CN103607734B - The monitoring of anomalous event based on compressed sensing and localization method - Google Patents

Abstract

The present invention provides a method for monitoring and locating abnormal events based on compressed sensing, comprising: selecting m _p sampling sensors whose monitoring signal strength is greater than a preset threshold ρ from n sampling sensors to form the first batch of sampling initiation nodes; Select log(q) sampling sensors for each sampling sensor in the first batch of sampling initiating nodes from n sampling sensors to form the second batch of sampling initiating nodes; take each sampling initiating node as the starting point and take the data collection node as End point, select H sampling sensors between each start point and end point among n sampling sensors to form a sampling transmission path corresponding to each start point and end point, where H=O(logq); the data collection node receives The location and intensity of the anomalous event are calculated from the sampled data of the new anomalous event. The invention can use the sampling sensor network to monitor in real time and locate the events occurring in the monitoring area on the basis of low energy consumption, with high accuracy and less false alarms.

Description

Monitoring and Location Method of Abnormal Events Based on Compressed Sensing

technical field

The invention belongs to the technical field of sensor applications, in particular to a method for monitoring and locating abnormal events based on compressed sensing.

Background technique

The sensor monitoring network is arranged in the monitoring area to monitor the environmental parameters and abnormal events in the area. The nodes in the sensor network have the functions of perception, calculation, storage, communication and so on. Perceive, record and analyze specific environmental parameters, and the sensor network can monitor whether there are abnormal events in the monitoring area in real time.

Existing methods for event detection suffer from many shortcomings. First, the modeling of the event does not correspond to the actual situation. The event is simply modeled as a signal area, and the signal strength in the area is the same. Second, the accuracy of the event monitoring mechanism of existing methods is low. By setting the signal strength threshold and monitoring that a specific signal strength exceeds the threshold, a single sensor can consider that an event has occurred, and there are many false positives and omissions in traditional methods. Third, the transmission cost of the traditional method is very high, which makes the sensor network high energy consumption and short life.

Contents of the invention

The purpose of the present invention is to provide a method for monitoring and locating abnormal events based on compressed sensing, which can use the sampling sensor network to monitor and locate events in the monitoring area in real time on the basis of low energy consumption, with high accuracy and few false alarms. The number of abnormal events may be multiple, and the signal sources generated by the abnormal events may have different signal intensities.

In order to solve the above problems, the present invention provides a method for monitoring and locating abnormal events based on compressed sensing, including:

Distributing n sampling sensors in the monitoring area and dividing the monitoring area into q intervals according to the situation of the monitoring area, and setting a data collection node in the monitoring area;

Select m _p sampling sensors whose monitoring signal strength is greater than the preset threshold ρ from n sampling sensors to form the first batch of sampling initiation nodes;

From n sampling sensors, select log(q) sampling sensors for each sampling sensor in the first batch of sampling initiating nodes to form the second batch of sampling initiating nodes;

Taking each sampling initiating node in the first batch of sampling initiating nodes and the second batch of sampling initiating nodes as a starting point and taking the data collection node as an end point, select between each starting point and end point among n sampling sensors The H sampling sensors form a corresponding sampling transmission path between each start point and end point, where H=O(logq);

Each sampling initiation node sends the sampling data of the detected abnormal event to the adjacent sampling sensor on the corresponding sampling transmission path, and each sampling sensor on each sampling transmission path compares the sampling data of the abnormal event it monitors with the sampling data from the sampling sensor. The sampling data of the abnormal event received by the initiating node or the previous adjacent sampling sensor on the sampling transmission path is weighted and formed into new sampling data of the abnormal event, and then sent to the next adjacent sampling sensor or the data on the sampling transmission path collection node;

The data collection node calculates the position and intensity of the abnormal event in the monitoring area according to the received sampling data of the new abnormal event.

Further, in the above method, the calculation formula of the threshold ρ is as follows:

$\mathrm{<span\; class="notranslate">\; \&rho;</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; n\&pi;</span>}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}}<span\; class="notranslate">\; -</span>\sqrt{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}}\mathrm{<span\; class="notranslate">\; er</span>}{\mathrm{<span\; class="notranslate">\; f</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0.5</span>}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>$

Among them, n is the number of sampling sensors distributed in the monitoring area, s is the area of the monitoring area, λ is the parameter value of an abnormal event, μ _λ is the average value of the parameter values of all abnormal events, and σ _λ is the abnormal event The standard deviation of the parameter value of , erf is the Gaussian error function, erf defines a cumulative probability distribution function of a normally distributed random variable, and the parameter value λ of each abnormal event conforms to the following normal distribution

