Intelligent security alarm method and system based on FMCW radar

Abstract

The invention relates to an intelligent security alarm method and system based on an FMCW radar, and belongs to the technical field of intelligent security. The method comprises the following steps: acquiring a detection target signal through an FMCW radar, and suppressing background clutter through a DPCA technology according to the detection target signal to obtain a moving target speed; obtaining a behavior sequence through a sequence conversion model according to the moving target speed; obtaining a video frame, and carrying out behavior feature modeling according to the video frame to obtain a human body behavior feature model; performing abnormal behavior detection through a human body behavior feature model according to the behavior sequence to obtain an abnormal behavior; and generating an alarm signal according to the abnormal behavior to enable the buzzer to give an alarm, and transmitting an abnormal behavior video frame to the intelligent terminal. The intelligent security alarm based on the FMCW radar is realized, and the accuracy, timeliness and intelligent level of the security alarm are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent security, and in particular relates to an intelligent security alarm method and system based on FMCW radar. Background Art

[0002] With the development of social economy and the progress of informatization, people have higher and higher requirements for security, and property and personal safety have become the primary requirements for home security. At present, the application of security has been relatively widespread. The most common method is to install cameras and remotely control and monitor through computers or mobile phones. However, this method lacks timeliness, and users cannot obtain information at the first time of an accident. In addition, the cost is relatively high. Therefore, how to achieve the timeliness, intelligence and economy of the security system has become an important development direction of the current security industry.

[0003] Security systems based on big data and artificial intelligence technology have made significant progress in improving timeliness and intelligence, but false alarms still exist. FMCW radar has broad application prospects in the field of intelligent security due to its advantages in moving target detection and motion parameter extraction, but its detection ability is easily interfered by background clutter. How to effectively suppress background clutter and integrate it with artificial intelligence to improve the accuracy, timeliness and intelligence level of security systems in risk identification still needs further research. Summary of the invention

[0004] In order to solve the above problems existing in the prior art, the present invention provides an intelligent security alarm method and system based on FMCW radar.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] An intelligent security alarm method based on FMCW radar, comprising:

[0007] Acquire a detection target signal by using an FMCW radar, and obtain a moving target speed by suppressing background clutter by using a DPCA technique according to the detection target signal;

[0008] Obtaining a behavior sequence through a sequence conversion model according to the moving target speed;

[0009] Acquire a video frame, and obtain a human behavior feature model by modeling the behavior features according to the video frame;

[0010] Performing abnormal behavior detection through the human behavior feature model according to the behavior sequence to obtain abnormal behavior;

[0011] An alarm signal is generated according to the abnormal behavior to make a buzzer sound an alarm, and the abnormal behavior video frame is transmitted to the intelligent terminal.

[0012] Preferably, the behavior sequence includes a target movement trajectory, a target posture sequence and a target speed sequence, the moving target speed includes a radial speed and an azimuth speed, and the behavior sequence obtained by using a sequence conversion model according to the moving target speed includes:

[0013] According to the moving target speed, a velocity component is obtained by converting the velocity of the moving target through a Cartesian coordinate system;

[0014] Obtaining the target moving trajectory by integral calculation according to the velocity component;

[0015] Obtaining a current azimuth angle through azimuth angle calculation according to the azimuth velocity;

[0016] Obtaining the target posture sequence by changing calculation according to the current azimuth;

[0017] A target speed sequence is obtained by calculating the overall speed according to the moving target speed.

[0018] Preferably, obtaining a human behavior feature model by modeling the behavior features according to the video frame comprises:

[0019] Detecting a person portrait in the video frame by using the Mask R-CNN algorithm according to the video frame;

[0020] Extracting the moving trajectory, posture sequence and speed sequence from the video frame portrait by using the DAHM algorithm;

[0021] According to the movement trajectory, a movement trajectory model is obtained by movement trajectory modeling;

[0022] Obtaining a posture model through posture modeling according to the posture sequence;

[0023] Obtaining a velocity model through velocity modeling according to the velocity sequence;

[0024] The human behavior characteristic model is obtained according to the movement trajectory model, the posture model and the speed model.

