CN106559749B - Multi-target passive positioning method based on radio frequency tomography - Google Patents
Multi-target passive positioning method based on radio frequency tomography Download PDFInfo
- Publication number
- CN106559749B CN106559749B CN201611035940.XA CN201611035940A CN106559749B CN 106559749 B CN106559749 B CN 106559749B CN 201611035940 A CN201611035940 A CN 201611035940A CN 106559749 B CN106559749 B CN 106559749B
- Authority
- CN
- China
- Prior art keywords
- point
- points
- target
- pixel
- targets
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/02—Services making use of location information
- H04W4/023—Services making use of location information using mutual or relative location information between multiple location based services [LBS] targets or of distance thresholds
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S5/00—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations
- G01S5/02—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations using radio waves
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W64/00—Locating users or terminals or network equipment for network management purposes, e.g. mobility management
Abstract
The invention discloses a multi-target passive positioning method based on radio frequency tomography, which comprises the following steps: radio frequency tomography; extracting a local maximum value; cross section scanning: after all local maximum value points in the imaging result are obtained, carrying out gray scale cross section scanning on each maximum value point, taking the maximum value point as the center of a scanning straight line L, rotating the straight line at a fixed angular interval theta, recording the pixel value of a pixel point through which the straight line passes after each rotation, and stopping after rotating for one circle; calculating the characteristics of the gray distribution map; according to the characteristics of the gray distribution diagram, points to be classified are divided into three classes by using a naive Bayes classifier: the method comprises the steps of judging each hot spot by a pseudo target point, a single target point and two target points, removing pseudo targets and distinguishing the number of targets when a plurality of targets are gathered together. The invention has the characteristic of higher precision.
Description
Technical Field
The invention belongs to the research field of indoor personnel passive positioning by using UHF RFID equipment, and particularly relates to a radio frequency tomography algorithm, a gray scale cross section scanning analysis method and a Bayesian classification method.
Background
And positioning, namely acquiring the coordinates of the interested target in the space. Since the human being enters the information era, navigation and communication technologies are rapidly developed in mutual interaction, wherein the personal positioning navigation technology of a user generates greater application value and plays an increasingly greater role in the daily life of people, and meanwhile, single position navigation is also converted into comprehensive position services such as monitoring, management, traffic, logistics, rescue, entertainment and the like. Location services have entered people's lives as a strategic emerging industry and are becoming an indispensable part of social life, economic construction, and even national defense security.
Currently, many positioning methods have been applied to intrusion detection, location and tracking systems for objects. In general, positioning systems can be divided into two categories: cooperative and uncooperative systems. The cooperative system requires the user to carry some auxiliary positioning device (such as a mobile phone, a GPS receiver or a near field communication tag) to assist the positioning, i.e. active positioning. For example, the GPS system predicts the position of the receiver by measuring received signals from a plurality of satellites. The RFID system in the Internet of things reads the information of the RFID label by utilizing the reader, and the reader can be used for automatically identifying and tracking the position of the label, so that automatic warehouse logistics distribution management is realized. Basically, the cooperative positioning method is to measure some physical quantity of a wireless signal emitted by a user or a target, and then to perform positioning by using a three-edge (or multi-edge) positioning method or a triangulation method.
Non-cooperative positioning is also known as passive positioning (DFL). Passive positioning[1]The method does not need any auxiliary equipment carried by the user, and has great application prospect in the fields of emergency rescue (such as fire and earthquake), intrusion detection, intelligent home and the like. The current passive positioning algorithm mainly performs positioning by analyzing interference caused by a target appearing in a positioning area to a signal. However, the realization of the target passive wireless positioning and tracking with high precision, strong anti-interference capability and adaptation to complex environments is still a problem to be solved urgently under the influence of factors such as complex channel environment and weak algorithm anti-interference.
