A multi-sensor data fusion method and system

Abstract

The present invention relates to the technical field of gas data processing, and particularly relates to a multi-sensor data fusion method and system. The method includes the steps of: respectively taking each data point in the time series sequence of environmental gas monitoring data as an initial clustering cluster; obtaining the local density of a data point through the reciprocal of the average Euclidean distance between the data point and each data point within its preset neighborhood; calculating the necessity of fusion between the initial clustering clusters; if the maximum value of the necessity of fusion between an initial clustering cluster and other initial clustering clusters is greater than a preset threshold, then merging the initial clustering cluster with the initial clustering cluster corresponding to the maximum value of the necessity of fusion to obtain a new initial clustering cluster, and repeating this process to obtain multiple target clustering clusters; obtaining the degree of abnormality of the target clustering cluster based on the data point distribution in the target clustering cluster, and obtaining the environmental gas monitoring result, effectively improving the accuracy of the environmental gas monitoring result.

Description

Technical Field

[0001] The present invention relates to the technical field of gas data processing, and in particular, to a multi-sensor data fusion method and system. Background Art

[0002] Gas leakage may not only lead to production accidents, but also cause serious harm to environmental safety and personnel safety. Therefore, gas leakage monitoring technology can be applied to fields such as petroleum, chemical industry, and electric power. By collecting environmental gas data, the monitoring and early warning of gas leakage can be realized.

[0003] Currently, gas detection technologies can be roughly divided into two categories: single-sensor detection and multi-sensor fusion detection. Among them, for single-sensor detection, such as electrochemical sensors, infrared sensors, etc., these sensors can only detect specific gas species, such as ammonia, methane, etc. For multi-sensor detection, thresholds corresponding to the sensor types are usually set for the data collected by multiple sensors respectively, and an alarm is given if the threshold is exceeded.

[0004] However, during the process of collecting gas data through sensors, the change in temperature may affect the sensitivity and response speed of the sensors; the change in humidity may cause the drift of the sensor measurement values. If, when conducting gas monitoring, the influence of these environmental factors on gas sensor data is not considered, and the gas concentration is simply judged whether it exceeds the standard according to the preset threshold, false alarms or missed alarms may occur, posing a safety hazard.

[0005] Based on this, how to fuse the gas-related parameters after collecting the gas-related parameters in the environment through various sensors, so as to accurately obtain the gas leakage monitoring result in the environment, is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0006] In order to solve the technical problem of how to fuse the gas-related parameters after collecting the gas-related parameters in the environment through various sensors, so as to accurately obtain the gas leakage monitoring result in the environment, the present invention provides a multi-sensor data fusion method and system.

[0007] In a first aspect, the present invention provides a multi-sensor data fusion method, adopting the following technical solution:

[0008] A multi-sensor data fusion method includes the steps of:

[0009] Regarding each data point in the time series of environmental gas monitoring data as an initial clustering cluster; obtaining the local density of the data point through the reciprocal of the average Euclidean distance between the data point and each data point within its preset neighborhood; calculating the necessity of fusion between the initial clustering clusters:

[0010] ;

[0011] For the necessity of fusion between the -th and the -th initial clustering clusters, , are respectively the local density means of all data points in the -th and the -th initial clustering clusters, is the Euclidean distance between the clustering centers of the -th and the -th initial clustering clusters, is the exponential function with base e, is the absolute value symbol; if the maximum value of the necessity of fusion between an initial clustering cluster and other initial clustering clusters is greater than a preset threshold, then merge this initial clustering cluster with the initial clustering cluster corresponding to the maximum value of the necessity of fusion to obtain a new initial clustering cluster, and cycle in this way to obtain multiple target clustering clusters; obtain the abnormality degree of the target clustering clusters through the data point distribution in the target clustering clusters, and obtain the environmental gas monitoring result.

