equipment, including computer vision, radio frequency, radar, etc. The convolutional neural network greatly accelerates the development speed of computer vision, and the functions of accurately positioning and tracking the target object in the video in real time are realized at present. However, computer vision has obvious defects, such as in dark nights or foggy days, the positioning performance is greatly reduced and target positioning cannot be performed on the partition wall. Radar-based positioning may overcome the deficiencies of the above techniques. However, the use of radar for accurate positioning requires the arrangement of radar arrays, which results in higher installation costs and makes it impossible to install the radar positioning system in a daily life environment. The widespread WIFI technology makes the wireless infrastructure ubiquitous, and positioning based on radio technology does not require modification of the current wireless infrastructure. Radio is ubiquitous in life, and it is not passing through our body at all times. The article thus explores positioning in radio networks in hopes that our method has wider application.

Disclosure of Invention

The invention aims to provide an indoor accurate positioning method without carrying equipment on a target, which obtains a sparse vector by solving a target equation containing a non-convex sparse regularization term; a method of joint analytical calculation is used to predict the position coordinates of the target.

The invention adopts the following technical scheme:

an indoor accurate positioning method of a target without carrying equipment comprises the following steps:

(1) acquiring quantitative data of shielding effects of the target on radio waves at different indoor positions and label data of the target position, and dividing the data to obtain a training sample set and a test sample set; the training sample set comprises receiving radio data of the target at different positions and position labels of the target;

(2) processing the quantized data on all positions, processing the data on each position into a column vector, and forming a dictionary by all the obtained column vectors in a column splicing mode;

(3) constructing a target equation, simultaneously considering a linear fitting error and a regularization term of a test signal, and performing iterative solution to obtain a final sparse vector;

(4) and performing joint analysis calculation on a plurality of maximum values in the sparse vector to obtain a target prediction position.

The method for acquiring the label data and dividing the label data in the step (1) comprises the following steps:

1) dividing an indoor positioning scene into square grids with the same size, wherein each grid represents a position where a target possibly exists and corresponds to a unique digital label;

2) the method comprises the following steps that L radio receiving sensors which are uniformly distributed are arranged around a monitoring area, the positions of the L radio receiving sensors are fixed, a radio transmitting sensor is configured, the radio transmitting sensor is provided with K variable transmitting positions, wireless communication between one transmitting sensor and one receiving sensor is called a group of links, and N-K-L groups of links are shared in the whole experimental scene;

3) the target is present at a certain position in the room, the radio transmitting sensor transmits a radio signal, and all receiving sensors receive the signal; the transmitting sensor moves all the optional positions and transmits radio signals, and all the receiving nodes receive arriving signals at the same time; then, extracting signal intensity from the frequency spectrum of each link as effective information, and obtaining an information matrix at each position of a target, wherein the dimension size is L rows and K columns; meanwhile, the digital label information of the target at each grid position is stored and can be used for corresponding to a real two-dimensional position coordinate;

4) for the collected data set, the ratio of 5: the scale of 1 is divided into a training sample set and a test sample set.

The column vector splicing method in the step (2) comprises the following steps:

1) firstly, converting an information matrix obtained by a target at a first position into a column vector, wherein the dimension is N, and the label information is 1;

2) repeating the measurement tau times at the first position, respectively converting into column vectors to obtain tau column vectors in total, wherein the label information is Lab₁＝{1,1,…,1}；

3) Repeating the steps 1) and 2) until the data of the target on all the optional positions are collected; and respectively carrying out column-column splicing on the column vectors at each position in sequence to form a final dictionary, carrying out two-norm normalization on each column of the dictionary, and simultaneously carrying out column-splicing on corresponding label information to form labeling information.

The method for constructing the target equation and iteratively solving in the step (3) comprises the following steps:

1) firstly, an objective equation is constructed, wherein the objective equation should include two parts, namely fitting error and regularization term, and the objective equation is as follows:

where y represents a test signal, W is a dictionary matrix, x is the sparse vector sought,represents the introduced non-convex GMC regularization; both α and λ are positive parameters; z is an auxiliary vector;

2) because the number of the targets is always far less than the total number of the network grid points, the sparse vector is solved by adopting a sparse coding method, and the sparse coding can be solved by adopting a forward and backward splitting algorithm.

The method for performing joint analysis calculation on several maximum values in the sparse vector in step (4) comprises the following steps:

1) firstly, a sparse vector x is obtained as x ═ x_1,1,x_1,2,…,x_1,τ,x_2,1,...x_S,τAfter that, x is equal to Rⁿ(ii) a By summing the x coefficients of each sparse vectorWherein 1. ltoreq. p. ltoreq.S, thus obtaining x^*＝{x₁,...,x_p,...x_S}，x^*In order to obtain the weight vectors corresponding to different predicted labels, the larger the weight is, the greater the probability of predicting the current position is, and x is calculated^*The elements in (a) are arranged in descending order;

2) extracting a plurality of maximum values in the weight vector, wherein each weight corresponds to a different position label, and the position labels can correspond to the two-dimensional coordinates one by one;

3) and normalizing the weights, and performing combined calculation to obtain the predicted target position.

Further, the normalization processing method and the target position predicting method in step 3) are as follows:

normalization treatment: first, several maximum value vectors are arranged according toNormalization is carried out, wherein num is the number of the maximum values, and normalized weights can be obtained;

predicting the target position: calculated by the following formula:where loc represents the position of the final target, a_iIndicating the normalized weight corresponding to location i, and loc (i) indicating the true two-dimensional coordinate corresponding to location i.

