CN116990883A - Remote person-carried dangerous object detection system based on multi-spectrum sensing fusion technology - Google Patents
Remote person-carried dangerous object detection system based on multi-spectrum sensing fusion technology Download PDFInfo
- Publication number
- CN116990883A CN116990883A CN202311263036.4A CN202311263036A CN116990883A CN 116990883 A CN116990883 A CN 116990883A CN 202311263036 A CN202311263036 A CN 202311263036A CN 116990883 A CN116990883 A CN 116990883A
- Authority
- CN
- China
- Prior art keywords
- data
- module
- sensor
- alarm
- carried
- 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.)
- Granted
Links
- 230000004927 fusion Effects 0.000 title claims abstract description 41
- 238000001228 spectrum Methods 0.000 title claims abstract description 38
- 238000001514 detection method Methods 0.000 title claims abstract description 23
- 238000005516 engineering process Methods 0.000 title claims abstract description 16
- 238000007781 pre-processing Methods 0.000 claims abstract description 34
- 238000012544 monitoring process Methods 0.000 claims abstract description 23
- 238000005457 optimization Methods 0.000 claims abstract description 9
- 238000012545 processing Methods 0.000 claims abstract description 6
- 230000006870 function Effects 0.000 claims description 39
- 230000007613 environmental effect Effects 0.000 claims description 27
- 238000012706 support-vector machine Methods 0.000 claims description 24
- 238000004422 calculation algorithm Methods 0.000 claims description 18
- 239000002184 metal Substances 0.000 claims description 14
- 239000013598 vector Substances 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 12
- 238000000034 method Methods 0.000 claims description 10
- 230000007246 mechanism Effects 0.000 claims description 9
- 238000007635 classification algorithm Methods 0.000 claims description 7
- 238000013135 deep learning Methods 0.000 claims description 7
- 238000013136 deep learning model Methods 0.000 claims description 6
- 230000003044 adaptive effect Effects 0.000 claims description 4
- 230000004044 response Effects 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 claims description 3
- 238000004891 communication Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 238000005286 illumination Methods 0.000 claims description 3
- 238000007689 inspection Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 230000005855 radiation Effects 0.000 claims description 3
- 231100001267 hazard identification Toxicity 0.000 claims 1
- 239000000383 hazardous chemical Substances 0.000 claims 1
- 238000010606 normalization Methods 0.000 claims 1
- 230000001953 sensory effect Effects 0.000 claims 1
- 238000007405 data analysis Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000035945 sensitivity Effects 0.000 description 2
- 230000001960 triggered effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V11/00—Prospecting or detecting by methods combining techniques covered by two or more of main groups G01V1/00 - G01V9/00
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Geophysics (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention relates to the technical field of dangerous object detection, in particular to a remote person-carried dangerous object detection system based on a multi-frequency spectrum sensing fusion technology, which comprises a multi-frequency spectrum sensor module, a data preprocessing module, a self-adaptive adjustment module, a data fusion module, a dangerous object identification module and an alarm module, wherein the multi-frequency spectrum sensor module is used for: data for collecting a plurality of frequency spectrums in a predetermined monitoring area; and a data preprocessing module: receiving original data, performing denoising and standardization processing, and preparing for data fusion; and the self-adaptive adjustment module is used for: automatically adjusting acquisition parameters and strategies of the multi-spectrum sensor module; and a data fusion module: and carrying out data fusion to generate a comprehensive data characteristic set. According to the invention, through the multi-element data acquisition, the data preprocessing and the intelligent optimization of the self-adaptive adjustment module of the multi-spectrum sensor module, the high-accuracy remote detection of dangerous objects carried by people and the real-time alarm are realized in complex and changeable environments.
Description
Technical Field
The invention relates to the technical field of dangerous object detection, in particular to a remote person-carried dangerous object detection system based on a multi-spectrum sensing fusion technology.
Background
Remote person-carried dangerous object detection is one of the key links of modern society safety precautions, and traditional person-carried dangerous object detection methods mostly depend on a single or a few sensors, such as a metal detector, an infrared camera or an X-ray scanning device, and the methods generally require that a human body is in close contact with the device or needs to be checked in a specific place, which limits the application of the method in large-scale and open places (such as airports, stations, stadiums and the like).
In addition, the conventional method lacks self-adaptive capability in terms of data preprocessing, sensor setting adjustment, data analysis and the like, and particularly, since a single sensor can only acquire one type of data, the accuracy and sensitivity of identification are generally affected when facing complex environments and various types of dangerous objects, and in addition, most of the conventional systems lack self-adaptive adjustment mechanisms, real-time optimization of sensor parameters or strategies according to environmental changes or preamble processing results cannot be performed.