Radio-frequency tomography (RTI) is a novel passive localization method that has emerged in recent years. The basic idea is to divide the positioning area into grids (pixel points) of equal size, and the attenuation of each link in the network is equivalent to the sum of the attenuation values of all grids passed by the link. And then establishing a mathematical model, and solving a matrix equation set to reversely solve the attenuated pixel value of each pixel point. The point at which the attenuation value is highest is taken as the final target position. The algorithm has the advantages of low calculation complexity, high positioning precision, no need of training data and good real-time performance. At present, radio frequency tomography algorithms are basically based on Wireless Sensor Networks (WSNs), and in fact, RFID systems widely applied in the internet of things can be used not only for active positioning but also for passive positioning. Meanwhile, the passive tag in the RFID system does not need to be powered by a battery, so that the cost is low and the maintenance cost is low.
Disclosure of Invention
The invention provides a multi-target passive positioning method based on radio frequency tomography, which can effectively eliminate false targets and realize multi-target positioning when the number of targets is unknown. The technical scheme of the invention is as follows:
a multi-target passive positioning method based on radio frequency tomography comprises the following steps:
1) radio frequency tomography
Firstly, arranging l passive UHFRFID tags around a positioning area, arranging a reader antenna at the midpoint of each edge, uniformly dividing the positioning area into N grids, wherein the grids are called pixel points, and each reader and the tags can communicate with each other. The radio frequency tomography algorithm linear model is that y is Wx + n, wherein y represents a Received Signal Strength (RSS) attenuation value of a link, x represents an RSS attenuation value of each pixel point in an area, W represents a weight value of each pixel point to each link, and n is additive white Gaussian noise; and calculating a weight matrix W, measuring the RSS variation y of each link, further calculating x, and displaying in an imaging mode to obtain a positioning result.
2) Local maximum extraction
Firstly, local maximum extraction is carried out on x of an imaging result image, namely all hot spots in the image are found out, and the hot spots can be real target positions or pseudo target points. The method comprises the following steps: the judgment is carried out by comparing the size of each pixel point with the size of eight or sixteen adjacent pixel points around the pixel point, if the value of the pixel point is maximum, the pixel point is considered as a local maximum value point, the algorithm is executed on all the pixel points in the image, and all the local maximum value points are found out.
3) Cross-section scanning
After all local maximum value points in the imaging result are obtained, gray scale cross section scanning is carried out on each maximum value point, the maximum value point is used as the center of a scanning straight line L, the straight line is rotated at a fixed angular interval theta, the pixel value of a pixel point through which the straight line passes after each rotation is recorded, the rotation is stopped after one circle, and the length unit of L is the number of the pixel points.
4) Feature calculation of gray-scale distribution map
Several variables including the width, height, rise and fall of the peak are set to record the gray profile characteristics for each direction.
5) Classifying points to be classified using a naive Bayes classifier
According to the characteristics of the gray distribution diagram, points to be classified are divided into three classes by using a naive Bayes classifier: the method comprises the steps of judging each hot spot by a pseudo target point, a single target point and two target points, removing pseudo targets and distinguishing the number of targets when a plurality of targets are gathered together.
Drawings
FIG. 1 is a block diagram of the process of the present invention.
Fig. 2(a) and (b) are schematic diagrams of the passive UHFRFID positioning scenario and its link, respectively.
FIG. 3 is a diagram illustrating a positioning result when a plurality of targets appear.
Fig. 4 is a schematic diagram of a cross-sectional scanning process.
Fig. 5 is a schematic diagram of feature extraction of a gray distribution map.
Fig. 6 is a schematic diagram of simulation results.
Detailed Description
The UHF RFID multi-target passive positioning method based on the radio frequency tomography comprises the following steps:
1) radio frequency tomography
Firstly, arranging l passive UHF RFID tags around a positioning area, and arranging a reader antenna at the midpoint of each edge. The positioning area is evenly divided into N grids, and the grids are called pixel points. Communication may be between each reader and tag. As the various nodes in fig. 2 communicate, the wireless link may traverse the area covered by the nodes. The presence of the target can affect the surrounding link because the target reflects, absorbs, diffracts or scatters a portion of the transmitted power. Thereby causing the power to the receiver of the link to drop. The radio frequency tomography algorithm linear model is that y is Wx + n, wherein y represents a Received Signal Strength (RSS) attenuation value of a link, x represents an RSS attenuation value of each pixel point in an area, W represents a weight value of each pixel point to each link, and n is additive white Gaussian noise. As long as we calculate the weight matrix W and measure the RSS variation y of each link. We can find x. After x is obtained, the x is displayed in an imaging mode, and then the positioning result can be obtained. However, when a plurality of targets are present in the positioning region, a false target point may appear in the final imaging result, as shown in fig. 3. The patent proposes an image processing method based on cross section scanning to remove false targets in the imaging result. The method comprises the following specific steps.