[0025] Preferably, obtaining a movement trajectory model by movement trajectory modeling according to the movement trajectory comprises:

[0026] The trajectory distance is obtained by calculating the trajectory change cost according to the moving trajectory;

[0027] Obtaining a regularized trajectory distance by regularization calculation according to the trajectory distance;

[0028] A cluster set is obtained by K-means clustering according to the regular trajectory distance, and the cluster set carries a sample number label;

[0029] Obtaining the total number of samples, and obtaining the sample ratio according to the sample number label and the total number of samples;

[0030] Preset a ratio threshold, and obtain a trajectory mark by ratio judgment according to the sample ratio and the ratio threshold, wherein the trajectory mark includes an abnormal trajectory and a normal trajectory;

[0031] The ratio judgment includes judging whether the sample ratio is less than the ratio threshold, if yes, the trajectory is marked as the abnormal trajectory to obtain abnormal trajectory clustering; if no, the trajectory is marked as the normal trajectory to obtain normal trajectory clustering.

[0032] Preferably, obtaining the posture model through posture modeling according to the posture sequence comprises:

[0033] Obtain a human body posture library and obtain human body dynamic feature samples;

[0034] Obtaining static human body feature samples by frame splitting according to the human body dynamic feature samples;

[0035] Obtain a human body posture library sample according to the human body posture library, obtain a posture distance according to the static human body feature sample and the human body posture library sample through posture distance calculation, and obtain a posture category of the static human body feature sample according to the posture distance;

[0036] Obtaining a static human body feature sample posture set by classification according to the static human body feature sample posture categories;

[0037] Obtain a normal trajectory cluster centroid trajectory point, obtain a static posture human body sample trajectory point according to the static human body feature sample posture set, the static posture human body sample trajectory point is a trajectory point of a sample in the static human body feature sample posture set, and calculate the posture trajectory distance according to the normal trajectory cluster centroid trajectory point and the static posture human body sample trajectory point through the trajectory transformation cost;

[0038] Obtaining a posture cluster set by K-means clustering according to the posture trajectory distance, wherein the posture cluster set carries a label of the number of posture samples;

[0039] Obtaining the total number of posture samples, and obtaining the posture sample ratio according to the number of posture samples and the total number of posture samples;

[0040] Preset a posture ratio threshold, and determine the posture category by posture ratio according to the posture sample ratio and the posture ratio threshold, wherein the posture category includes abnormal posture and normal posture;

[0041] The posture ratio judgment includes judging whether the posture sample ratio is less than the posture ratio threshold value. If yes, the posture category is the abnormal posture; if no, the posture category is the normal posture.

[0042] Preferably, obtaining a velocity model through velocity modeling according to the velocity sequence comprises:

[0043] Obtaining a normal posture speed of a normal trajectory cluster centroid trajectory point, and calculating a speed threshold through a Gaussian function according to the normal posture speed of the normal trajectory cluster centroid trajectory point;

[0044] A speed interval is obtained according to the speed threshold, and a speed category is obtained by speed anomaly judgment according to the normal posture speed of the centroid trajectory point of the normal trajectory cluster and the speed interval, wherein the speed category includes a normal speed and an abnormal speed.

[0045] Preferably, the speed abnormality judgment includes judging whether the normal posture speed of the centroid trajectory point of the normal trajectory cluster is within the speed range, if yes, the speed category is the normal speed; if no, the speed category is the abnormal speed.

[0046] An intelligent security alarm method based on FMCW radar and an intelligent security alarm system based on FMCW radar include a data preprocessing module, a speed conversion module, a behavior modeling module, a behavior detection module, and an abnormal response module, including:

[0047] The data preprocessing module is used to obtain a detection target signal through the FMCW radar, and obtain the moving target speed by suppressing background clutter through the DPCA technology according to the detection target signal;

[0048] The speed conversion module is used to obtain a behavior sequence according to the moving target speed through a sequence conversion model;

[0049] The behavior modeling module is used to obtain video frames and obtain a human behavior feature model through behavior feature modeling based on the video frames;

[0050] The behavior detection module is used to detect abnormal behavior according to the behavior sequence through the human behavior feature model to obtain abnormal behavior;

[0051] The abnormal response module is used for the central controller to control the buzzer alarm and send the abnormal behavior to the smart terminal.

[0052] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the intelligent security alarm method based on FMCW radar is implemented.

[0053] A storage medium containing computer executable instructions, wherein the computer executable instructions are used to execute the above-mentioned intelligent security alarm method based on FMCW radar when executed by a computer processor.