Data fusion and advanced data analysis are also often neglected links in the traditional method, and most systems adopt simple threshold judgment or basic classification algorithms, so that the system is easy to generate false alarm or missing report under a complex or noisy background.
Therefore, a new method for detecting dangerous objects carried by people efficiently and accurately in long-distance and open occasions is urgently needed.
Disclosure of Invention
Based on the purposes, the invention provides a remote person-carried dangerous object detection system based on a multi-frequency sensing fusion technology.
The remote person carrying dangerous object detection system based on the multi-frequency sensing fusion technology comprises a multi-frequency sensor module, a data preprocessing module, a self-adaptive adjustment module, a data fusion module, a dangerous object identification module and an alarm module, wherein,
a multi-spectral sensor module: data for collecting a plurality of frequency spectrums in a predetermined monitoring area;
and a data preprocessing module: receiving the original data from the multi-spectrum sensor module, performing denoising and standardization processing, and preparing for data fusion;
and the self-adaptive adjustment module is used for: the system is linked with the multi-frequency spectrum sensor module and the data preprocessing module, and the acquisition parameters and strategies of the multi-frequency spectrum sensor module are automatically adjusted according to the environmental factors and the output results of the preprocessing module so as to optimize the data quality and reduce the noise;
and a data fusion module: the multi-frequency spectrum data adjusted by the self-adaptive adjusting module and the standardized data output by the data preprocessing module are used as inputs by adopting a deep learning algorithm, and data fusion is carried out to generate a comprehensive data characteristic set;
dangerous object identification module: analyzing according to the generated comprehensive data characteristic set, and judging whether dangerous articles are carried or not by using a classification algorithm of a support vector machine;
and an alarm module: for alerting after the identification of dangerous objects carried.
Further, the multi-spectrum sensor module comprises an infrared sensor, a millimeter wave radar and a metal detector, in particular,
an infrared sensor: the infrared sensors are arranged at four corners of a preset monitoring area and are 1.5 meters away from the ground so as to cover the horizontal and vertical directions of the whole monitoring area, and the infrared sensors are used for collecting the heat radiation information of individuals in the area;
millimeter wave radar: the millimeter wave radar is arranged in the center of the monitoring area and is 2 meters away from the ground so that the scanning angle can cover the whole preset area, and the millimeter wave radar is used for acquiring the shape and speed information of a moving object in the area so as to identify nonmetallic dangerous goods hidden under clothes;
a metal detector: are distributed at the entrance and the exit of the monitoring area and are arranged below 20 cm of the ground for detecting whether an individual passing through the area carries metal objects.
Further, the denoising and standardization process specifically comprises:
denoising: the method comprises the steps of firstly, respectively carrying out Fourier transform on raw data from an infrared sensor, a millimeter wave radar and a metal detector, converting the raw data into a frequency domain, filtering environmental noise and non-target related frequency components by using a preset frequency threshold value in the frequency domain, and then carrying out inverse Fourier transform to recover to a time domain;
and (3) standardization treatment: the denoised data is normalized, and specifically, the data of each sensor is Z-score normalized, i.e., the mean value of the sensor data is subtracted from each data point and divided by the standard deviation, which ensures that data from different types of sensors and different magnitudes can be compared and fused under the same standard.
Further, the self-adaptive adjustment module is embedded with an environment sensor group, the sensor group comprises a temperature sensor, a humidity sensor and an illumination sensor, and the environment sensor group is used for receiving current environment data.
Furthermore, the self-adaptive adjustment module automatically adjusts the acquisition parameters and strategies of the multi-frequency sensor module by adopting a gradient descent optimization algorithm, and the method comprises the following specific steps:
s1: let the current environmental factors be:, wherein ,/>Is a vector containing n environmental factors, and sets metadata output by the preprocessing module as follows: /> wherein ,/>Is a kind of->A vector of individual metadata;
s2: based on the two types of data and />And running a gradient descent optimization algorithm, specifically, optimizing an objective function as follows:
，
wherein ,is an objective function, intended to be minimized;
is a vector containing all the sensor parameters to be optimized;
and />Is a weight factor for balancing the environmental factors +.>And preprocessing output +.>An effect in the objective function;
and />Is a mapping function for quantifying environmental factors and preprocessing output versus sensor parameters +.>A relationship between;
s3: by calculating gradientsUpdate sensor parameters +.>To minimize +.>The update rule is:
, wherein ,
and />Sensor parameter vectors after and before updating, respectively;
is the learning rate, is used for controlling the step length of parameter updating;
s4: when a new parameter is determinedThe adaptive adjustment module updates parameters of the multi-frequency sensor module via the control signal.