2) Local maximum extraction
Firstly, local maximum extraction is carried out on the imaging result image x, namely all 'hot spots' in the image are found out. As "hot spots" may all be true target locations. The simplest method is to judge by comparing the size of each pixel point with the size of eight or sixteen adjacent pixel points around the pixel point, and if the value of the pixel point is the maximum, the pixel point is regarded as a local maximum value point. The algorithm is executed for all pixel points in the image, and all local maximum value points are found out.
3) Cross-section scanning
After all the local maximum points in the imaging result are obtained, the gray scale cross section scanning is performed on each maximum point, which is also the core idea of the method proposed by the present invention. And (3) taking the maximum value point as the center of the scanning straight line L, rotating the straight line at a fixed angular interval theta, recording the pixel value of the pixel point through which the straight line passes after each rotation, and stopping after one rotation. The length unit of L is the number of pixels. As shown in fig. 4.
4) Feature calculation of gray-scale distribution map
How to convert the difference information on the graph into digital quantity to realize positioning is an essential step. Because the actual positioning is processed by a computer to finally give the coordinates of the target, rather than being artificially searched in the image. First we set several variables to record the gray profile characteristics for each direction, including peak width, height, rise rate, fall rate, etc. As shown in fig. 5.
5) Classifying points to be classified using a naive Bayes classifier
A naive bayes classifier is the simplest of bayesian classifications. The basic idea is as follows: for a given item to be classified, the probability of each class appearing on the premise that each feature of the item appears is solved, and the item to be classified is assigned to the class with the highest probability. The only problem is how to get the posterior probability, and bayes' theorem tells us that the posterior probability can be expressed as long as the prior probability is known. And the prior probability can be obtained by learning the sample.
The problem under investigation of this patent can be seen as a classification problem. It should be emphasized that the proposed method not only can remove the false target, but also can further distinguish the number of targets when multiple targets are gathered together. Because when a plurality of targets are close to each other, the gray scale distribution map at the position has obvious changes in the characteristics of the height, the width and the like of the peak compared with the single target. From this we can estimate the number of objects at that location. The Bayes classifier aims to judge whether a 'hot spot' belongs to a pseudo target point, a single target point, a double target point, a triple target point and the like. In theory it should be possible to resolve a larger number of targets. However, in consideration of the real scene, the RSS value of the signal is easily affected by the surrounding environment, and the communication distance between the tag and the reader is limited, the experiment only considers the case when two targets are very close to each other. Namely, the points to be classified are divided into three categories: a pseudo target point, a single target point, and two target points.
In order to further illustrate the present invention, reference is made to the following examples.
(1) Fig. 2 shows a schematic diagram of a UHF RFID multi-target passive location scene, which is a location area of 10m × 10m, and 40 tags and 4 readers are respectively placed around the location area. There are 160 signal links and the location area is divided into 1600 pixels of 40 x 40. Here we model the target (here the person) as a circle with a diameter set to 0.5 meters. If the Euclidean distance between a certain link and the center of the circle is smaller than the radius of the circle, the link is considered to be influenced by the target. For an RFID network, a complete communication link consists of two parts, a forward link from the reader antenna to the tag and a backward link from the tag to the reader. The forward link is mainly where the reader antenna energizes the tag by electromagnetic waves, i.e., activates the tag, and the activated tag reflects its stored information back to the reader through the backward link. Researchers have long studied the impact of personnel on RFID communication links and proposed several models. We use one of the models:
in the above formula, the first and second carbon atoms are,
for the RSS variation of the affected link, dexcRepresents the path difference, i.e., the sum of the sender-to-target distance and the target-to-recipient distance minus the sender-to-recipient distance.