[0054] The beneficial effects of the present invention are:

[0055] (1) The target signal is acquired through FMCW radar detection, and the background clutter is suppressed by combining DPCA technology, which enhances the detection capability of moving targets, reduces the impact of background interference on detection, improves the detection accuracy of moving targets, reduces false alarms, and reduces the cost of intelligent security.

[0056] (2) By associating the moving target speed with the behavior sequence and using the sequence conversion model, the behavior sequence can be quickly inferred. This reduces the delay of pure visual information processing and improves the real-time and accuracy of intelligent security detection anomalies.

[0057] (3) The Mask R-CNN algorithm is used to detect human portraits in video frames, and the DAHM algorithm is used to extract the movement trajectory, posture sequence, and speed sequence of the human portraits in video frames. A human behavior feature model is constructed based on the movement trajectory, posture sequence, and speed sequence, which improves the accuracy of human behavior feature modeling and lays the foundation for subsequent intelligent security detection anomalies.

[0058] (4) The human behavior feature model is used to detect abnormal behavior sequences, realize multiple verifications of movement trajectory, posture sequence and speed sequence, reduce the possibility of false alarms, and improve the intelligence of smart security. At the same time, the central controller controls the buzzer alarm and sends the abnormal behavior to the smart terminal, which improves the real-time performance of smart security alarms. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0060] Figure 1 The present invention is a flowchart of an intelligent security alarm method based on FMCW radar. DETAILED DESCRIPTION

[0061] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0062] See also Figure 1 , an intelligent security alarm method based on FMCW radar, comprising:

[0063] S1: Acquire a detection target signal through the FMCW radar, and obtain the moving target speed by suppressing background clutter through the DPCA technology according to the detection target signal;

[0064] S2: Obtaining a behavior sequence through a sequence conversion model according to the moving target speed;

[0065] S3: Obtain a video frame, and obtain a human behavior feature model through behavior feature modeling according to the video frame;

[0066] S4: performing abnormal behavior detection using the human behavior feature model according to the behavior sequence to obtain abnormal behavior;

[0067] S5: The central controller controls the buzzer to alarm and sends the abnormal behavior to the smart terminal.

[0068] Specifically, in step S1, the FMCW radar is a frequency modulated continuous wave radar, which can use frequency modulation to measure the distance, speed and azimuth of the detected target; the moving target speed includes radial speed and azimuth speed.

[0069] In step S2, the behavior sequence includes a target moving trajectory, a target posture sequence and a target speed sequence, and the behavior sequence obtained by the sequence conversion model according to the moving target speed parameter includes:

[0070] S201: Obtaining velocity components according to the moving target velocity through a Cartesian coordinate system velocity conversion, wherein the velocity components include an x-direction velocity and a y-direction velocity;

[0071] The Cartesian coordinate system velocity conversion expression is:

[0072]

[0073] Where t represents time, v x (t) represents the x-direction velocity, v y (t) represents the y-direction velocity, v r (t) represents the radial velocity, v θ (t) represents the azimuth velocity, θ(t) represents the azimuth angle of the moving target relative to the polar coordinates of the FMCW radar;

[0074] S202: Obtaining the target moving trajectory by integral calculation according to the velocity component;

[0075] The integral calculation expression is:

[0076]

[0077] Among them, x0 represents the x coordinate of the initial position of the moving target, y0 represents the y coordinate of the initial position of the moving target, and vx (t) represents the x-direction velocity, v y (t) represents the y-direction speed, x(t) represents the x-coordinate of the trajectory position of the moving target, and y(t) represents the y-coordinate of the trajectory position of the moving target;

[0078] S203: obtaining a current azimuth angle through azimuth angle calculation according to the azimuth velocity;

[0079] The direction angle calculation expression is:

[0080]

[0081] Among them, θ(t) represents the azimuth of the moving target relative to the polar coordinates of the FMCW radar, θ(0) represents the azimuth of the initial position of the moving target relative to the polar coordinates of the FMCW radar, and v θ (t) represents the azimuth velocity, R(t) represents the distance between the moving target and the FMCW radar;

[0082] S204: Obtaining the target posture sequence by changing calculation according to the current azimuth angle;

[0083] The change calculation expression is:

[0084]

[0085] Wherein, θ(t) represents the azimuth of the moving target relative to the polar coordinates of the FMCW radar, φ(t) represents the target posture sequence, and φ(0) represents the initial posture sequence;

[0086] S205: Obtaining the target speed sequence through overall speed calculation according to the moving target speed;

[0087] The overall speed calculation expression is:

[0088]

[0089] Among them, v θ (t) represents the azimuthal velocity, v r (t) represents the radial velocity, and v(t) represents the target velocity sequence.