Further, the deep learning algorithm specifically performs data fusion on a multi-layer perceptron deep learning model, where the deep learning model specifically includes:
input layer: the layer receives two types of data as input, the first type is multi-frequency data adjusted by the self-adaptive adjustment module, the second type is standardized data output by the data preprocessing module, and the first type is preset as followsThe second type is；
Hidden layer: comprising a plurality of hidden layers, each hidden layer being non-linearly transformed using an activation function, said hidden layers being for capturing and />Features and modes of (a);
output layer: the layer outputs a comprehensive data feature setThe system is used for further analysis of the dangerous object identification module;
loss function: using cross-picking loss functionsTraining by a back propagation algorithm, wherein the specific loss function is as follows: />，
wherein ,
is the actual label;
probability of model prediction;
is the number of samples.
Further, the step of judging whether dangerous goods are carried by using a classification algorithm of the support vector machine specifically includes:
input characteristics: receiving the comprehensive data feature set output by the data fusion moduleAs input;
support vector machine model: the support vector machine classifier with radial basis function RBF kernel is adopted, and the specific decision function is expressed as follows:
，
wherein ,
is a decision function for determining a sample +.>Whether dangerous objects are carried;
and />Is a parameter of a support vector machine model; />Is->Labels of the individual training samples;
is->Training samples;
is an RBF kernel function for calculating samples +.> and />Similarity between;
is the number of training samples;
classification decision: when (when)When it is, determine sample->Carrying dangerous objects and triggering an alarm module; when (when)When it is, determine sample->Is not dangerous; model updating: the dangerous object recognition module also comprises an online learning mechanism, and periodically updates parameters of the support vector machine model by continuously collecting new sample data and corresponding labels> and />To accommodate new types of hazards or environmental changes.
Further, the alarm module includes a multi-level alarm mechanism that, in particular,
primary alarm: support vector machine decision function of dangerous object recognition moduleOutputting a positive value but below a predetermined level one alarm threshold +.>When the system triggers a first-level alarm, and the system informs an operator of further manual inspection through a yellow LED lamp of the young and a low-volume alarm age;
secondary alarm: when (when)Outputting a positive value and higher than or equal to +.>But below the secondary alarm threshold +.>When the system is used, the system can trigger a secondary alarm, an operator is notified through a continuous red LED lamp and a high-volume alarm bell, and the entrance and exit of the monitoring area are automatically locked until manual release is achieved.
Further, the alarm module further comprises an automatic notification unit, specifically, when the secondary alarm is triggered, the automatic notification unit sends relevant alarm information and dangerous object characteristic data to communication equipment of a pre-configured security personnel or emergency response team immediately, wherein the notification content comprises the current threat level, the specific position of the monitoring area and a dangerous object identification module support vector machine decision functionIs a function of the output value of (a).
The invention has the beneficial effects that:
according to the invention, through the design and implementation of the multi-spectrum sensor module, data of various spectrums can be acquired in a preset monitoring area, so that high-accuracy identification of various dangerous objects is realized, and the module adopts various sensors including but not limited to infrared sensors, millimeter wave radars and chemical sensors, and the sensors work together to provide data with more comprehensive and higher resolution than a single sensor, so that the system has higher accuracy and sensitivity when carrying out large-range and long-distance dangerous object detection on open occasions such as airports, stations and stadiums.
According to the invention, through the linkage of the data preprocessing module and the self-adaptive adjustment module, the acquisition parameters and strategies of the sensor can be automatically adjusted according to the environmental factors and the data quality, so that the risks of false alarm and missing alarm are reduced, the self-adaptive adjustment module can analyze the output of the preprocessing module and the current environmental factors in real time, and the setting of the multi-frequency spectrum sensor module is automatically adjusted according to the preset optimization algorithm, so that the mechanism greatly improves the adaptability and stability of the system in complex and changeable environments.
According to the invention, a plurality of frequency spectrum data and environment information are synthesized through a deep learning algorithm adopted by the data fusion module, a comprehensive data feature set is generated, the dangerous object identification module adopts a support vector machine SVM algorithm, rapid and accurate classification is carried out according to the generated feature set, when the system judges that dangerous objects exist, the alarm module is immediately started to carry out real-time early warning and automatic notification, and a safety person or an emergency response team is ensured to rapidly carry out proper countermeasures.
Drawings
In order to more clearly illustrate the invention or the technical solutions of the prior art, the drawings which are used in the description of the embodiments or the prior art will be briefly described, it being obvious that the drawings in the description below are only of the invention and that other drawings can be obtained from them without inventive effort for a person skilled in the art.