Is the wavelength of the radio signal, phireflIs the initial reference phase. A and B are undetermined constants and are related to the surrounding environment.
The weight matrix W is calculated by:
in the above formula diRepresents the length of the ith link, dij(1) And dij(2) Respectively representing the distances from the center of the jth pixel point to the two end points of the link i. Lambda is an adjustable parameter and controls the size of the ellipse. This model is commonly referred to as an elliptical weight model. The method of least square method and regularization is applied to the RTI linear model y ═ Wx + n, and the method can be obtained as follows:
wherein
As a regularization parameter，djkRepresenting the distance between the center of pixel point j and the center of pixel point k,
is the variance of the pixel attenuation values, δcIs a space constant.
(2) For the imaging result image
And extracting local maximum values, and finding out all 'hot spots' in the image. These local maximum points may be true targets or false targets. Here, we compare the pixel value (attenuation value) of each pixel point in the image with the sixteen adjacent pixel points around the pixel point to judge, and if the value of the pixel point is maximum, the pixel point is considered as a local maximum point. The algorithm is executed for all pixel points in the image, and all local maximum value points are found out. The coordinates of all local maximum points are recorded.
(3) After all the local maximum points in the imaging result are obtained, the gray scale cross section scanning is performed on each maximum point, which is also the core idea of the method proposed by the present invention. And (3) taking the maximum value point as the center of the scanning straight line L, rotating the straight line at a fixed angular interval theta, recording the pixel value of the pixel point through which the straight line passes after each rotation, and stopping after one rotation. Here, the length of L is 9 pixels, and θ is 15 degrees.
(4) After obtaining the gray profile of each direction, we set several variables to record the characteristics of the gray profile of each direction, including the width, height, rising speed, falling speed, etc. of the peak:
a) height of ascending section: hrise＝G[risee]-G[rises]
b) Height of descending section: hdown＝G[downs]-G[downe]
c) Width of rising section: wrise＝X[risee]-X[rises]
d) Width of falling interval: wdown＝X[downe]-X[downs]
e) Slope of rising interval: srise＝Hrise/Wrise
f) Slope of the falling interval: sdown＝Hdown/Wdown
g) Peak width (end of falling interval minus start of rising interval): wpeak＝X[downe]-X[rises]
h) Peak height:
Hpeak＝G[center]-(G[downe]-G[rises])/Wpeak·(X[center]-X[rises])+G[rises]
a) kurtosis:
in order to make the obtained data features more concise, corresponding statistics are defined to record the feature information. Common statistics are expectation, variance, standard deviation, median, mode, etc. We define the following set to store the characteristics of each point:
T＝{μWpeak,σWpeak,μHpeak,μSrise,μSdown,μκ,σκ}
the above feature set records statistical information of different features of the gray distribution map in all directions of a point, muiAs desired for the variable i, σiIs the variance of i. Mu.sWpeakReflecting the width, σ, of the main peakWpeakReflects the symmetry of the main peak, muHpeakReflects the height of the main peak, muSriseAnd muSdownThe steepness of the peak, μκAnd σκThe degree of similarity of the gray distribution to the normal distribution is described.
(5) And finally, learning the set T by using a naive Bayes classifier, obtaining the prior probability of each category through sample training, solving the posterior probability of the target belonging to different categories by using the prior probability in a positioning stage, and judging whether the target is a real target or a false target by comparing the sizes of the targets. The simulation results are shown in fig. 6. The positioning results of four, five and six targets are shown in the figure from top to bottom. The imaging result of the traditional RTI algorithm in the passive RFID scene is shown on the left, and the positioning result processed by the method in the patent is shown on the right. It can be found that when the number of targets reaches more than four, a simple RTI algorithm cannot provide an accurate positioning result, and the occurrence of false targets greatly increases the positioning difficulty. After the processing is carried out by the method provided by the invention, the false targets can be accurately identified, and the positions of all real targets can be obtained.