[0090] In step S3, the behavior feature modeling includes:

[0091] S301: Detecting a human portrait in the video frame by using the Mask R-CNN algorithm according to the video frame;

[0092] S302: extracting a moving trajectory, a posture sequence and a speed sequence according to the video frame portrait by using a DAHM algorithm;

[0093] S303: Obtaining a movement trajectory model by movement trajectory modeling according to the movement trajectory;

[0094] S304: Obtaining a posture model through posture modeling according to the posture sequence;

[0095] S305: Obtaining a velocity model through velocity modeling according to the velocity sequence;

[0096] S306: Obtaining the human behavior characteristic model according to the movement trajectory model, the posture model and the speed model.

[0097] Specifically, the movement trajectory modeling includes:

[0098] S303-1: Calculating a trajectory distance according to the moving trajectory by using a trajectory change cost;

[0099] The trajectory change cost calculation expression is:

[0100]

[0101] oth_cost(R a ,R b )=min{RERP(R a ,REST(R b ))+ins_cost(r b，n ),

[0102] RERP(REST(R a ),REST(R b ))+rep_cost(r a,m , r b，n ),

[0103] RERP(REST(R a ), R b )+del_cost(r a，m , r b，n )},

[0104]

[0105] Among them, RERP (R a , R b ) represents the trajectory distance, R a and R b represents the moving trajectory, r b，j Represents the moving trajectory R b The jth trajectory point in a，i Represents the moving trajectory R a The i-th trajectory point in ins_cost(r b，j) represents the trajectory point r b，j Insert track R a The cost of Represents the trajectory R a Delete the trajectory point r a，i The cost, oth_cost(R a , R b ) represents the cost in other cases, represents the empty set, m is the trajectory R a length, n is the trajectory R b Length, REST(R a ) represents the trajectory R a The remaining set of trajectory points after removing a trajectory point, REST (R b ) represents the trajectory R b The remaining set of trajectory points after removing a trajectory point, RERP (R a REST(R b )) indicates that from trajectory R b The remaining trajectory point set after removing a trajectory point and the trajectory R a The trajectory distance, RERP(REST(R a ), REST(R b )) indicates that from trajectory R a The remaining trajectory point set after removing a trajectory point is the same as the set of trajectory points from trajectory R b The trajectory distance between the remaining trajectory points after removing a trajectory point, RERP (REST (R a ), R b ) represents the trajectory R a The remaining trajectory point set after removing a trajectory point and the trajectory R b The trajectory distance between them, ins_cost(r b，n ) represents the trajectory point r b，n Insert track R a The cost in rep_cost(ra,m,r b，n ) represents the trajectory point r a，m Replaced by trajectory point r b，n The cost, || represents the Euclidean distance calculation;

[0106] S303-2: Obtaining a regularized trajectory distance by regularization calculation according to the trajectory distance;

[0107] The regularized calculation expression is:

[0108]

[0109] Among them, D(R a , R b) represents the regular trajectory distance, RERP (R a , R b ) represents the trajectory distance, and max represents the maximum value calculation;

[0110] S303-3: Obtaining a cluster set by K-means clustering according to the regular trajectory distance, wherein the cluster set carries a sample number label;

[0111] S303-4: Obtain the total number of samples, and obtain the sample ratio according to the sample number label and the total number of samples;

[0112] S303-5: Preset a ratio threshold, and obtain a trajectory mark by ratio judgment according to the sample ratio and the ratio threshold, wherein the trajectory mark includes an abnormal trajectory and a normal trajectory;

[0113] The ratio judgment includes judging whether the sample ratio is less than the ratio threshold, if yes, the trajectory is marked as the abnormal trajectory to obtain abnormal trajectory clustering; if no, the trajectory is marked as the normal trajectory to obtain normal trajectory clustering;

[0114] S303 - 6 : Obtain the movement trajectory model through steps S303 - 1 to S303 - 5 .