Compared with the prior art, the present invention distributes n sampling sensors in the monitoring area according to the situation of the monitoring area and divides the monitoring area into q intervals, and sets a data collection node in the monitoring area; from n Among the sampling sensors, select m _p sampling sensors whose monitoring signal strength is greater than the preset threshold ρ to form the first batch of sampling initiation nodes; select log for each sampling sensor in the first batch of sampling initiation nodes from the n sampling sensors (q) sampling sensors form a second batch of sampling initiating nodes; each sampling initiating node in the first batch of sampling initiating nodes and the second batch of sampling initiating nodes is a starting point and the data collection node is an end point, respectively Select H sampling sensors between each start point and end point among n sampling sensors to form a sampling transmission path corresponding to each start point and end point, wherein, H=O(logq); each sampling initiation node will monitor The sampling data of the abnormal event is sent to the adjacent sampling sensor on the corresponding sampling transmission path, and each sampling sensor on each sampling transmission path compares the sampling data of the abnormal event detected by itself with the sampling data from the sampling initiation node or the sampling transmission path The sampling data of the abnormal event received by the previous adjacent sampling sensor is weighted and formed into the sampling data of the new abnormal event, and then sent to the next adjacent sampling sensor or the data collection node on the sampling transmission path; the data collection node Calculate the position and intensity of the abnormal event in the monitoring area according to the received sampling data of the new abnormal event, and can use the sampling sensor network to monitor and locate the events occurring in the monitoring area in real time on the basis of low energy consumption, with high accuracy High, less false positives, the number of abnormal events can be multiple, and the signal sources of abnormal events can have different signal strengths.

Description of drawings

1 is a flowchart of a method for monitoring and locating abnormal events based on compressed sensing according to an embodiment of the present invention;

2 is a schematic diagram of a method for monitoring and locating abnormal events based on compressed sensing according to an embodiment of the present invention;

Fig. 3 is an example of signal strength attenuation of an abnormal event signal source and an example of signal strength monitored by sampling sensors in multiple event overlapping regions according to an embodiment of the present invention.

detailed description

In order to make the above objects, features and advantages of the present invention more comprehensible, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

The invention provides a method for monitoring and locating abnormal events based on compressed sensing. The compressed sensing technology is originally used to restore sparse signals sampled at low sampling frequencies. Based on the sparsity of the signal, the sampling frequency can be greatly reduced, and the original signal can be recovered correctly. The invention applies the compressed sensing technology to the event monitoring and positioning of the sensor monitoring network. Since the simultaneous emergencies in the monitoring area are sparse, the location of the event and the characteristic parameters of the event can be estimated by applying compressed sensing technology and only needing less data sampling of the sensor network. Sampling sensor network data can greatly reduce the sampling frequency of sensor network data, reduce transmission and calculation energy consumption, and prolong the service life of sensors and network life, as shown in Figures 1-3. The methods include:

Step S1, distributing n sampling sensors 1 in the monitoring area according to the situation in the monitoring area 5 and dividing the monitoring area into q intervals 2, and setting a data collection node 3 in the monitoring area;

Step S2, select m _p sampling sensors whose monitoring signal strength is greater than the preset threshold ρ from n sampling sensors to form the first batch of sampling initiating nodes 11; specifically, the sampling initiating process can ensure that a sufficient number of sampling initiating nodes are generated , assuming that the monitoring area is divided into q intervals on average, the positioning accuracy can be controlled by the interval size, and there are k abnormal events in the area, then according to the compressed sensing technology, only O(k log n) data sampling is required, that is The positions and signal strengths of multiple abnormal events can be accurately estimated, and the sampling initiation process of the present invention can ensure that O(klogq) sampling initiation nodes are generated, and each sampling initiation node is generated in a distributed manner. When a certain sampling sensor node detects When the received event signal strength meets the set threshold, it becomes the first batch of sampling initiation nodes;

Step S3, select log(q) sampling sensors for each sampling sensor in the first batch of sampling initiating nodes 11 from the n sampling sensors to form the second batch of sampling initiating nodes 11; specifically, how many measurement m is determined by k (the number of abnormal events) and q (the number of divided intervals), that is, m=o(poly(k,logq));

Step S4, starting from each sampling initiating node in the first batch of sampling initiating nodes and the second batch of sampling initiating nodes and taking the data collection node as an end point, selecting each starting point and The H sampling sensors between the end points form a sampling transmission path corresponding to each start point and end point, where H=O(logq); specifically, according to the compressed sensing technology, an effective data sampling is a plurality of sensor data The weighted sum, assuming H, then H needs to meet certain conditions, that is, H=O(logq), therefore, a data sampling packet must pass through H sensors, and finally be delivered to a data collection node, that is, the sampling path There needs to be a certain length limit (H hops, H=O(logq)). On the sampling path, each sampling sensor will determine which neighbor to transmit to the next hop according to its hop distance from the data collection node (optional transmission Adjust the length of the sampling path to nodes closer or farther away from the collection node); the basis of the present invention is the sampling sensor monitoring network arranged in the monitoring area, and the sampling sensor monitoring network needs to have basic sensing, communication, computing and storage function. In order to monitor different emergencies, the sensor monitoring network should have corresponding sensing functions, for example, the ability to perceive environmental parameters such as temperature, humidity, light, noise, air pressure, etc. The occurrence of emergencies is usually accompanied by abnormal environmental parameters For example, forest fires are usually accompanied by abnormal temperature, light, and smoke. Some existing event monitoring methods simply set thresholds for specific environmental parameters. When environmental parameters higher than the threshold are detected, it is determined that an event has occurred. , this method is easy to cause false positives, false negatives and omissions. The present invention models the event-related environmental parameters. The event center is modeled as a signal source with the highest signal strength value, and the signal strength attenuates with the increase of distance. Heterogeneous events mean that the signal strength of the event signal source is different. As shown in Figure 3, the signal strength of abnormal event A or abnormal event B decreases with the increase of distance, and the sampling sensor i is between abnormal event A and abnormal event B. The signal intensity monitored in the overlapping area is y _i =y _i (B)+y _i (A); the present invention does not simply collect the monitoring data of all sensors and then analyze whether there is an abnormal event, so It will generate a huge transmission cost, and then quickly consume the electric energy of the sensor. The present invention uses compressed sensing technology to perform a small amount of data sampling on the sensor data, and can recover the positions of the signal sources of multiple events, thereby extremely Greatly reduce the energy consumption of sensor networks;

Step S5, each sampling initiation node sends the sampled data of the monitored abnormal event to the adjacent sampling sensor on the corresponding sampling transmission path, and each sampling sensor on each sampling transmission path sends the sampled data of the abnormal event it monitors Weighted with the sampling data of the abnormal event received from the sampling initiation node or the previous adjacent sampling sensor on the sampling transmission path to form new sampling data of the abnormal event, and then send it to the next adjacent sampling sensor on the sampling transmission path or The data collection node; specifically, in the case of an unknown number of abnormal events, the present invention includes a distributed sampling initiation process to select an appropriate number of sampling initiation nodes, each sampling initiation node initiates a sampling, and each sampling A sampling data packet will be generated, which contains the weighted sum of a sufficient amount of sensor data, and is transmitted to the data collection node through multi-hops. In order to ensure that each sampling data packet contains a sufficient amount of sensor data, each sampling needs to go through A sufficient number of sampling sensor nodes on the sampling transmission path, specifically O(logq), the present invention models abnormal events as signal sources, and the signal strength of sampling sensors in a certain area decays with distance, and different abnormal events Different signal source strengths and attenuation parameters are called heterogeneous events. The heterogeneous events in the monitoring area are modeled as signal sources with different positions, signal source strengths, and signal attenuation parameters. The invention monitors and locates multiple heterogeneous events. There is no additional cost for qualitative events, accurate positioning, no need for any central control mechanism, and the sampling sensor node operates autonomously; each sampling initiation node will initiate a sampling process, and the sampling process can ensure that the sampling data of a sufficient number of sampling sensors is passed through the weighted sum The method is included in the sampling data packet, and each sampling data packet can finally be delivered to the data collection node;

Step S6, the data collection node calculates the position and intensity of the abnormal event in the monitoring area according to the received sampling data of the new abnormal event. Specifically, the only data collection node in the network has the functions of data collection, calculation and analysis, and reporting of event monitoring and positioning results. After collecting the sampling data packets formed in the sampling stage, the data collection node uses compressed sensing technology to estimate and report the number and location of events, and then gives the location and intensity of the event signal source in the monitoring area. The process of data recovery can take into account normally distributed data noise.

Preferably, the formula for calculating the threshold ρ is as follows:

$\mathrm{<span\; class="notranslate">\; \&rho;</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; n\&pi;</span>}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}}<span\; class="notranslate">\; -</span>\sqrt{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}}\mathrm{<span\; class="notranslate">\; er</span>}{\mathrm{<span\; class="notranslate">\; f</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0.5</span>}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>$

Among them, n is the number of sampling sensors distributed in the monitoring area, s is the area of the monitoring area, λ is the parameter value of an abnormal event, μ _λ is the mean value of the parameter values of all abnormal events, σ _λ is all abnormal events The standard deviation of the parameter value of , erf is the Gaussian error function, erf defines a cumulative probability distribution function of a normally distributed random variable, and the parameter value λ of each abnormal event conforms to the following normal distribution Specifically, if $\mathrm{\&rho;}=\frac{\mathrm{n\&pi;}}{\mathrm{the\; s}}({\mathrm{\&mu;}}_{\mathrm{\&lambda;}}-\sqrt{2}{\mathrm{\&sigma;}}_{\mathrm{\&lambda;}}\mathrm{er}{f}^{-1}(2{\mathrm{\&delta;}}_p-1)),0.5<{\mathrm{\&delta;}}_p<1,$ Then Pr(m _p >k)>δ _p .

According to the situation of the monitoring area, the present invention distributes n sampling sensors in the monitoring area and divides the monitoring area into q intervals, and sets a data collection node in the monitoring area; selects monitoring signals from n sampling sensors m _p sampling sensors whose intensity is greater than the preset threshold ρ form the first batch of sampling initiation nodes; from n sampling sensors, log(q) sampling sensors are selected for each sampling sensor in the first batch of sampling initiation nodes to form The second batch of sampling initiating nodes; starting from each sampling initiating node in the first batch of sampling initiating nodes and the second batch of sampling initiating nodes and taking the data collection node as an end point, selecting among n sampling sensors respectively H sampling sensors between each start point and end point form a sampling transmission path corresponding to each start point and end point, wherein, H=O(logq); each sampling initiation node sends the sampled data of the detected abnormal event To the adjacent sampling sensor on the corresponding sampling transmission path, each sampling sensor on each sampling transmission path compares the sampling data of the abnormal event detected by itself with the sampling data received from the sampling initiation node or the previous adjacent sampling sensor on the sampling transmission path The sampling data of the abnormal event is weighted and the sampling data of the new abnormal event is sent to the next adjacent sampling sensor or the data collection node on the sampling transmission path; the data collection node according to the received new abnormal event The sampling data of the event calculates the position and intensity of the abnormal event in the monitoring area, which can use the sampling sensor network to monitor and locate the event in the monitoring area in real time on the basis of low energy consumption, with high accuracy, less false alarms, and abnormal The number of events may be multiple, and the signal sources generated by the abnormal events may have different signal strengths.

Each embodiment in this specification is described in a progressive manner, each embodiment focuses on the difference from other embodiments, and the same and similar parts of each embodiment can be referred to each other. As for the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and for relevant information, please refer to the description of the method part.

Professionals can further realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, computer software or a combination of the two. In order to clearly illustrate the possible For interchangeability, in the above description, the composition and steps of each example have been generally described according to their functions. Whether these functions are executed by hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be regarded as exceeding the scope of the present invention.

Obviously, those skilled in the art can make various changes and modifications to the invention without departing from the spirit and scope of the invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and equivalent technologies thereof, the present invention also intends to include these modifications and variations.

Claims (2)

1. A method for monitoring and locating abnormal events based on compressed sensing, characterized in that, comprising: Distributing n sampling sensors in the monitoring area and dividing the monitoring area into q intervals according to the situation of the monitoring area, and setting a data collection node in the monitoring area; Select m _p sampling sensors whose monitoring signal strength is greater than the preset threshold ρ from n sampling sensors to form the first batch of sampling initiation nodes; From n sampling sensors, select log(q) sampling sensors for each sampling sensor in the first batch of sampling initiating nodes to form the second batch of sampling initiating nodes; Taking each sampling initiating node in the first batch of sampling initiating nodes and the second batch of sampling initiating nodes as a starting point and taking the data collection node as an end point, select between each starting point and end point among n sampling sensors H sampling sensors form a corresponding sampling transmission path between each starting point and end point, wherein, H=O(logq); Each sampling initiation node sends the sampling data of the detected abnormal event to the adjacent sampling sensor on the corresponding sampling transmission path, and each sampling sensor on each sampling transmission path compares the sampling data of the abnormal event it monitors with the sampling data from the sampling sensor. The sampling data of the abnormal event received by the initiating node or the previous adjacent sampling sensor on the sampling transmission path is weighted and formed into new sampling data of the abnormal event, and then sent to the next adjacent sampling sensor or the data on the sampling transmission path collection node; The data collection node calculates the position and intensity of the abnormal event in the monitoring area according to the received sampling data of the new abnormal event; Wherein, the data collection node uses compressed sensing technology to provide the location and intensity of the abnormal event in the monitoring area. 2. The monitoring and location method of abnormal events based on compressed sensing as claimed in claim 1, wherein the calculation formula of the threshold ρ is as follows: $\mathrm{<span\; class="notranslate">\; \&rho;</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; \&pi;</span>}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}}<span\; class="notranslate">\; -</span>\sqrt{\mathrm{<span\; class="notranslate">\; 2</span>}}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}}{\mathrm{<span\; class="notranslate">\; erf</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0.5</span>}<span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; \&delta;</span>}}_{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>$ Among them, n is the number of sampling sensors distributed in the monitoring area, s is the area of the monitoring area, λ is the parameter value of an abnormal event, μ _λ is the average value of the parameter values of all abnormal events, and σ _λ is the abnormal event The standard deviation of the parameter value of , erf is the Gaussian error function, erf defines a cumulative probability distribution function of a normally distributed random variable, and the parameter value λ of each abnormal event conforms to the following normal distribution