Fig. 1 is a schematic diagram of a remote person-carried dangerous object detection system according to an embodiment of the present invention.
Detailed Description
The present invention will be further described in detail with reference to specific embodiments in order to make the objects, technical solutions and advantages of the present invention more apparent.
It is to be noted that unless otherwise defined, technical or scientific terms used herein should be taken in a general sense as understood by one of ordinary skill in the art to which the present invention belongs. The terms "first," "second," and the like, as used herein, do not denote any order, quantity, or importance, but rather are used to distinguish one element from another. The word "comprising" or "comprises", and the like, means that elements or items preceding the word are included in the element or item listed after the word and equivalents thereof, but does not exclude other elements or items. The terms "connected" or "connected," and the like, are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "upper", "lower", "left", "right", etc. are used merely to indicate relative positional relationships, which may also be changed when the absolute position of the object to be described is changed.
As shown in fig. 1, the remote person-carried dangerous object detection system based on the multi-frequency sensing fusion technology comprises a multi-frequency sensor module, a data preprocessing module, a self-adaptive adjustment module, a data fusion module, a dangerous object identification module and an alarm module, wherein,
a multi-spectral sensor module: data for collecting a plurality of frequency spectrums in a predetermined monitoring area;
and a data preprocessing module: receiving the original data from the multi-spectrum sensor module, performing denoising and standardization processing, and preparing for data fusion;
and the self-adaptive adjustment module is used for: the system is linked with the multi-frequency spectrum sensor module and the data preprocessing module, and the acquisition parameters and strategies of the multi-frequency spectrum sensor module are automatically adjusted according to the environmental factors and the output results of the preprocessing module so as to optimize the data quality and reduce the noise;
and a data fusion module: the multi-frequency spectrum data adjusted by the self-adaptive adjusting module and the standardized data output by the data preprocessing module are used as inputs by adopting a deep learning algorithm, and data fusion is carried out to generate a comprehensive data characteristic set;
dangerous object identification module: analyzing according to the generated comprehensive data characteristic set, and judging whether dangerous articles are carried or not by using a classification algorithm of a support vector machine;
and an alarm module: for alerting after the identification of dangerous objects carried.
The multi-spectrum sensor module comprises an infrared sensor, a millimeter wave radar and a metal detector, in particular,
an infrared sensor: the infrared sensors are arranged at four corners of a preset monitoring area and are 1.5 meters away from the ground so as to cover the horizontal and vertical directions of the whole monitoring area, and the infrared sensors are used for collecting the heat radiation information of individuals in the area;
millimeter wave radar: the millimeter wave radar is arranged in the center of the monitoring area and is 2 meters away from the ground so that the scanning angle can cover the whole preset area, and the millimeter wave radar is used for acquiring the shape and speed information of a moving object in the area so as to identify nonmetallic dangerous goods hidden under clothes;
a metal detector: an entrance and an exit distributed in the monitoring area and installed 20 cm below the ground for detecting whether an individual passing through the area carries a metal object such as a gun or an explosive made of metal;
the three sensors acquire data through a time synchronization mechanism and transmit the data to the data preprocessing module in real time, and the layout and the operation mode of the multi-spectrum sensor module can ensure that the data acquisition is carried out on a preset monitoring area in an all-around, multi-angle and high-precision manner so as to reduce uncertainty and false alarm to the greatest extent.
The denoising and standardization process specifically comprises the following steps:
denoising: firstly, respectively carrying out Fourier transform (FFT) on the original data from the infrared sensor, the millimeter wave radar and the metal detector, converting the original data into a frequency domain, filtering out environmental noise and non-target related frequency components by utilizing a preset frequency threshold value in the frequency domain, and then carrying out inverse Fourier transform (IFFT) to recover to a time domain;
and (3) standardization treatment: the denoised data is subjected to standardization, specifically, the data of each sensor is subjected to Z-score standardization, namely, the average value of the sensor data is subtracted from each data point and then divided by the standard deviation, so that the data from different types of sensors and different magnitudes can be compared and fused under the same standard;
denoising and standardization processing are completed in real time or near real time so as to ensure that the whole system can react quickly; the parameter setting and execution logic of the two steps are correspondingly adjusted according to the output of the self-adaptive adjustment module, so that the data quality is ensured to always meet the requirements of the data fusion module and the dangerous object identification module.
The self-adaptive adjustment module is embedded with an environment sensor group, wherein the sensor group comprises a temperature sensor, a humidity sensor and an illumination sensor, and the environment sensor group is used for receiving current environment data.