[0115] Specifically, the posture modeling includes:

[0116] S304-1: Acquire a human body posture library and obtain human body dynamic feature samples, where the human body dynamic feature samples are dynamic samples in the normal trajectory cluster;

[0117] S304-2: Obtaining static human body feature samples by frame splitting according to the human body dynamic feature samples;

[0118] S304-3: obtaining a human body posture library sample according to the human body posture library, obtaining a posture distance according to the static human body feature sample and the human body posture library sample through posture distance calculation, and obtaining a posture category of the static human body feature sample according to the posture distance;

[0119] The posture distance calculation expression is:

[0120]

[0121] Among them, k i represents the static human feature sample posture, k j represents the sample posture of the human posture library, D(k i , k j ) represents the posture distance, D(k i，n , k j，n) represents the posture distance between the static human feature sample posture and the nth skeleton key point of the human posture library sample posture, k j，n =(0,0) indicates that the nth skeletal key point of the sample pose in the human pose library does not exist, ki,n indicates the nth skeletal key point of the static human feature sample pose, kj,n indicates the nth skeletal key point of the sample pose in the human pose library, k i，n，x Represents the horizontal coordinate of the nth skeletal key point of the static human feature sample posture, k i，n，y Indicates the ordinate of the nth skeletal key point of the static human feature sample posture, k j，n，x Indicates the horizontal coordinate of the nth skeletal key point of the human body posture library sample posture, k j，n，y Indicates the vertical coordinate of the nth bone key point of the human body posture library sample posture, Indicates Euclidean distance calculation;

[0122] S304-4: Obtaining a static human body feature sample posture set by classification according to the static human body feature sample posture categories;

[0123] S304-5: Obtain a normal trajectory cluster centroid trajectory point, obtain a static posture human body sample trajectory point according to the static human body feature sample posture set, the static posture human body sample trajectory point is a trajectory point of a sample in the static human body feature sample posture set, and calculate the posture trajectory distance according to the normal trajectory cluster centroid trajectory point and the static posture human body sample trajectory point through the trajectory transformation cost;

[0124] S304-6: Obtaining a posture cluster set by K-means clustering according to the posture trajectory distance, wherein the posture cluster set carries a label of the number of posture samples;

[0125] S304-7: Obtain the total number of posture samples, and obtain the posture sample ratio according to the number of posture samples and the total number of posture samples;

[0126] S304-8: Preset a posture ratio threshold, and determine the posture category by posture ratio according to the posture sample ratio and the posture ratio threshold, wherein the posture category includes abnormal posture and normal posture;

[0127] The posture ratio judgment includes judging whether the posture sample ratio is less than the posture ratio threshold value, if yes, the posture category is the abnormal posture; if no, the posture category is the normal posture;

[0128] S304-9: Obtain the posture model through steps S304-1 to S304-8.

[0129] In the present embodiment, the human posture library includes standing, walking, running, squatting, sitting, kicking, lying, jumping, bowing, drinking water, riding a bicycle, pushing a cart, roller skating, punching and making a phone call, and each human posture is composed of 17 human skeleton key point coordinates and the categories described in the posture. The distance calculation of the human posture in the posture sequence and the human posture library is converted into the distance between the human skeleton key points, and the posture category of the human body is determined as the posture category in the posture library that is closest to it.

[0130] Specifically, the velocity modeling includes:

[0131] S305-1: Obtain the normal posture speed of the centroid trajectory point of the normal trajectory cluster, and calculate the speed threshold through the Gaussian function according to the normal posture speed of the centroid trajectory point of the normal trajectory cluster;

[0132] The Gaussian function calculation expression is:

[0133]

[0134] Where F(x) is the cumulative distribution function of the normal distribution, σ, μ represent the average speed, represents the normal posture speed of the centroid trajectory point of the normal trajectory cluster, i represents the i-th normal trajectory cluster, j represents the centroid trajectory point of the j-th normal trajectory cluster, k represents the normal posture of the centroid trajectory point of the k-th normal trajectory cluster, x is the normal speed threshold variable, p represents the total number of samples, q represents the q-th sample, σ 2 represents the speed variance;

[0135] S305-2: obtaining a speed interval according to the speed threshold, and obtaining a speed category by speed abnormality judgment according to the normal posture speed of the centroid trajectory point of the normal trajectory cluster and the speed interval, wherein the speed category includes a normal speed and an abnormal speed;

[0136] The speed abnormality judgment includes judging whether the normal posture speed of the centroid trajectory point of the normal trajectory cluster is within the speed interval, if yes, the speed category is the normal speed; if no, the speed category is the abnormal speed;

[0137] S305-3: Obtain the velocity model through steps S305-1 to S305-2.

[0138] In this embodiment, the posture modeling is based on the normal trajectory cluster centroid trajectory point, and the speed modeling is based on each normal posture of the normal trajectory cluster centroid trajectory point.

[0139] In this embodiment, the x value when the normal distribution cumulative distribution function F(x)=0.85 is calculated as the upper threshold of the normal speed; the x value when the normal distribution cumulative distribution function F(x)=0.15 is calculated as the lower threshold of the normal speed. After obtaining the upper and lower thresholds of the normal speed, they are recorded in each normal posture category, representing the normal speed range of this posture.

[0140] An intelligent security alarm method based on FMCW radar and an intelligent security alarm system based on FMCW radar include a data preprocessing module, a speed conversion module, a behavior modeling module, a behavior detection module, and an abnormal response module, including:

[0141] The data preprocessing module is used to obtain a detection target signal through the FMCW radar, and obtain the moving target speed by suppressing background clutter through the DPCA technology according to the detection target signal;

[0142] The speed conversion module is used to obtain a behavior sequence according to the moving target speed through a sequence conversion model;

[0143] The behavior modeling module is used to obtain video frames and obtain a human behavior feature model through behavior feature modeling based on the video frames;

[0144] The behavior detection module is used to detect abnormal behavior according to the behavior sequence through the human behavior feature model to obtain abnormal behavior;

[0145] The abnormal response module is used for the central controller to control the buzzer alarm and send the abnormal behavior to the smart terminal.

[0146] The computer storage medium of the embodiment of the present invention may adopt any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device.

[0147] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0148] The program code included on the computer readable medium can be transmitted with any appropriate medium, including but not limited to wireless, electric wire, optical cable, RF, etc., or any suitable combination of the above. The computer program code for performing the operation of the present invention can be written in one or more programming languages ​​or their combinations, and the programming language includes object-oriented programming languages-such as Java, Smalltalk, C++, and also includes conventional procedural programming languages-such as "C" language or similar programming languages. The program code can be executed completely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on the remote computer, or completely on the remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (for example, using an Internet service provider to connect through the Internet).

[0149] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. An intelligent security alarm method based on FMCW radar, characterized in that: The following steps are involved: Acquire a detection target signal by using an FMCW radar, and obtain a moving target speed by suppressing background clutter by using a DPCA technique according to the detection target signal; Obtaining a behavior sequence through a sequence conversion model according to the moving target speed; Acquire a video frame, and obtain a human behavior feature model by modeling the behavior features according to the video frame; Performing abnormal behavior detection through the human behavior feature model according to the behavior sequence to obtain abnormal behavior; An alarm signal is generated according to the abnormal behavior to make a buzzer sound an alarm, and the abnormal behavior video frame is transmitted to the intelligent terminal.

2. The FMCW radar-based intelligent security alarm method according to claim 1, characterized in that: The behavior sequence includes a target moving trajectory, a target posture sequence and a target speed sequence, the moving target speed includes a radial speed and an azimuth speed, and the behavior sequence obtained by the sequence conversion model according to the moving target speed includes: According to the moving target speed, a velocity component is obtained by converting the velocity of the moving target through a Cartesian coordinate system; Obtaining the target moving trajectory by integral calculation according to the velocity component; Obtaining a current azimuth angle through azimuth angle calculation according to the azimuth velocity; Obtaining the target posture sequence by changing calculation according to the current azimuth; A target speed sequence is obtained by calculating the overall speed according to the moving target speed.

3. The FMCW radar-based intelligent security alarm method according to claim 1, characterized in that: The obtaining of a human behavior feature model by modeling the behavior features according to the video frame comprises: Detecting a person portrait in the video frame by using the Mask R-CNN algorithm according to the video frame; Extracting the moving trajectory, posture sequence and speed sequence from the video frame portrait by using the DAHM algorithm; According to the movement trajectory, a movement trajectory model is obtained by movement trajectory modeling; Obtaining a posture model through posture modeling according to the posture sequence; Obtaining a velocity model through velocity modeling according to the velocity sequence; The human behavior characteristic model is obtained according to the movement trajectory model, the posture model and the speed model.