[0012] By dividing the time series of environmental gas monitoring data into multiple target clustering clusters for anomaly recognition, the present invention can accurately fuse the parameter data of each sensor, and accurately obtain the abnormal clustering clusters based on the fusion result, effectively improving the accuracy of the gas detection data monitoring result. In this process, by analyzing the differences in local density characteristics between the initial clustering clusters, the present invention can accurately obtain the necessity of fusion between the initial clustering clusters, and merge based on the necessity of fusion between the initial clustering clusters to obtain multiple target clustering clusters, reducing the possibility that abnormal data is fused into normal clustering clusters during the clustering process, thereby effectively improving the accuracy of the gas detection data monitoring result.

[0013] According to a multi-sensor data fusion method provided by the present invention, before taking each data point in the time series of environmental gas monitoring data as an initial clustering cluster, it further includes: collecting gas-related parameter data in the environment through a sensor at each acquisition moment and preprocessing it as a data point to obtain the time series of environmental gas monitoring data; wherein, the dimensions of the gas-related parameter data include gas concentration, environmental temperature, and humidity.

[0014] The present invention takes into account that there may be data missing, etc. in the originally collected gas-related parameter data in the environment, so the overall quality of the data is improved through preprocessing to facilitate subsequent data processing.

[0015] A multi-sensor data fusion method provided by the present invention, a method for obtaining a preset neighborhood of data points, includes: arranging the Euclidean distances between a data point and surrounding data points in ascending order; obtaining a preset number of surrounding data points in the ascending order as the preset neighborhood of the data point.

[0016] A multi-sensor data fusion method provided by the present invention, before calculating the necessity of fusion between initial clustering clusters, further includes: screening target clustering clusters from the initial clustering clusters according to the mergibility of the initial clustering clusters.

[0017] A multi-sensor data fusion method provided by the present invention, screening target clustering clusters from the initial clustering clusters according to the mergibility of the initial clustering clusters, includes: taking the initial clustering clusters with mergibility not greater than a preset merging threshold as the target clustering clusters; wherein, the mergibility of the initial clustering clusters satisfies the relational expression:

[0018] ;

[0019] is the mergibility of the th initial clustering cluster, is the number of dimensions of the data points in the th initial clustering cluster, is the th data point number in the th dimension in the th initial clustering cluster, is the th dimension in the th data point value in the th initial clustering cluster, is the mean value of the data points in the th dimension in the is the exponential function with base e, is the absolute value symbol.

[0020] The present invention provides an accurate calculation method for the mergibility of initial clustering clusters. By analyzing the differences between the data points in the initial clustering clusters and the data in each dimension, the mergibility of the initial clustering clusters can be accurately obtained. The smaller the difference, the higher the mergibility of the initial clustering clusters. Thus, high-quality initial clustering clusters can be accurately screened based on this, improving the accuracy of gas data monitoring while effectively reducing the data processing volume.

[0021] A multi-sensor data fusion method provided by the present invention, if the maximum value of the necessity for fusion between an initial clustering cluster and other initial clustering clusters is greater than a preset threshold, then merge this initial clustering cluster with the initial clustering cluster corresponding to the maximum value of the necessity for fusion. It further includes: if the maximum value of the necessity for fusion between an initial clustering cluster and other initial clustering clusters is not greater than the preset threshold, then take this initial clustering cluster as the target clustering cluster.

[0022] A multi-sensor data fusion method provided by the present invention, the degree of abnormality of the target clustering cluster satisfies the relational expression:

[0023] ;

[0024] is the degree of abnormality of the th target clustering cluster, is the number of data points in the th target clustering cluster, is the th data point in the th target clustering cluster and the distance between the th target clustering cluster's clustering center, is the exponential function with base e.

[0025] The present invention provides an accurate calculation method for the degree of abnormality of the target clustering cluster. By analyzing the distribution of data points in the target clustering cluster, the degree of abnormality of each target clustering cluster can be accurately obtained.

[0026] A multi-sensor data fusion method provided by the present invention, obtaining the degree of abnormality of the target clustering cluster through the distribution of data points in the target clustering cluster, and obtaining the environmental gas monitoring result, including: if the degree of abnormality of the target clustering cluster is greater than the preset abnormality threshold, then the monitoring results of each data point in the target clustering cluster are abnormal; otherwise, the monitoring results of each data point in the target clustering cluster are normal.

[0027] A multi-sensor data fusion method provided by the present invention, after obtaining the environmental gas monitoring result, it further includes: in response to the monitoring results of each data point in the target clustering cluster being abnormal, sending out an abnormal prompt for environmental gas leakage.

[0028] The present invention takes into account that when there is a leakage of environmental gas, it may cause serious harm to environmental safety and personnel safety. Therefore, by sending out an abnormal prompt, it is convenient for the staff to handle it in time and reduce potential safety hazards.

[0029] In a second aspect, the present invention provides a multi-sensor data fusion system, adopting the following technical solution:

[0030] A multi-sensor data fusion system includes: a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the above multi-sensor data fusion method.

[0031] By adopting the above technical solution, a computer program is generated for the above multi-sensor data fusion method and stored in the memory to be loaded and executed by the processor, so as to manufacture a terminal device according to the memory and the processor, which is convenient to use.

[0032] The present invention has the following technical effects:

[0033] Based on the above technical solution, for a multi-sensor data fusion method and system provided by the present invention, when monitoring whether there is gas leakage in the environment, by dividing the time series sequence of environmental gas monitoring data into multiple target clustering clusters and then performing anomaly identification, the parameter data of each sensor can be accurately fused, so as to accurately obtain the abnormal clustering clusters, effectively improving the accuracy of the gas detection data monitoring result. In this process, by analyzing the local density feature differences between the initial clustering clusters, the necessity for fusion between the initial clustering clusters can be accurately obtained, and based on the necessity for fusion between the initial clustering clusters, merging is performed to obtain multiple target clustering clusters, reducing the possibility that abnormal data is fused into normal clustering clusters during the clustering process, thereby effectively improving the accuracy of the gas detection data monitoring result. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a flowchart in a multi-sensor data fusion method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0035] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention.

[0036] In order to fuse the gas-related parameters after collecting the gas-related parameters in the environment through various sensors, so as to accurately obtain the gas leakage monitoring result in the environment, an embodiment of the present invention discloses a multi-sensor data fusion method, which can accurately obtain the gas leakage monitoring result in the environment by integrating and clustering various sensor data.

[0037] Specifically, please refer to Figure 1 as shown in Figure 1 It is a flowchart in a multi-sensor data fusion method provided by an embodiment of the present invention. The method specifically includes the following steps:

[0038] S1: Obtain each data point in the time series of environmental gas monitoring data collected by multiple sensors.

[0039] It should be noted that when monitoring whether there is gas leakage, there is a close relationship among gas concentration, environmental temperature, and humidity. For example, when monitoring gas, an infrared sensor is used for monitoring, and the light source spectrum, detector responsivity, and filter parameters of the infrared sensor will change with the change of temperature; an increase in temperature may cause the emission spectrum of the light source to shift or the detector sensitivity to decrease, resulting in a deviation in the detection of the characteristic absorption peak of the target gas by the sensor and affecting the accuracy of the gas leakage monitoring result. The humidity-sensitive element in the gas sensor is very sensitive to humidity changes. If the humidity is too high or too low, it may change the surface state of the humidity-sensitive element, thereby affecting its adsorption and desorption ability to the target gas, resulting in a drift of the measured value and ultimately affecting the accuracy of the gas leakage monitoring result.

[0040] Based on this, in the embodiment of the present invention, when monitoring gas concentration, it is necessary to comprehensively consider the data in three dimensions: concentration, temperature, and humidity, so as to accurately obtain the abnormal monitoring result of gas leakage.

[0041] Exemplarily, in the embodiment of the present invention, each data point in the time series of environmental gas monitoring data is respectively used as an initial clustering cluster. It also includes: collecting gas-related parameter data in the environment through a sensor at each acquisition moment and preprocessing it as a data point to obtain a time series of environmental gas monitoring data; among them, the dimensions of the gas-related parameter data include gas concentration, environmental temperature, and humidity.

[0042] Among them, the preprocessing can be data denoising, missing data interpolation, data format conversion, etc., which can be specifically set according to actual needs, and the embodiment of the present invention does not limit it too much here.

[0043] Specifically, through an infrared sensor, non-contact detection is carried out by using the absorption characteristics of gas to infrared light of a specific wavelength to collect gas concentration data in the environment; temperature data is collected by using a temperature sensor; humidity data is collected by using a humidity sensor. The concentration data, temperature data, and humidity data collected at the same moment are used as a data point, and the gas-related parameter data of each data point in the finally obtained time series of environmental gas monitoring data correspond one by one.

[0044] Among them, the acquisition frequency and acquisition duration of the sensor can be specifically set according to actual needs.

[0045] After collecting the time series of environmental gas monitoring data based on the above steps, continue to execute the following steps.

[0046] S2: Each data point in the time series of environmental gas monitoring data is used as an initial clustering cluster; the local density of the data point is obtained by the reciprocal of the average Euclidean distance between the data point and each data point within its preset neighborhood.

[0047] It should be noted that for environmental gas monitoring data, under normal circumstances, most data points will be in a relatively stable state and distribution, that is, most data points are relatively close to each other; when the environment changes, the data will shift as a whole and anomalies will occur, and the distance between the abnormal data and most normal data points is relatively far.

[0048] Based on this, in the embodiment of the present invention, the data distribution within the neighborhood of the data point can be characterized by calculating the local density of the data point. The greater the local density, the more concentrated the distribution of the data points around the data point.

[0049] Exemplarily, in the embodiment of the present invention, when obtaining the preset neighborhood of the data point, the Euclidean distances between the data point and the surrounding data points can be sorted in ascending order; the surrounding data points are obtained in the ascending order.

[0050] Obtaining the surrounding data points in the ascending order specifically includes the following two possible implementation manners:

[0051] In one possible implementation manner, a preset number of surrounding data points can be obtained in the ascending order as the preset neighborhood of the data point.

[0052] Among them, the preset number can be set to 50; the number can be specifically set according to actual needs.

[0053] In another possible implementation manner, a distance threshold can be preset; the surrounding data points greater than the distance threshold are used as the data points within the preset neighborhood of the data point.

[0054] Among them, the distance threshold can be specifically set according to actual needs, and the embodiment of the present invention does not limit it too much here.

[0055] After obtaining the preset neighborhood of each data point based on the above manner, the local density of the data point can be obtained based on the data distribution within the preset neighborhood of the data point.

[0056] Exemplarily, in the embodiment of the present invention, to calculate the local density of the data point, the following relational expression can be specifically referred to:

[0057] ;

[0058] is the local density of the i-th data point, is the number of data points within the neighborhood of the i-th data point, is the Euclidean distance between the i-th data point and the j-th data point within its preset neighborhood.

[0059] In the above formula, represents the average value of the Euclidean distances between the i-th data point and each data point within its preset neighborhood. The smaller this value is, the closer the Euclidean distances between the i-th data point and each data point within its preset neighborhood are, the denser the data points within the preset neighborhood of the i-th data point are, and the greater the corresponding local density is.

[0060] After obtaining the local densities of each data point based on the above steps, the data points can be clustered, and abnormal clustering clusters can be obtained based on the clustering results. During the clustering process, in order to avoid classifying abnormal data points with relatively small local densities into the nearest normal clustering cluster, in the embodiments of the present invention, each data point is used as an independent initial clustering cluster. By calculating the similarity degree and distance of the data distribution between the initial clustering clusters, the necessity of fusion between the initial clustering clusters and other initial clustering clusters is obtained, and the initial clustering clusters with a greater necessity of fusion are merged, so as to accurately implement the clustering analysis of environmental gas data.

[0061] S3: Calculate the necessity of fusion between the initial clustering clusters. If the maximum value of the necessity of fusion between an initial clustering cluster and other initial clustering clusters is greater than a preset threshold, then merge this initial clustering cluster with the initial clustering cluster corresponding to the maximum value of the necessity of fusion to obtain a new initial clustering cluster, and cycle in this way to obtain multiple target clustering clusters.

[0062] Among them, the preset threshold can be set to 0.8; the preset threshold can be specifically set according to actual needs, and the embodiments of the present invention do not limit this too much here.

[0063] It should be noted that for any initial clustering cluster, if the differences between the data points in the initial clustering cluster are too large, it means that the similarity between the data points within the clustering cluster is relatively low, and continuing to cluster it will transmit the lower similarity features to the clustering cluster obtained after merging, resulting in a relatively poor final clustering effect.

[0064] Based on this, before calculating the necessity of fusion between the initial clustering clusters in the embodiments of the present invention, the mergability of the initial clustering clusters can also be calculated first, so as to screen out the initial clustering clusters with better quality and reduce the data processing volume.

[0065] Exemplarily, in the embodiments of the present invention, before calculating the necessity of fusion between the initial clustering clusters, it further includes: screening target clustering clusters from the initial clustering clusters according to the mergability of the initial clustering clusters, and taking the initial clustering clusters with mergability not greater than a preset merging threshold as the target clustering clusters.

[0066] Among them, the preset merging threshold can be set to 0.7; the preset merging threshold can be specifically set according to actual needs.

[0067] It is understandable that the target clustering cluster is the initial clustering cluster that finally completes clustering and cannot be merged with other clustering clusters anymore.

[0068] Exemplarily, in the embodiments of the present invention, to calculate the mergability of the initial clustering clusters, the following relational expression can be specifically referred to:

[0069] ;

[0070] is the mergability of the th initial clustering cluster, is the number of dimensions of the data points in the th initial clustering cluster, is the number of data points in the th initial clustering cluster in the th dimension, is the value of the th data point in the th dimension in the th initial clustering cluster, is the mean value of the data points in the th initial clustering cluster in the th dimension, is the exponential function with base e, is the absolute value symbol.

[0071] In the above formula, represents the difference between the data values of each dimension in the th initial clustering cluster in the th dimension and the mean value of the data points in the th initial clustering cluster in the th dimension. The larger this value is, the greater the data difference in the th initial clustering cluster in the th dimension, the lower the similarity, and the weaker the mergable feature of the th initial clustering cluster in the th dimension, and the lower the mergability of the corresponding th initial clustering cluster.

[0072] is used to reduce the influence of data in different dimensions on the calculation result of mergability.

[0073] is a hyperparameter used to avoid the denominator of the formula being 0 and can be specifically set according to actual needs. The embodiments of the present invention do not impose too many restrictions here.

[0074] It can be understood that if the mergability of the initial clustering clusters is not greater than the preset merging threshold, it indicates that the data within the initial clustering clusters varies greatly, and the similarity between the data point dimension data is relatively low. By directly using the initial clustering clusters with relatively low mergability as the target clustering clusters without participating in the calculation of the necessity for fusion, the quality of the initial clustering clusters for calculating the necessity for fusion can be effectively improved, and the data processing volume can be reduced.

[0075] After analyzing the data characteristics within the initial clustering clusters based on the above formula to obtain the mergability of each initial clustering cluster, the initial clustering clusters with relatively low mergability can be screened out and directly used as the final clustering clusters, and the necessity for fusion between the initial clustering clusters with relatively high mergability and other initial clustering clusters can be calculated.

[0076] Exemplarily, in the embodiments of the present invention, the necessity for fusion between the initial clustering clusters can be calculated specifically with reference to the following relational expression:

[0077] ;

[0078] is the necessity for fusion between the th and the th initial clustering clusters, is the average local density of all data points in the th initial clustering cluster, is the average local density of all data points in the th initial clustering cluster, is the Euclidean distance between the clustering centers of the th and the th initial clustering clusters, is the exponential function with e as the base, is the absolute value symbol.

[0079] Among them, when obtaining the clustering center of the initial clustering cluster, the average value of the data points in the initial clustering cluster in one dimension can be used as the value of the clustering center in that dimension. The specific steps for obtaining the clustering center of the clustering cluster can be implemented through the prior art, and the embodiments of the present invention will not elaborate here.

[0080] In the above formula, represents the difference in the average local density between the th initial clustering cluster and the th initial clustering cluster. The larger this value is, the lower the similarity of the data distribution between the th initial clustering cluster and the th initial clustering cluster, the greater the difference, and the lower the necessity for fusion between the corresponding th initial clustering cluster and the th initial clustering cluster.

[0081] is the Euclidean distance between the -th initial clustering cluster and the clustering center of the -th initial clustering cluster. The smaller this value is, the closer the distance between the -th initial clustering cluster and the -th initial clustering cluster is. The higher the possibility that these two initial clustering clusters can be merged, the greater the corresponding necessity for merging.

[0082] In summary, if the difference in the local density means between an initial clustering cluster and other initial clustering clusters except itself is smaller, and the distance between the clustering centers is closer, then the necessity for merging between the initial clustering cluster and the corresponding other clustering clusters is greater, and the possibility that the two can be merged is also higher.

[0083] When merging after obtaining the necessity for merging between initial clustering clusters based on the above formula, in order to avoid errors caused by the same initial clustering cluster being merged with multiple other initial clustering clusters simultaneously during the merging process, after obtaining the necessity for merging between an initial clustering cluster and other initial clustering clusters based on the above steps, the maximum value of the necessity for merging between the initial clustering cluster and other initial clustering clusters can be obtained, and based on the comparison result between the maximum value of the necessity for merging and a preset threshold, it can be determined whether the initial clustering cluster needs to be merged with other clustering clusters.

[0084] Exemplarily, in the embodiments of the present invention, based on the comparison result between the maximum value of the necessity for merging and a preset threshold, determining whether the initial clustering cluster needs to be merged with other clustering clusters includes: if the maximum value of the necessity for merging between the initial clustering cluster and other initial clustering clusters is greater than the preset threshold, then merge the initial clustering cluster with the initial clustering cluster corresponding to the maximum value of the necessity for merging to obtain a new initial clustering cluster; if the maximum value of the necessity for merging between the initial clustering cluster and other initial clustering clusters is not greater than the preset threshold, then use the initial clustering cluster as the target clustering cluster.

[0085] After obtaining the new initial clustering cluster, the above steps S2 - S3 can be continuously looped to process the new initial clustering cluster until all the initial clustering clusters are target clustering clusters. Finally, the time series of environmental gas monitoring data is divided into multiple target clustering clusters, and the abnormal clustering clusters in the target clustering clusters are screened, that is, the following steps are continued to be executed.

[0086] S4: Obtain the degree of abnormality of the target clustering cluster based on the distribution of data points in the target clustering cluster to obtain the environmental gas monitoring result.

[0087] It should be noted that based on the above steps, each data point in the time series of environmental gas monitoring data can be divided into the corresponding target clustering cluster. The fewer the number of data points in the target clustering cluster, the more discrete the data distribution in the target clustering cluster, and the higher the corresponding degree of abnormality.

[0088] Based on this, the embodiments of the present invention can accurately obtain the degree of abnormality of each target clustering cluster by obtaining the data point distribution in the target clustering cluster.

[0089] It can be understood that there is a unique clustering center in each target clustering cluster. Therefore, each data point in the target clustering cluster has a unique corresponding clustering center.

[0090] Exemplarily, in the embodiments of the present invention, to determine the degree of abnormality of the target clustering cluster, the following relational expression can be specifically referred to:

[0091] ;

[0092] is the degree of abnormality of the th target clustering cluster, is the number of data points in the th target clustering cluster, is the th data point in the th target clustering cluster and the distance between the clustering center of the th target clustering cluster, is the exponential function with base e.

[0093] Among them, the distance between the data point in the target clustering cluster and the clustering center of the data points in other target clustering clusters can be the Euclidean distance.

[0094] In the above formula, is the average distance between the data points in the th target clustering cluster and the clustering center of the th target clustering cluster. The smaller this value is, the closer the distance between the data points in the th target clustering cluster and the clustering center of the th target clustering cluster, and the lower the degree of abnormality.

[0095] In summary, if the number of data points in the th target clustering cluster is larger and the distance between each data point and the corresponding clustering center is closer, then the degree of abnormality of the th target clustering cluster is lower.

[0096] It can be understood that the higher the degree of abnormality of the target clustering cluster, the higher the possibility that the target clustering cluster is an abnormal clustering cluster. Based on this, the abnormal clustering clusters in the target clustering cluster can be accurately obtained.

[0097] Exemplarily, in an embodiment of the present invention, the environmental gas monitoring result is obtained through the abnormality degree of the target clustering cluster, including: obtaining the environmental gas monitoring result through the comparison result between the abnormality degree of the target clustering cluster and a preset abnormality threshold.

[0098] Among them, the preset abnormality threshold can be set to 0.85; the preset abnormality threshold can be specifically set according to actual needs, and this is not elaborated in the embodiments of the present invention.

[0099] Specifically, if the abnormality degree of the target clustering cluster is greater than the preset abnormality threshold, the monitoring results of each data point in the target clustering cluster are abnormal; otherwise, the monitoring results of each data point in the target clustering cluster are normal.

[0100] It can be understood that if the monitoring result of a data point is abnormal, it indicates that there is an abnormality in the gas data corresponding to the acquisition moment of this data point, and gas leakage may occur. In order to reduce potential safety hazards, an abnormality prompt can be given after abnormal gas data is generated.

[0101] Exemplarily, in an embodiment of the present invention, after obtaining the environmental gas monitoring result, it further includes: in response to the monitoring results of each data point in the target clustering cluster being abnormal, sending out an abnormal prompt for environmental gas leakage.

[0102] Among them, the abnormal prompt can be data highlighting, automatically establishing an abnormal database for abnormal data, etc., which can be specifically set according to actual needs, and this is not overly limited in the embodiments of the present invention.

[0103] It can be seen that in an embodiment of the present invention, when monitoring whether there is gas leakage in the environment, each data point in the time series of environmental gas monitoring data can be used as an initial clustering cluster respectively; through the reciprocal of the average Euclidean distance between a data point and each data point within its preset neighborhood, the local density of the data point is obtained; calculate the necessity of fusion between the initial clustering clusters:

[0104] ;

[0105] is the necessity of fusion between the th and the th initial clustering clusters, 、 are respectively the average local densities of all data points in the th and the th initial clustering clusters, is the Euclidean distance between the clustering centers of the th and the th initial clustering clusters, is the exponential function with e as the base, is the absolute value symbol; if the maximum value of the necessity for fusion between the initial clustering cluster and other initial clustering clusters is greater than the preset threshold, then merge this initial clustering cluster with the initial clustering cluster corresponding to the maximum value of the necessity for fusion to obtain a new initial clustering cluster, and so on in a loop to obtain multiple target clustering clusters; obtain the degree of abnormality of the target clustering cluster based on the distribution of data points in the target clustering cluster, and obtain the environmental gas monitoring result, effectively improving the accuracy of the environmental gas monitoring result.

[0106] An embodiment of the present invention also discloses a multi-sensor data fusion system, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, a multi-sensor data fusion method provided by the present invention is implemented.

[0107] The above system also includes other components well-known to those skilled in the art such as a communication bus and a communication interface, and their settings and functions are known in the art, so they will not be described in detail here.

[0108] In the present invention, the aforementioned memory can be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, apparatus, or device.

[0109] The above are all preferred embodiments of the present invention, and the protection scope of the present invention is not limited accordingly. Therefore, all equivalent changes made according to the structure, shape, and principle of the present invention should be covered within the protection scope of the present invention.

Claims

1. A multi-sensor data fusion method, characterized in that: include: Each data point in the time series of environmental gas monitoring data is taken as an initial cluster; The local density of the data point is obtained by taking the inverse of the mean of the Euclidean distance between the data point and each data point in its preset neighborhood; the necessity of fusion between the initial clusters is calculated: ； For the and The necessity of fusion between the initial clusters, , Respectively , The local density mean of all data points in the initial clusters, For the and The Euclidean distance between the initial cluster centers, is an exponential function with base e, is the absolute value symbol; By analyzing the differences in local density features between the initial clusters, the necessity of fusion between the initial clusters can be accurately obtained. If the maximum fusion necessity between the initial cluster and other initial clusters is greater than a preset threshold, the initial cluster is merged with the initial cluster corresponding to the maximum fusion necessity to obtain a new initial cluster. This cycle is repeated to obtain multiple target clusters. The abnormality degree of the target cluster is obtained through the distribution of data points in the target cluster, and the environmental gas monitoring result is obtained; The abnormality degree of the target cluster satisfies the relationship: ； For the The abnormality of the target clusters, For the The number of data points in the target cluster, For the The first target cluster Data points and The distance between the cluster centers of the target clusters, is an exponential function with base e.

2. A multi-sensor data fusion method according to claim 1, characterized in that: The method of taking each data point in the time series of the environmental gas monitoring data as an initial cluster also includes: At each collection moment, the gas-related parameter data in the environment is collected by the sensor and preprocessed as a data point to obtain a time series sequence of environmental gas monitoring data; wherein the dimensions of the gas-related parameter data include gas concentration, ambient temperature, and humidity.

3. A multi-sensor data fusion method according to claim 1, characterized in that: The preset neighborhood acquisition method of the data point includes: Arrange the Euclidean distances between the data point and surrounding data points in ascending order; obtain a preset number of surrounding data points in ascending order as a preset neighborhood of the data point.

4. The multi-sensor data fusion method according to claim 1, characterized in that: The calculation of the necessity of fusion between the initial clustering clusters also includes: The target cluster is selected from the initial clusters according to the mergibility of the initial clusters.

5. A multi-sensor data fusion method according to claim 4, characterized in that: The step of selecting a target cluster from the initial clusters according to the mergibility of the initial clusters comprises: The initial cluster whose mergibility is not greater than the preset merging threshold is used as the target cluster; Among them, the mergeability of the initial clusters satisfies the relationship: ； For the The mergibility of the initial clusters. For the The number of dimensions of the data points in the initial clusters, For the The first The number of data points in the dimension, For the The first The dimension The value of the data point, For the The first The mean of the data points in the dimension, is an exponential function with base e, is the absolute value symbol.

6. The multi-sensor data fusion method according to claim 1, characterized in that: If the maximum value of the fusion necessity between the initial clustering cluster and other initial clustering clusters is greater than a preset threshold, the initial clustering cluster is merged with the initial clustering cluster corresponding to the maximum value of the fusion necessity, and further includes: If the maximum value of the fusion necessity between the initial clustering cluster and other initial clustering clusters is not greater than a preset threshold, the initial clustering cluster is used as the target clustering cluster.

7. The multi-sensor data fusion method according to claim 1, characterized in that: The abnormality degree of the target cluster is obtained by the distribution of data points in the target cluster, and the environmental gas monitoring result is obtained, including: If the abnormality degree of the target cluster is greater than the preset abnormality threshold, the monitoring result of each data point in the target cluster is abnormal; otherwise, the monitoring result of each data point in the target cluster is normal.

8. The multi-sensor data fusion method according to claim 1, characterized in that: The environmental gas monitoring result is obtained, and then further includes: In response to the monitoring results of each data point in the target cluster being abnormal, an abnormal environmental gas leakage prompt is issued externally.

9. A multi-sensor data fusion system, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a multi-sensor data fusion method according to any one of claims 1 to 8 is implemented.

Images