The self-adaptive adjustment module automatically adjusts the acquisition parameters and strategies of the multi-spectrum sensor module by adopting a gradient descent optimization algorithm, and comprises the following specific steps:
s1: let the current environmental factors be:, wherein ,/>Is a vector containing n environmental factors, and sets metadata output by the preprocessing module as follows: /> wherein ,/>Is a kind of->A vector of individual metadata;
s2: based on the two types of data and />And running a gradient descent optimization algorithm, specifically, optimizing an objective function as follows:
, wherein ,
is an objective function, intended to be minimized;
is a vector containing all the sensor parameters to be optimized;
and />Is a weight factor for balancing the environmental factors +.>And preprocessing output +.>An effect in the objective function;
and />Is a mapping function for quantifying environmental factors and preprocessing output versus sensor parameters +.>A relationship between;
s3: by calculating gradientsUpdate sensor parameters +.>To minimize +.>The update rule is:
，
wherein , and />Sensor parameter vectors after and before updating, respectively;
is the learning rate, is used for controlling the step length of parameter updating;
s4: when a new parameter is determinedThe adaptive adjustment module updates parameters of the multi-frequency sensor module via the control signal.
The deep learning algorithm specifically performs data fusion on a multi-layer perceptron deep learning model, and the deep learning model specifically comprises the following components:
input layer: the layer receives two types of data as input, the first type is multi-frequency data adjusted by the self-adaptive adjustment module, the second type is standardized data output by the data preprocessing module, and the first type is preset as followsThe second type is；
Hidden layer: comprising a plurality of hidden layers, each hidden layer being non-linearly transformed using an activation function, said hidden layers being for capturing and />Features and modes of (a);
output layer: the layer outputs a comprehensive data feature setThe system is used for further analysis of the dangerous object identification module;
loss function: using cross-picking loss functionsTraining by a back propagation algorithm, wherein the specific loss function is as follows: />, wherein ,
is the actual tag (dangerous or non-dangerous); />Probability of model prediction; />Is the number of samples.
The step of judging whether dangerous goods are carried by using a classification algorithm of a support vector machine specifically comprises the following steps:
input characteristics: receiving the comprehensive data feature set output by the data fusion moduleAs input;
support vector machine model: the support vector machine classifier with radial basis function RBF kernel is adopted, and the specific decision function is expressed as follows:
, wherein ,
is a decision function for determining a sample +.>Whether dangerous objects are carried;
and />Is a parameter of a support vector machine model; />Is->Labels of the individual training samples;
is->Training samples; />Is an RBF kernel function for calculating samples +.> and />Similarity between;
is the number of training samples;
classification decision: when (when)When it is, determine sample->Carrying dangerous objects and triggering an alarm module; when (when)When it is, determine sample->Is not dangerous; model updating: the dangerous object recognition module also comprises an online learning mechanism, and periodically updates parameters of the support vector machine model by continuously collecting new sample data and corresponding labels> and />To accommodate new types of hazards or environmental changes.
The alarm module includes a multi-level alarm mechanism that, in particular,
primary alarm: support vector machine decision function of dangerous object recognition moduleOutputting a positive value but below a predetermined level one alarm threshold +.>When the system triggers a first-level alarm, and the system informs an operator of further manual inspection through a yellow LED lamp of the young and a low-volume alarm age;
secondary alarm: when (when)Outputting a positive value and higher than or equal to +.>But below the secondary alarm threshold +.>When the system is used, the system can trigger a secondary alarm, an operator is notified through a continuous red LED lamp and a high-volume alarm bell, and the entrance and exit of the monitoring area are automatically locked until manual release is achieved.
The alarm module further comprises an automatic notification unit, and specifically, when the secondary alarm is triggered, the automatic notification unit immediately sends related alarm information and dangerous object characteristic data to communication equipment of a pre-configured security personnel or emergency response team, wherein notification content comprises the current threat level, the specific position of a monitoring area and a dangerous object identification module support vector machine decision functionIs a function of the output value of (a).
The present invention is intended to embrace all such alternatives, modifications and variances which fall within the broad scope of the appended claims. Therefore, any omission, modification, equivalent replacement, improvement, etc. of the present invention should be included in the scope of the present invention.