Claims (1)
1. A multi-target passive positioning method based on radio frequency tomography comprises the following steps:
1) radio frequency tomography
Firstly, arranging m passive UHF RFID tags around a positioning area, arranging a reader antenna at the midpoint of each edge, uniformly dividing the positioning area into N grids, wherein the grids are called pixel points, and each reader and the tags can communicate; the radio frequency chromatography algorithm linear model is that y is Wx + n, wherein y represents a Received Signal Strength (RSS) attenuation value of a link, x represents an RSS attenuation value of each pixel point in an area, W is a weight matrix and represents the weight value of each pixel point to each link, and n is additive white Gaussian noise; calculating a weight matrix W, measuring the RSS variation y of each link, further calculating x, and displaying in an imaging manner to obtain a positioning result,
the weight matrix W elements are calculated by:
in the above formula diRepresents the length of the ith link, dij(1) And dij(2) Respectively representing the distance from the center of the jth pixel point to two end points of the link i, wherein lambda is an adjustable parameter, and controlling the size of the ellipse;
2) local maximum extraction
Firstly, extracting local maximum values of x of an imaging result image, namely finding out all hot spots in the image, wherein the hot spots can be real target positions or pseudo target points; the method comprises the following steps: judging by comparing the size of each pixel point with the size of eight or sixteen adjacent pixel points around the pixel point, if the value of the pixel point is maximum, considering the pixel point as a local maximum point, executing a local maximum extraction algorithm on all the pixel points in the image, and finding out all the local maximum points;
3) cross-section scanning
After all local maximum value points in the imaging result are obtained, carrying out gray scale cross section scanning on each maximum value point, taking the maximum value point as the center of a scanning straight line L, rotating the straight line at a fixed angular interval theta, recording the pixel value of a pixel point through which the straight line passes after each rotation, and stopping after one rotation, wherein the length unit of L is the number of the pixel points;
4) feature calculation of gray-scale distribution map
Setting variables to record the characteristics of the gray distribution map in each direction, wherein the variables are as follows:
a) height of ascending section: hrise
b) Height of descending section: hdown
c) Width of rising section: wrise
d) Width of falling interval: wdown
e) Slope of rising interval: srise＝Hrise/Wrise
f) Slope of the falling interval: sdown＝Hdown/Wdown
g) Peak width, i.e. the end point of the falling interval minus the start point of the rising interval: wpeak
h) Peak height: hpeak
i) Kurtosis: kappa
5) Classifying points to be classified using a naive Bayes classifier
According to the characteristics of the gray distribution diagram, points to be classified are divided into three classes by using a naive Bayes classifier: the method comprises the steps of judging each point to be classified according to a pseudo target point, a single target point and two target points, removing pseudo targets and distinguishing the number of targets when a plurality of targets are gathered together.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611035940.XA CN106559749B (en) | 2016-11-22 | 2016-11-22 | Multi-target passive positioning method based on radio frequency tomography |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611035940.XA CN106559749B (en) | 2016-11-22 | 2016-11-22 | Multi-target passive positioning method based on radio frequency tomography |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106559749A CN106559749A (en) | 2017-04-05 |
| CN106559749B true CN106559749B (en) | 2020-04-17 |
Family
ID=58444880
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611035940.XA Active CN106559749B (en) | 2016-11-22 | 2016-11-22 | Multi-target passive positioning method based on radio frequency tomography |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106559749B (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107480606A (en) * | 2017-07-28 | 2017-12-15 | 天津大学 | Pseudo- target identification method based on radio frequency tomography |
| CN107422296B (en) * | 2017-07-31 | 2018-07-13 | 河南工业大学 | Wireless tomography determines method with grid pixel pad value |