4. The FMCW radar-based intelligent security alarm method according to claim 3, characterized in that: The obtaining of a movement trajectory model by movement trajectory modeling according to the movement trajectory comprises: The trajectory distance is obtained by calculating the trajectory change cost according to the moving trajectory; Obtaining a regularized trajectory distance by regularization calculation according to the trajectory distance; A cluster set is obtained by K-means clustering according to the regular trajectory distance, and the cluster set carries a sample number label; Obtaining the total number of samples, and obtaining the sample ratio according to the sample number label and the total number of samples; Preset a ratio threshold, and obtain a trajectory mark by ratio judgment according to the sample ratio and the ratio threshold, wherein the trajectory mark includes an abnormal trajectory and a normal trajectory; The ratio judgment includes judging whether the sample ratio is less than the ratio threshold. If yes, the trajectory is marked as the abnormal trajectory to obtain abnormal trajectory clustering; if no, the trajectory is marked as the normal trajectory to obtain normal trajectory clustering.

5. The FMCW radar-based intelligent security alarm method according to claim 3, characterized in that: The obtaining of the posture model by posture modeling according to the posture sequence comprises: Obtain a human body posture library and obtain human body dynamic feature samples; Obtaining static human body feature samples by frame splitting according to the human body dynamic feature samples; Obtain a human body posture library sample according to the human body posture library, obtain a posture distance according to the static human body feature sample and the human body posture library sample through posture distance calculation, and obtain a posture category of the static human body feature sample according to the posture distance; Obtaining a static human body feature sample posture set by classification according to the static human body feature sample posture categories; Obtain a normal trajectory cluster centroid trajectory point, obtain a static posture human body sample trajectory point according to the static human body feature sample posture set, the static posture human body sample trajectory point is a trajectory point of a sample in the static human body feature sample posture set, and calculate the posture trajectory distance according to the normal trajectory cluster centroid trajectory point and the static posture human body sample trajectory point through the trajectory transformation cost; Obtaining a posture cluster set by K-means clustering according to the posture trajectory distance, wherein the posture cluster set carries a label of the number of posture samples; Obtaining the total number of posture samples, and obtaining the posture sample ratio according to the number of posture samples and the total number of posture samples; Preset a posture ratio threshold, and determine the posture category by posture ratio according to the posture sample ratio and the posture ratio threshold, wherein the posture category includes abnormal posture and normal posture; The posture ratio judgment includes judging whether the posture sample ratio is less than the posture ratio threshold value. If yes, the posture category is the abnormal posture; if no, the posture category is the normal posture.

6. The FMCW radar-based intelligent security alarm method according to claim 3, characterized in that: The obtaining of the speed model by speed modeling according to the speed sequence comprises: Obtaining a normal posture speed of a normal trajectory cluster centroid trajectory point, and calculating a speed threshold through a Gaussian function according to the normal posture speed of the normal trajectory cluster centroid trajectory point; A speed interval is obtained according to the speed threshold, and a speed category is obtained by speed anomaly judgment according to the normal posture speed of the centroid trajectory point of the normal trajectory cluster and the speed interval, wherein the speed category includes a normal speed and an abnormal speed.

7. The FMCW radar-based intelligent security alarm method according to claim 6, characterized in that: The speed abnormality judgment includes judging whether the normal posture speed of the centroid trajectory point of the normal trajectory cluster is within the speed range. If yes, the speed category is the normal speed; if not, the speed category is the abnormal speed.

8. An intelligent security alarm system based on FMCW radar using the intelligent security alarm method based on FMCW radar according to claims 1-7, comprising a data preprocessing module, a speed conversion module, a behavior modeling module, a behavior detection module, and an abnormal response module, characterized in that: include: The data preprocessing module is used to obtain a detection target signal through the FMCW radar, and obtain the moving target speed by suppressing background clutter through the DPCA technology according to the detection target signal; The speed conversion module is used to obtain a behavior sequence according to the moving target speed through a sequence conversion model; The behavior modeling module is used to obtain video frames and obtain a human behavior feature model through behavior feature modeling based on the video frames; The behavior detection module is used to detect abnormal behavior according to the behavior sequence through the human behavior feature model to obtain abnormal behavior; The abnormal response module is used for the central controller to control the buzzer alarm and send the abnormal behavior to the smart terminal.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the FMCW radar-based intelligent security alarm method as described in any one of claims 1 to 7 is implemented.

10. A storage medium containing computer executable instructions, characterized in that: The computer executable instructions are used to execute the FMCW radar-based intelligent security alarm method as described in any one of claims 1 to 7 when executed by a computer processor.

Images