Claims (9)
1. The remote person-carried dangerous object detection system based on the multi-frequency spectrum sensing fusion technology is characterized by comprising a multi-frequency spectrum sensor module, a data preprocessing module, a self-adaptive adjustment module, a data fusion module, a dangerous object identification module and an alarm module, wherein,
a multi-spectral sensor module: data for collecting a plurality of frequency spectrums in a predetermined monitoring area;
and a data preprocessing module: receiving the original data from the multi-spectrum sensor module, performing denoising and standardization processing, and preparing for data fusion;
and the self-adaptive adjustment module is used for: the system is linked with the multi-frequency spectrum sensor module and the data preprocessing module, and the acquisition parameters and strategies of the multi-frequency spectrum sensor module are automatically adjusted according to the environmental factors and the output results of the preprocessing module so as to optimize the data quality and reduce the noise;
and a data fusion module: the multi-frequency spectrum data adjusted by the self-adaptive adjusting module and the standardized data output by the data preprocessing module are used as inputs by adopting a deep learning algorithm, and data fusion is carried out to generate a comprehensive data characteristic set;
dangerous object identification module: analyzing according to the generated comprehensive data characteristic set, and judging whether dangerous articles are carried or not by using a classification algorithm of a support vector machine;
and an alarm module: for alerting after the identification of dangerous objects carried.
2. The remote person-carried hazardous materials detection system based on multi-spectral sensing fusion technology of claim 1, wherein the multi-spectral sensor module comprises an infrared sensor, a millimeter wave radar, and a metal detector, in particular,
an infrared sensor: the infrared sensors are arranged at four corners of a preset monitoring area and are 1.5 meters away from the ground so as to cover the horizontal and vertical directions of the whole monitoring area, and the infrared sensors are used for collecting the heat radiation information of individuals in the area;
millimeter wave radar: the millimeter wave radar is arranged in the center of the monitoring area and is 2 meters away from the ground so that the scanning angle can cover the whole preset area, and the millimeter wave radar is used for acquiring the shape and speed information of a moving object in the area so as to identify nonmetallic dangerous goods hidden under clothes;
a metal detector: are distributed at the entrance and the exit of the monitoring area and are arranged below 20 cm of the ground for detecting whether an individual passing through the area carries metal objects.
3. The remote person-carried hazard detection system based on multi-spectral sensing fusion technology of claim 1, wherein the denoising and normalization process is specifically:
denoising: the method comprises the steps of firstly, respectively carrying out Fourier transform on raw data from an infrared sensor, a millimeter wave radar and a metal detector, converting the raw data into a frequency domain, filtering environmental noise and non-target related frequency components by using a preset frequency threshold value in the frequency domain, and then carrying out inverse Fourier transform to recover to a time domain;
and (3) standardization treatment: the denoised data is normalized, and specifically, the data of each sensor is Z-score normalized, i.e., the mean value of the sensor data is subtracted from each data point and divided by the standard deviation, which ensures that data from different types of sensors and different magnitudes can be compared and fused under the same standard.
4. The remote person-carried hazard detection system based on multi-spectral sensing fusion technology of claim 3, wherein said adaptive adjustment module is embedded with an environmental sensor set comprising a temperature sensor, a humidity sensor and an illumination sensor, said environmental sensor set being configured to receive current environmental data.
5. The remote person-carried dangerous object detection system based on the multi-spectrum sensing fusion technology according to claim 4, wherein the self-adaptive adjustment module specifically adopts a gradient descent optimization algorithm to automatically adjust the acquisition parameters and strategies of the multi-spectrum sensor module, and the specific steps are as follows: s1: let the current environmental factors be:, wherein ,/>Is a vector containing n environmental factors, and sets metadata output by the preprocessing module as follows: /> wherein ,is a kind of->A vector of individual metadata;
s2: based on the two types of data and />And running a gradient descent optimization algorithm, specifically, optimizing an objective function as follows:，
wherein ,
is an objective function aimed at minimizing；
Is a vector containing all the sensor parameters to be optimized;
and />Is a weight factor for balancing the environmental factors +.>And preprocessing output +.>An effect in the objective function;
and />Is a mapping function for quantifying environmental factors and preprocessing output versus sensor parameters +.>A relationship between;
s3: by calculating gradientsUpdate sensor parameters +.>To minimize +.>The update rule is:
，
wherein ,
and />Sensor parameter vectors after and before updating, respectively;
is the learning rate, is used for controlling the step length of parameter updating;
s4: when a new parameter is determinedThe adaptive adjustment module updates parameters of the multi-frequency sensor module via the control signal.
6. The remote person-carried dangerous object detection system based on the multi-spectrum sensing fusion technology of claim 5, wherein the deep learning algorithm specifically performs data fusion for a multi-layer perceptron deep learning model, and the deep learning model specifically comprises:
input layer: the layer receives two types of data as input, the first type is multi-frequency data adjusted by the self-adaptive adjustment module, the second type is standardized data output by the data preprocessing module, and the first type is preset as followsThe second type is；
Hidden layer: comprising a plurality of hidden layers, each hidden layer being non-linearly transformed using an activation function, said hidden layers being for capturing and />Features and modes of (a);
output layer: the layer outputs a comprehensive data feature setThe system is used for further analysis of the dangerous object identification module;
loss function: using cross-picking loss functionsTraining by a back propagation algorithm, wherein the specific loss function is as follows:，
wherein ；
is the actual label;
probability of model prediction;
is the number of samples.
7. The remote person-carried dangerous object detection system based on multi-spectrum sensing fusion technology of claim 6, wherein said step of using a classification algorithm of a support vector machine to determine whether there is a carried dangerous object specifically comprises:
input characteristics: receiving the comprehensive data feature set output by the data fusion moduleAs input;
support vector machine model: the support vector machine classifier with radial basis function RBF kernel is adopted, and the specific decision function is expressed as follows:
，
wherein ：
is a decision function for determining a sample +.>Whether dangerous objects are carried;
and />Is a parameter of a support vector machine model;
is->Labels of the individual training samples;
is->Training samples;
is an RBF kernel function for calculating samples +.> and />Similarity between;
is the number of training samples;
classification decision: when (when)When it is, determine sample->Carrying dangerous objects and triggering an alarm module; when->When it is, determine sample->Is not dangerous; model updating: the dangerous object recognition module also comprises an online learning mechanism, and periodically updates parameters of the support vector machine model by continuously collecting new sample data and corresponding labels> and />To accommodate new types of hazards or environmental changes.
8. The remote person-carried hazard detection system based on multi-frequency sensory fusion technology according to claim 7, wherein said alarm module comprises a multi-level alarm mechanism, in particular,
primary alarm: when dangerous object is identifiedSupport vector machine decision function of other modulesOutputting a positive value but below a predetermined level one alarm threshold +.>When the system triggers a first-level alarm, and the system informs an operator of further manual inspection through a yellow LED lamp of the young and a low-volume alarm age;
secondary alarm: when (when)Outputting a positive value and higher than or equal to +.>But below the secondary alarm threshold +.>When the system is used, the system can trigger a secondary alarm, an operator is notified through a continuous red LED lamp and a high-volume alarm bell, and the entrance and exit of the monitoring area are automatically locked until manual release is achieved.
9. The remote person-carried hazard detection system based on multi-frequency sensing fusion technology of claim 8, wherein said alarm module further comprises an automatic notification unit that, upon triggering a secondary alarm, immediately sends the associated alarm information and hazard characteristic data to a communication device of a pre-configured security personnel or emergency response team, wherein the notification content comprises the current threat level, the specific location of the monitored area, and the hazard identification module support vector machine decision functionIs a function of the output value of (a).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311263036.4A CN116990883B (en) | 2023-09-27 | 2023-09-27 | Remote person-carried dangerous object detection system based on multi-spectrum sensing fusion technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311263036.4A CN116990883B (en) | 2023-09-27 | 2023-09-27 | Remote person-carried dangerous object detection system based on multi-spectrum sensing fusion technology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116990883A true CN116990883A (en) | 2023-11-03 |
| CN116990883B CN116990883B (en) | 2023-12-15 |
Family
ID=88528690
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311263036.4A Active CN116990883B (en) | 2023-09-27 | 2023-09-27 | Remote person-carried dangerous object detection system based on multi-spectrum sensing fusion technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116990883B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117520753A (en) * | 2024-01-05 | 2024-02-06 | 河北中体善建体育产业有限公司 | Early warning system and method for ice and snow sports |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050110672A1 (en) * | 2003-10-10 | 2005-05-26 | L-3 Communications Security And Detection Systems, Inc. | Mmw contraband screening system |
| US20050232459A1 (en) * | 2004-04-14 | 2005-10-20 | Rowe Richard L | Multi-source surveillance portal |
| WO2007011391A2 (en) * | 2004-10-22 | 2007-01-25 | Set Associates | System and method for standoff detection of human carried exposives |
| US20090279664A1 (en) * | 2008-05-08 | 2009-11-12 | L-3 Communications Security And Detection Systems, Inc. | Adaptive scanning in an imaging system |
| WO2010076261A1 (en) * | 2008-12-31 | 2010-07-08 | Thales | Security system comprising sensors in a corridor for uncovering hazardous items |
| US20140028457A1 (en) * | 2006-10-11 | 2014-01-30 | Thermal Matrix USA, Inc. | Real Time Threat Detection System |
| US20210271922A1 (en) * | 2018-06-29 | 2021-09-02 | Logistics and Supply Chain MultiTech R&D Centre Limited | Multimodal imaging sensor calibration method for accurate image fusion |
| CN114355462A (en) * | 2021-12-28 | 2022-04-15 | 重庆邮电大学 | Human hidden dangerous object detection method and medium based on micro-Doppler characteristics |
| US20230245452A1 (en) * | 2023-04-06 | 2023-08-03 | Intel Corporation | Adaptive multimodal event detection |
-
2023
- 2023-09-27 CN CN202311263036.4A patent/CN116990883B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050110672A1 (en) * | 2003-10-10 | 2005-05-26 | L-3 Communications Security And Detection Systems, Inc. | Mmw contraband screening system |
| US20050232459A1 (en) * | 2004-04-14 | 2005-10-20 | Rowe Richard L | Multi-source surveillance portal |
| WO2007011391A2 (en) * | 2004-10-22 | 2007-01-25 | Set Associates | System and method for standoff detection of human carried exposives |
| US20140028457A1 (en) * | 2006-10-11 | 2014-01-30 | Thermal Matrix USA, Inc. | Real Time Threat Detection System |
| US20090279664A1 (en) * | 2008-05-08 | 2009-11-12 | L-3 Communications Security And Detection Systems, Inc. | Adaptive scanning in an imaging system |
| WO2010076261A1 (en) * | 2008-12-31 | 2010-07-08 | Thales | Security system comprising sensors in a corridor for uncovering hazardous items |
| US20210271922A1 (en) * | 2018-06-29 | 2021-09-02 | Logistics and Supply Chain MultiTech R&D Centre Limited | Multimodal imaging sensor calibration method for accurate image fusion |
| CN114355462A (en) * | 2021-12-28 | 2022-04-15 | 重庆邮电大学 | Human hidden dangerous object detection method and medium based on micro-Doppler characteristics |
| US20230245452A1 (en) * | 2023-04-06 | 2023-08-03 | Intel Corporation | Adaptive multimodal event detection |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117520753A (en) * | 2024-01-05 | 2024-02-06 | 河北中体善建体育产业有限公司 | Early warning system and method for ice and snow sports |
| CN117520753B (en) * | 2024-01-05 | 2024-04-05 | 河北中体善建体育产业有限公司 | Early warning system and method for ice and snow sports |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116990883B (en) | 2023-12-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7178343B2 (en) | How to set up safety inspection systems and safety inspection equipment | |
| US10713914B2 (en) | Intelligent security management system | |
| CN116990883B (en) | Remote person-carried dangerous object detection system based on multi-spectrum sensing fusion technology | |
| US7961096B2 (en) | System and method for detection of EAS marker shielding | |
| Qureshi et al. | QuickBlaze: early fire detection using a combined video processing approach | |
| US8102260B2 (en) | Methods, systems and devices for detecting threatening objects and for classifying magnetic data | |
| WO2008103206A1 (en) | Surveillance systems and methods | |
| CN110167344B (en) | System and method for detecting flying animals | |
| US10949677B2 (en) | Method and system for detecting concealed objects using handheld thermal imager | |
| WO2021156153A1 (en) | System, method, and computer program product for automatically configuring a detection device | |
| KR101876797B1 (en) | Method And Apparatus for Detecting Target in Radar for Border Fences | |
| Ihekoronye et al. | Aerial supervision of drones and other flying objects using convolutional neural networks | |
| CN114355462A (en) | Human hidden dangerous object detection method and medium based on micro-Doppler characteristics | |
| US20160133023A1 (en) | Method for image processing, presence detector and illumination system | |
| JP7388532B2 (en) | Processing system, processing method and program | |
| Shafay et al. | Programmable broad learning system to detect concealed and imbalanced baggage threats | |
| WO2020086520A1 (en) | Artificial intelligence based motion detection | |
| US20210201074A1 (en) | Method and system for detecting concealed objects using handheld thermal imager | |
| Gaddipati et al. | Real-time human intrusion detection for home surveillance based on IOT | |
| Khan et al. | Recognizing foreign object debris (FOD): false alarm reduction implementation | |
| CN117331138B (en) | Intelligent detection system of intelligent analyzer | |
| RU2781768C1 (en) | Method, system and device for detection of people for inspection during passage of metal detector | |
| Muthukkumarasamy et al. | Intelligent illicit object detection system for enhanced aviation security | |
| KR102361089B1 (en) | Security inspection system using radar | |
| Sudha et al. | Smart X-Ray Baggage Security System using Machine Learning Algorithm |
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 |