The invention has the advantages that: the indoor positioning model is modeled by using a sparse coding method, noise is processed by considering the minimum fitting error in a target equation, the sparsity of the obtained sparse vector is better ensured by introducing non-convex regularization, meanwhile, the convexity of the target equation is kept, and the rapid solution is facilitated; after the sparse vector is obtained, the final position of the target is estimated by simultaneously considering the position information of several maximum values, so that the indoor position of the target can be effectively predicted, and technical bases are provided for intelligent home control, target monitoring, bank anti-theft, intelligent medical treatment and the like.

Drawings

Fig. 1 is a schematic structural diagram of a sensor network provided in an embodiment of the present invention;

fig. 2 is a general schematic diagram of an indoor positioning method without a device according to an embodiment of the present invention;

fig. 3 is a flow chart of an indoor positioning method without a device according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

(as shown in fig. 1), the present invention first models an indoor positioning environment, and since the number of targets in the environment is always much less than the total number of meshes, we model the positioning problem as a sparsely coded problem. Firstly, test signals at all positions are collected to form a dictionary, sparse vectors containing positioning information can be obtained for the collected test signals in a sparse coding mode, and finally a method of joint analysis and calculation is used for predicting the target position.

As shown in fig. 2-3, the indoor positioning method without carrying equipment of the present invention sequentially comprises the following steps:

s101, obtaining quantitative data of shielding effects of the target on radio waves at different indoor positions and label data of the target position, and dividing the data to obtain a training sample set and a test sample set; the training sample set comprises received radio data of the target at different positions and a position label of the target; the method specifically comprises the following steps:

s1011, dividing the indoor positioning scene into square grids with the same size, wherein each grid represents the position where the target possibly exists and corresponds to a unique digital label;

s1012, arranging L radio receiving sensors uniformly distributed around a monitoring area, fixing the positions, configuring a radio transmitting sensor having K variable transmitting positions, wherein wireless communication between one transmitting sensor and one receiving sensor is called a set of links, and N ═ K ×, L sets of links are shared in the whole experimental scene;

s1013, when the target exists at a certain position in the room, the radio transmitting sensor transmits a radio signal, and all receiving sensors receive the signal; the transmitting sensor moves all the selectable positions and transmits radio signals, and all the receiving nodes receive arriving signals simultaneously; then, extracting signal strength from the frequency spectrum of each link as effective information, and obtaining an information matrix at each position of a target, wherein the dimension size is L rows and K columns; meanwhile, the digital label information of the target at each grid position is stored and can be used for corresponding to a real two-dimensional position coordinate;

s1014, for the collected data set, calculating the ratio of 5: 1 into a training sample set and a test sample set;

s102, processing the quantized data at all positions, processing the data at each position into a column vector, and forming a dictionary by all the obtained column vectors in a column splicing mode; the method specifically comprises the following steps:

s1021, firstly, converting an information matrix obtained by a target at a first position into a column vector, wherein the dimension is N, and the label information is 1;

s1022, repeating tau times of measurement on the first position, respectively converting the tau times of measurement into column vectors, and obtaining tau column vectors in total, wherein the label information is Lab₁＝{1,1,…,1}；

S1023, repeating the steps S1021 and S1022 until the data of the completion target on all the selectable positions are collected; respectively carrying out row-column splicing on the column vector at each position in sequence to form a final dictionary, and simultaneously carrying out column-splicing on corresponding label information to form labeling information;

s103, constructing a target equation, considering linear fitting errors and regularization terms of the test signals at the same time, and performing iterative solution to obtain a final sparse vector; the method specifically comprises the following steps:

s1031, firstly, constructing an object equation, wherein the object equation should include two parts, namely a fitting error and a regularization term, and the object equation is as follows:

where y represents a test signal, W is a dictionary matrix, x is the sparse vector sought,represents the introduced non-convex GMC regularization; both α and λ are positive parameters; z is an auxiliary vector;

s1032, because the number of the targets is always far smaller than the total number of the network grid points, a sparse coding method can be adopted to solve sparse vectors, and a forward and backward splitting algorithm can be adopted to solve the sparse coding;

s104, performing joint analysis calculation on a plurality of maximum values in the sparse vector to obtain a target prediction position, and specifically comprising the following specific steps:

s1041, first, a sparse vector x ═ x is obtained_1,1,x_1,2,…,x_1,τ,x_2,1,…x_S,τAfter that, x is equal to Rⁿ(ii) a By summing the x coefficients of each sparse vectorWherein 1. ltoreq. p. ltoreq.S, thus obtaining x^*＝{x₁,…,x_p,…x_S}，x^*In order to obtain the weight vectors corresponding to different predicted labels, the larger the weight is, the greater the probability of predicting the current position is, and x is calculated^*The elements in (1) are arranged in descending order;

s1042, extracting a plurality of maximum values in the weight vector, wherein each weight corresponds to a different position label, and the position labels can correspond to two-dimensional coordinates one by one;

s1043, normalizing the weights, and performing combined calculation to obtain the predicted target position. The specific mode is as follows:

first, several maximum value vectors are arranged according toNormalization is carried out, wherein num is the number of the maximum values, and normalized weights can be obtained;

the position of the predicted target may be calculated by:where loc denotes the position of the final target, a_iIndicating the normalized weight corresponding to the position i, and loc (i) indicating the true two-dimensional coordinate corresponding to the position i.