| CN107666705B (en) * | 2017-08-23 | 2020-04-10 | 中山大学 | Dual space back projection radio frequency tomography method, positioning method and device |
| CN107945240B (en) * | 2017-10-13 | 2021-04-13 | 西北大学 | Target imaging method |
| CN108036730B (en) * | 2017-12-22 | 2019-12-10 | 福建和盛高科技产业有限公司 | Fire point distance measuring method based on thermal imaging |
| CN108769898B (en) * | 2018-05-16 | 2020-12-01 | 辽宁师范大学 | Wireless network passive positioning method based on layered radio frequency imaging |
| CN110471030A (en) * | 2019-08-23 | 2019-11-19 | 天津大学 | Based on the radio frequency tomography passive type Position-Solving method for improving conjugate gradient |
| CN111313999B (en) * | 2020-02-18 | 2022-04-15 | 五邑大学 | Radio frequency tomography method based on zero sparse data driven weight model |
| CN111999697B (en) * | 2020-04-30 | 2022-05-20 | 南京理工大学 | Self-positioning method for nodes of three-dimensional wireless tomography system on two sides of road |
| CN111476790A (en) * | 2020-05-14 | 2020-07-31 | 武汉中旗生物医疗电子有限公司 | Method and device for enhancing display of puncture needle in ultrasonic puncture |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105960017A (en) * | 2016-06-24 | 2016-09-21 | 北京理工大学 | Ultra-wideband node network-based device-free localization method |
| CN106019209A (en) * | 2016-05-12 | 2016-10-12 | 天津大学 | Indoor person device-free localization method based on radio tomography imaging |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2010030956A2 (en) * | 2008-09-12 | 2010-03-18 | University Of Utah Research Foundation | Method and system for tracking objects using radio tomographic imaging |
-
2016
- 2016-11-22 CN CN201611035940.XA patent/CN106559749B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106019209A (en) * | 2016-05-12 | 2016-10-12 | 天津大学 | Indoor person device-free localization method based on radio tomography imaging |
| CN105960017A (en) * | 2016-06-24 | 2016-09-21 | 北京理工大学 | Ultra-wideband node network-based device-free localization method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106559749A (en) | 2017-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106559749B (en) | Multi-target passive positioning method based on radio frequency tomography | |
| Luo et al. | Temporal convolutional networks for multiperson activity recognition using a 2-d lidar | |
| US20200143561A1 (en) | Multi-Channel Spatial Positioning System | |
| Chen et al. | Activity recognition based on micro-Doppler signature with in-home Wi-Fi | |
| Huang et al. | An accurate and efficient device-free localization approach based on sparse coding in subspace | |
| Ding et al. | Human object estimation via backscattered radio frequency signal | |
| Geng et al. | Indoor tracking with RFID systems | |
| CN111505632A (en) | Ultra-wideband radar action attitude identification method based on power spectrum and Doppler characteristics | |
| CN115620212B (en) | Behavior identification method and system based on monitoring video | |
| Ding et al. | Counting human objects using backscattered radio frequency signals | |
| Han et al. | A wireless fingerprint location method based on target tracking | |
| Wang et al. | Multi-classification of UWB signal propagation channels based on one-dimensional wavelet packet analysis and CNN | |
| CN114814832A (en) | Millimeter wave radar-based real-time monitoring system and method for human body falling behavior | |
| Hu et al. | Building occupancy detection and localization using cctv camera and deep learning | |
| CN110413655B (en) | Floor identification method based on improved hidden Markov model | |
| CN111444926A (en) | Radar-based regional people counting method, device, equipment and storage medium | |
| CN111598158B (en) | Warehouse inventory state sensing method based on AI analysis technology | |
| US11495111B2 (en) | Indoor occupancy estimation, trajectory tracking and event monitoring and tracking system | |
| Tian et al. | Indoor device-free passive localization for intrusion detection using multi-feature PNN | |
| Mu et al. | Human indoor positioning via passive spectrum monitoring | |
| CN115291184B (en) | Attitude monitoring method combining millimeter wave radar and deep learning | |
| Haferkamp et al. | A low cost modular radio tomography system for bicycle and vehicle detection and classification | |
| CN116561700A (en) | Indoor human body posture recognition method based on millimeter wave radar | |
| CN105844756A (en) | People counting method based on radio frequency backscatter signal | |
| Zhu et al. | Environmentally adaptive real-time detection of RFID false readings in a new practical scenario |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |