Article detection method and device and security check door

Abstract

The invention discloses an article detection method and device and a security check door, and belongs to the technical field of security check. The article detection method comprises the following steps: acquiring an electromagnetic signal of each to-be-detected sub-region through scanning; correcting the electromagnetic signal of each to-be-detected sub-region to obtain a corrected electromagnetic signal of each to-be-detected sub-region; and according to the correction electromagnetic signal of each to-be-detected sub-region, whether a target object exists in the to-be-detected region is determined, and the to-be-detected region is composed of the to-be-detected sub-regions. According to the invention, article detection is carried out based on the corrected electromagnetic signal, and the accuracy and anti-interference performance of article detection are improved.

Description

Technical Field

[0001] The present application belongs to the field of security inspection technology, and in particular, relates to an object inspection method, device and security inspection door. Background Art

[0002] At present, most of the security gates on the market are based on the technical route of magnetic field scanning to obtain signal characteristics, and scan the detected objects by generating a strong magnetic field. This process involves magnetic field sensing technology, that is, using the change of magnetic field to detect objects. When the object passes through the security gate, its metal component will change the distribution of the magnetic field. This change will be detected by the sensor installed inside the security gate. During the magnetic field scanning process, the security gate will collect a large amount of original electromagnetic signals. However, affected by the surrounding environment, the original electromagnetic signal is prone to problems such as signal offset or missing. It needs to be corrected before it can be used to identify and distinguish different objects and subsequently determine whether the target object exists. It is difficult to solve the problems existing in the electromagnetic signal from the hardware perspective alone. It is necessary to propose a method to realize the correction of electromagnetic signals and apply it to the object detection scenario to improve the accuracy of object detection. Summary of the invention

[0003] The present application aims to solve at least one of the technical problems existing in the related art. To this end, the present application proposes an object detection method, device and security door, which detect objects based on the corrected electromagnetic signal, thereby improving the accuracy and anti-interference of object detection.

[0004] In a first aspect, the present application provides an object detection method, the method comprising:

[0005] Acquire electromagnetic signals of each sub-area to be detected by scanning;

[0006] Correcting the electromagnetic signals of the sub-regions to be detected to obtain corrected electromagnetic signals of the sub-regions to be detected;

[0007] It is determined whether a target object exists in the area to be detected according to the corrected electromagnetic signals of the sub-areas to be detected, and the area to be detected is composed of the sub-areas to be detected.

[0008] In the above technical scheme, the electromagnetic signals of each sub-area to be detected are acquired by scanning, and the electromagnetic signals of each sub-area to be detected are corrected to obtain the corrected electromagnetic signals of each sub-area to be detected. Then, based on the corrected electromagnetic signals of each sub-area to be detected, it is determined whether there is a target object in the area to be detected composed of the sub-areas to be detected. By dividing the sub-areas to be detected and correcting the electromagnetic signals, object detection is performed based on the corrected electromagnetic signals, thereby improving the accuracy and anti-interference ability of object detection.

[0009] According to an embodiment of the present application, determining whether there is a target object in the area to be detected based on the corrected electromagnetic signals of each sub-area to be detected includes:

[0010] Screening the corrected electromagnetic signals of each sub-area to be detected according to the allowable value, and retaining the stable electromagnetic signals of each sub-area to be detected;

[0011] Whether there is a target object in the to-be-detected area is determined according to the characteristic values ​​of the stable electromagnetic signals of each to-be-detected sub-area and its adjacent to-be-detected sub-area and the threshold range corresponding to the target object.

[0012] In the above technical scheme, the corrected electromagnetic signals of each sub-area to be detected are screened according to the allowed value, and the stable electromagnetic signals of each sub-area to be detected are retained. According to the characteristic values ​​of the stable electromagnetic signals of each sub-area to be detected and its adjacent sub-areas to be detected and the threshold range corresponding to the target object, it is determined whether there is a target object in the area to be detected, thereby improving the accuracy and anti-interference of object detection.

[0013] According to an embodiment of the present application, the correcting the electromagnetic signal of each sub-region to be detected to obtain the corrected electromagnetic signal of each sub-region to be detected includes:

[0014] Inputting the electromagnetic signals of each sub-region to be detected into a multivariate time series interpolation diffusion model to obtain a corrected electromagnetic signal of each sub-region to be detected;

[0015] The multivariate time series interpolation diffusion model is obtained based on electromagnetic signal sample training.

[0016] In the above technical scheme, the electromagnetic signal of each sub-area to be detected is input into a multivariate time series interpolation diffusion model obtained by training based on electromagnetic signal samples, so as to obtain a corrected electromagnetic signal of each sub-area to be detected, which solves the problem of electromagnetic signal defects and realizes the correction of electromagnetic signals of each sub-area to be detected. It is applied to object detection to improve the accuracy and anti-interference of object detection.

[0017] According to one embodiment of the present application, the multivariate time series interpolation diffusion model is trained by the following steps:

[0018] Extracting features from the electromagnetic signal sample to obtain features of the electromagnetic signal sample;

[0019] Performing frequency division processing on the electromagnetic signal sample characteristics to determine the main frequency signal characteristics of the electromagnetic signal sample and the non-main frequency signal characteristics of the electromagnetic signal sample;

[0020] Determining a joint Gaussian distribution according to the non-main frequency signal characteristics of the electromagnetic signal sample;

[0021] According to the joint Gaussian distribution, forward diffusion processing is performed on the electromagnetic signal sample features to obtain forward diffusion data of the electromagnetic signal sample features, wherein the forward diffusion data of the electromagnetic signal sample features includes the noisy electromagnetic signal sample features;

[0022] According to the forward diffusion data of the electromagnetic signal sample characteristics, the noisy electromagnetic signal sample characteristics are subjected to reverse diffusion processing. In the reverse diffusion processing, the loss of the multivariate time series interpolation diffusion model is determined, and the parameters of the multivariate time series interpolation diffusion model are optimized based on the loss.

[0023] In the above technical scheme, feature extraction is performed on the electromagnetic signal sample to obtain the electromagnetic signal sample feature, and the electromagnetic signal sample feature is subjected to frequency division processing to determine the main frequency signal feature and non-main frequency signal feature of the electromagnetic signal sample. According to the non-main frequency signal feature of the electromagnetic signal sample, a joint Gaussian distribution is determined. According to the joint Gaussian distribution, the electromagnetic signal sample feature is subjected to forward diffusion processing to obtain forward diffusion data of the electromagnetic signal sample feature. According to the forward diffusion data of the electromagnetic signal sample feature, the noisy electromagnetic signal sample feature is subjected to reverse diffusion processing. In the reverse diffusion processing, the loss of the multivariate time series interpolation diffusion model is determined, and the parameters of the model are optimized based on the loss, thereby realizing the training of the multivariate time series interpolation diffusion model, improving the accuracy and stability of the model in predicting the missing values ​​of the electromagnetic signal, and thereby improving the accuracy and stability of the electromagnetic signal correction. Object detection is performed based on the corrected electromagnetic signal, which can improve the accuracy and anti-interference of object detection.

[0024] According to an embodiment of the present application, performing forward diffusion processing on the electromagnetic signal sample feature according to the joint Gaussian distribution to obtain forward diffusion data of the electromagnetic signal sample feature includes:

[0025] According to the current time step and the preset time step, the joint Gaussian distribution is sampled to determine the current step forward noise;

[0026] Adding the current step forward noise to the previous step electromagnetic signal sample forward feature to obtain the current step electromagnetic signal sample forward feature;

[0027] Gradually increase the current time step and repeat the above steps until the current time step is equal to the preset time step, and determine that the positive feature of the electromagnetic signal sample at the current time step is the noisy electromagnetic signal sample feature;

[0028] The forward diffusion data of the electromagnetic signal sample characteristics also includes each current step forward noise and each current step electromagnetic signal sample forward characteristics.

[0029] In the above technical scheme, the joint Gaussian distribution is sampled according to the current time step and the preset time step, the forward noise of the current step is determined, the forward noise of the current step is added to the forward feature of the electromagnetic signal sample of the previous step, and the forward feature of the electromagnetic signal sample of the current step is obtained. The current time step is gradually increased and the above steps are repeated until the current time step is equal to the preset time step. The forward feature of the electromagnetic signal sample of the current step is determined to be the noisy electromagnetic signal sample feature, and the forward diffusion processing of the electromagnetic signal sample feature is realized. The forward diffusion data of the electromagnetic signal sample feature is obtained, which can be used for reverse diffusion processing of the noisy electromagnetic signal sample feature, and the multivariate time series interpolation diffusion model is optimized in the reverse diffusion processing.

[0030] According to an embodiment of the present application, when the current time step is 1, the electromagnetic signal sample feature is determined to be the positive feature of the electromagnetic signal sample in the previous step.

[0031] In the above technical solution, when the current time step is 1, the electromagnetic signal sample feature is determined to be the forward feature of the electromagnetic signal sample in the previous step, thereby realizing the first noise addition to the electromagnetic signal sample feature during the forward diffusion process.

[0032] According to an embodiment of the present application, the forward diffusion data of the electromagnetic signal sample feature is used to perform reverse diffusion processing on the noisy electromagnetic signal sample feature. In the reverse diffusion processing, the loss of the multivariate time series interpolation diffusion model is determined, and the parameters of the multivariate time series interpolation diffusion model are optimized based on the loss, including:

[0033] Determine the current step reverse noise according to the current time step and the reverse characteristics of the electromagnetic signal sample of the current step;

[0034] According to the reverse noise of the current step, the reverse feature of the electromagnetic signal sample of the current step is removed to determine the reverse feature of the electromagnetic signal sample of the previous step;

[0035] Determining the current step loss of the multivariate time series interpolation diffusion model through an error estimation model;

[0036] Optimizing the parameters of the multivariate time series interpolation diffusion model based on the current step loss;

[0037] Gradually reduce the current time step and repeat the above steps until the current time step is 1.

[0038] In the above technical scheme, the current step reverse noise is determined according to the current time step and the reverse characteristics of the current step electromagnetic signal sample, the previous step electromagnetic signal sample reverse characteristics are determined according to the reverse characteristics of the current step electromagnetic signal sample and the current step reverse noise, the current step loss of the multivariate time series interpolation diffusion model is determined through the error estimation model, the parameters of the multivariate time series interpolation diffusion model are optimized based on the current step loss, the current time step is gradually reduced and the above steps are repeated until the current time step is 1, the reverse diffusion processing of the noisy electromagnetic signal sample characteristics is realized, and the multivariate time series interpolation diffusion model is optimized in the reverse diffusion processing, the accuracy and stability of the model in predicting the missing values ​​of the electromagnetic signal are improved, and then the accuracy and stability of the electromagnetic signal correction are improved, and the object detection is performed based on the corrected electromagnetic signal, which can improve the accuracy and anti-interference of the object detection.

[0039] According to an embodiment of the present application, when the current time step is equal to the preset time step, the noisy electromagnetic signal sample feature is determined to be the inverse feature of the current step electromagnetic signal sample.

[0040] In the above technical solution, when the current time step is equal to the preset time step, the noisy electromagnetic signal sample characteristics are determined as the reverse characteristics of the current step electromagnetic signal sample, thereby realizing the first noise removal of the noisy electromagnetic signal sample characteristics during the reverse diffusion processing.

[0041] According to one embodiment of the present application, determining the current step loss of the multivariate time series interpolation diffusion model through an error estimation model includes:

[0042] determining a first current step loss according to the current step reverse noise and the current step forward noise;

[0043] Determining a second current step loss according to the reverse feature of the electromagnetic signal sample of the previous step and the forward feature of the electromagnetic signal sample of the previous step;

[0044] The current step loss of the multivariate time series interpolation diffusion model is determined according to the first current step loss and the second current step loss.

[0045] In the above technical scheme, the first current step loss is determined according to the current step reverse noise and the current step forward noise, the second current step loss is determined according to the previous step electromagnetic signal sample reverse feature and the previous step electromagnetic signal sample forward feature, the current step loss of the multivariate time series interpolation diffusion model is determined according to the first current step loss and the second current step loss, the parameters of the multivariate time series interpolation diffusion model are optimized based on the current step loss, the accuracy and stability of the model in predicting the missing values ​​of the electromagnetic signal are improved, the accuracy and stability of the electromagnetic signal correction are improved, and object detection is performed based on the corrected electromagnetic signal, thereby improving the accuracy and anti-interference of object detection.

[0046] According to one embodiment of the present application, the error estimation model includes:

[0047] a step embedding module, for determining an embedding vector according to the current time step, so that the multivariate time series interpolation diffusion model determines the current time step;

[0048] A time embedding module, used to add time dimension information to the electromagnetic signal sample features;

[0049] An attention mechanism estimator is used to perform high-dimensional computation processing on the outputs of the step embedding module and the time embedding module according to the attention mechanism.

[0050] In the above technical scheme, the error estimation model includes a step embedding module, a time embedding module and an attention mechanism estimator. The step embedding module is used to determine the embedding vector according to the current time step so that the multivariate time series interpolation diffusion model can determine the current time step. The time embedding module is used to add time dimension information to the electromagnetic signal sample characteristics. The attention mechanism estimator is used to perform high-dimensional calculation processing on the outputs of the step embedding module and the time embedding module according to the attention mechanism. By maintaining the learnable embedding vector and adding time dimension information, the error estimation model can more accurately capture the characteristic differences of the electromagnetic signal time series at different time steps, improve the accuracy of error estimation, and thus improve the training effect of the multivariate time series interpolation diffusion model, improve the accuracy and stability of electromagnetic signal correction, and perform object detection based on the corrected electromagnetic signal, thereby improving the accuracy and anti-interference of object detection.

[0051] In a second aspect, the present application provides an electromagnetic signal identification device, the device comprising:

[0052] An acquisition unit, used for acquiring electromagnetic signals of each sub-area to be detected;

[0053] A correction unit, used for correcting the electromagnetic signal of each sub-area to be detected to obtain a corrected electromagnetic signal of each sub-area to be detected;

[0054] The distinguishing unit is used to determine whether there is a target object in the area to be detected according to the calibrated electromagnetic signals of each sub-area to be detected, and the area to be detected is composed of the sub-areas to be detected.

[0055] In the above technical scheme, the object detection device is used to obtain the electromagnetic signal of each sub-area to be detected by scanning, correct the electromagnetic signal of each sub-area to be detected, and obtain the corrected electromagnetic signal of each sub-area to be detected, and then determine whether there is a target object in the area to be detected composed of the sub-areas to be detected based on the corrected electromagnetic signal of each sub-area to be detected. By dividing the sub-areas to be detected and correcting the electromagnetic signal, object detection is performed based on the corrected electromagnetic signal, thereby improving the accuracy and anti-interference ability of object detection.

[0056] In a third aspect, the present application provides a security gate comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the object detection method as described in the first aspect above when executing the computer program.

[0057] In a fourth aspect, the present application provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the object detection method as described in the first aspect above.

[0058] In a fifth aspect, the present application provides a chip, comprising a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement the object detection method as described in the first aspect above.

[0059] In a sixth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the object detection method as described in the first aspect above.

[0060] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0062] Figure 1 is a flowchart of an object detection method provided in some embodiments of the present application;

[0063] Figure 2 is a schematic diagram of electromagnetic signal deviation provided by some embodiments of the present application;

[0064] Figure 3 is a schematic diagram of electromagnetic signal loss provided by some embodiments of the present application;

[0065] Figure 4 is a schematic diagram of a correction electromagnetic signal provided by some embodiments of the present application;

[0066] Figure 5 It is one of the flowcharts of multivariate time series interpolation diffusion model training provided by some embodiments of the present application;

[0067] Figure 6 This is the second flowchart of multivariate time series interpolation diffusion model training provided by some embodiments of the present application;

[0068] Figure 7 is a schematic diagram of an error estimation model provided by some embodiments of the present application;

[0069] Figure 8 is a schematic diagram of the structure of an object detection device provided in some embodiments of the present application;

[0070] Fig. 9 It is a schematic diagram of the structure of the security door provided in some embodiments of the present application.

[0071] Description of reference numerals:

[0072] 800: object detection device; 801: acquisition unit; 802: correction unit; 803: determination unit;

[0073] 900: security door; 901: processor; 902: memory. DETAILED DESCRIPTION

[0074] The following will be combined with the drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of this application.

[0075] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0076] The object detection method, device and security door provided in the embodiments of the present application are described in detail below with reference to the accompanying drawings through specific embodiments and their application scenarios.

[0077] The object detection method provided in the embodiment of the present application can be executed by an electronic device or a functional module or functional entity in the electronic device that can implement the object detection method. The electronic devices mentioned in the embodiment of the present application include but are not limited to security doors or security inspection machines, etc. The object detection method provided in the embodiment of the present application is described below using the electronic device as an example of the execution subject.

[0078] Figure 1 is a flow chart of an object detection method provided by some embodiments of the present application. Figure 1 As shown, the object detection method includes: step 110, step 120 and step 130.

[0079] Step 110: Acquire electromagnetic signals of each sub-area to be detected by scanning.

[0080] It can be understood that by scanning each sub-area to be detected through a detection device (such as a security gate), the electromagnetic signal obtained can be converted into a relatively stable sine wave through a fixed-frequency alternating voltage and signal orthogonal demodulation. The signal stability of this sine wave at different frequencies is different and can be converted into (x, y) coordinates in an XY two-dimensional rectangular coordinate system. However, since the electromagnetic signal is easily affected by the surrounding environment, numerical offset or missing values ​​may occur. Figure 2 Schematic diagram of electromagnetic signal deviation provided by some embodiments of the present application. Figure 2 As shown, it can be seen from the fourth quadrant of the coordinate axis that the electromagnetic signal has undergone obvious deviation and signal interference. Figure 3 Schematic diagram of electromagnetic signal loss provided by some embodiments of the present application. Figure 3 As shown, it can be seen from the second quadrant of the coordinate axis that the electromagnetic signal has data missing and cannot form a complete and continuous curve.

[0081] Step 120: calibrate the electromagnetic signal of each sub-region to be detected to obtain a calibrated electromagnetic signal of each sub-region to be detected.

[0082] It can be understood that the purpose of correcting the electromagnetic signal of each sub-region to be detected is to complete and fit the electromagnetic signal data to solve the above-mentioned possible problems in the electromagnetic signal of each sub-region to be detected. Figure 2 and Figure 3 The numerical deviation or missing value shown in the figure can be solved by obtaining a calibrated and restored electromagnetic signal, that is, a corrected electromagnetic signal of each sub-area to be detected. Object detection is performed based on the corrected electromagnetic signal, which helps to improve the accuracy and anti-interference of object detection.

[0083] Step 130: Determine whether there is a target object in the area to be detected based on the calibrated electromagnetic signals of the sub-areas to be detected, where the area to be detected is composed of the sub-areas to be detected.

[0084] It can be understood that by dividing the area to be detected into sub-areas to be detected, calibrating and detecting each sub-area to be detected respectively, and determining the detection result of the area to be detected by combining the detection results of each sub-area to be detected, the risk of false detection and missed detection is reduced, which helps to improve the accuracy of the object detection method.

[0085] In some embodiments, the target items include at least one of the following: security restricted items; electronic devices; metal items; controlled items. Among them, security restricted items refer to items that are restricted to be carried when passing through security checks; electronic devices include mobile phones, tablet computers, laptops, smart wearable devices, mobile power supplies, Bluetooth headsets, radio equipment, etc.; metal items refer to materials, components or products made of metal, such as tools and equipment made of metal, metal jewelry and accessories, metal daily necessities, metal components in electronic devices such as mobile phones, computers, headphones, or metal materials; controlled items include knives, weapons, etc.

[0086] It should be noted that the present application does not limit the specific types of target objects, which can be determined according to the needs of specific application scenarios. For example, in an invigilation scenario, based on the calibrated electromagnetic signals of each sub-area to be detected, it is determined whether there are electronic devices in the area to be detected to achieve security inspection.

[0087] In the above technical scheme, the electromagnetic signals of each sub-area to be detected are acquired by scanning, and the electromagnetic signals of each sub-area to be detected are corrected to obtain the corrected electromagnetic signals of each sub-area to be detected. Then, based on the corrected electromagnetic signals of each sub-area to be detected, it is determined whether there is a target object in the area to be detected composed of the sub-areas to be detected. By dividing the sub-areas to be detected and correcting the electromagnetic signals, object detection is performed based on the corrected electromagnetic signals, thereby improving the accuracy and anti-interference ability of object detection.

[0088] In some embodiments of the present application, determining whether there is a target object in the area to be detected according to the calibrated electromagnetic signals of each sub-area to be detected includes:

[0089] Screening the corrected electromagnetic signals of each sub-area to be detected according to the allowable value, and retaining the stable electromagnetic signals of each sub-area to be detected;

[0090] Whether there is a target object in the to-be-detected area is determined according to the characteristic values ​​of the stable electromagnetic signals of each to-be-detected sub-area and its adjacent to-be-detected sub-area and the threshold range corresponding to the target object.

[0091] It is understandable that the allowable value is set according to the actual application scenario and detection requirements, and represents an acceptable range of electromagnetic signal fluctuations. Even if the electromagnetic signals of each sub-area to be detected should be stable after correction, there may still be situations where the allowable value is exceeded. These corrected electromagnetic signals that exceed the allowable value are screened and eliminated, making the electromagnetic signals used for object detection more stable, further improving the accuracy and anti-interference of object detection.

[0092] The characteristic value of the stable electromagnetic signal of each sub-area to be detected may be a phase characteristic value or a frequency characteristic value, etc. The threshold range corresponding to the target object is a pre-set threshold range determined based on the characteristic value of the electromagnetic signal obtained by scanning the target object. If the characteristic value of the stable electromagnetic signal of the sub-area to be detected is within the threshold range corresponding to the target object, it can be considered that the target object exists in the current sub-area to be detected.

[0093] It is understandable that different parts of the target object may be distributed in different sub-areas to be detected, so the detection of the object based on the electromagnetic signal of the sub-area to be detected is not limited to a single sub-area to be detected, but may also span to adjacent sub-areas to be detected. Therefore, combining the characteristic values ​​of the stable electromagnetic signals of each sub-area to be detected and its adjacent sub-area to be detected helps to identify situations that are not obvious in a single sub-area to be detected, but when multiple sub-areas to be detected are combined, they can meet the threshold range corresponding to the target object. In addition, by combining the characteristic values ​​of the stable electromagnetic signals of each sub-area to be detected and its adjacent sub-area to be detected, it is possible to reduce misjudgments caused by abnormal electromagnetic signals or interference in a single sub-area to be detected, which also helps to improve the accuracy and reliability of object detection.

[0094] In the above technical scheme, the corrected electromagnetic signals of each sub-area to be detected are screened according to the allowed value, and the stable electromagnetic signals of each sub-area to be detected are retained. According to the characteristic values ​​of the stable electromagnetic signals of each sub-area to be detected and its adjacent sub-areas to be detected and the threshold range corresponding to the target object, it is determined whether there is a target object in the area to be detected, thereby improving the accuracy and anti-interference of object detection.

[0095] In some embodiments of the present application, the calibrating the electromagnetic signal of each sub-region to be detected to obtain the calibrated electromagnetic signal of each sub-region to be detected includes:

[0096] Inputting the electromagnetic signals of each sub-region to be detected into a multivariate time series interpolation diffusion model to obtain a corrected electromagnetic signal of each sub-region to be detected;

[0097] The multivariate time series interpolation diffusion model is obtained based on electromagnetic signal sample training.

[0098] Figure 4Schematic diagram of correcting electromagnetic signals provided by some embodiments of the present application. Figure 4 As shown, the corrected electromagnetic signal is the electromagnetic signal obtained by solving the electromagnetic signal offset or electromagnetic signal missing problem in the electromagnetic signal. The electromagnetic signal sample refers to the electromagnetic signal set used to train the multivariate time series interpolation diffusion model.

[0099] Time series interpolation refers to the process of filling missing values ​​in time series data. Multivariate time series interpolation refers to focusing on time series data containing multiple dimensions or variables. The diffusion model gradually injects noise into the data and applies a learnable inverse process to restore the data until the original clear data is obtained.

[0100] In the above technical scheme, the electromagnetic signal of each sub-area to be detected is input into a multivariate time series interpolation diffusion model obtained by training based on electromagnetic signal samples, so as to obtain a corrected electromagnetic signal of each sub-area to be detected, which solves the problem of electromagnetic signal defects and realizes the correction of electromagnetic signals of each sub-area to be detected. It is applied to object detection to improve the accuracy and anti-interference of object detection.

[0101] Figure 5 This is one of the flow charts of multivariate time series interpolation diffusion model training provided by some embodiments of the present application. Figure 5 As shown, the multivariate time series interpolation diffusion model is trained through the following steps, including: step 510, step 520, step 530, step 540 and step 550.

[0102] Step 510: extract features from the electromagnetic signal sample to obtain electromagnetic signal sample features.

[0103] Figure 6 This is the second flow chart of multivariate time series interpolation diffusion model training provided by some embodiments of the present application. Figure 6 As shown, during the model training process, the electromagnetic signal samples are feature extracted to obtain the electromagnetic signal sample features, which include the electromagnetic signal sample observation value features. And the missing value characteristics of electromagnetic signal samples t=0 means that the diffusion process has not yet started. and They correspond to the observed value characteristics and missing value characteristics of the electromagnetic signal sample in the initial state respectively.

[0104] Step 520: Perform frequency division processing on the electromagnetic signal sample characteristics to determine the main frequency signal characteristics of the electromagnetic signal sample and the non-main frequency signal characteristics of the electromagnetic signal sample;

[0105] Step 530: determining a joint Gaussian distribution according to the non-main frequency signal characteristics of the electromagnetic signal sample;

[0106] In an embodiment of the present application, a conditional distribution is generated from a potential representation of the observed value, and noise sampling is performed based on this distribution to explicitly retain the intrinsic correlation between the observed data and the missing data. In each step of noise addition, a joint Gaussian distribution with different covariance matrices is used for noise sampling. The sampling source of the noise is the characteristic values ​​of electromagnetic signal samples of other interference frequencies other than the main frequency, so that the added noise can be closer to the noise characteristics in the actual application scenario; based on the non-main frequency signal characteristics of the electromagnetic signal sample, the parameters of the joint Gaussian distribution, including the mean and the covariance matrix, are determined to match the statistical characteristics of the noise with the actual interference frequency characteristics.

[0107] It can be understood that a joint Gaussian distribution model based on observed values ​​and missing values ​​is constructed in the model. Assuming that the electromagnetic signal observation value x c and missing values ​​x m They all follow a joint Gaussian distribution P(x), that is, (x c ,x m )~P(x), by accurately estimating the joint distribution parameters, including calculating the infimum of P(x), it is possible to use known observations to infer missing values, thereby achieving effective supplementation and recovery of the multi-dimensional electromagnetic signal.

[0108] Step 540: Perform forward diffusion processing on the electromagnetic signal sample features according to the joint Gaussian distribution to obtain forward diffusion data of the electromagnetic signal sample features, wherein the forward diffusion data of the electromagnetic signal sample features includes the noisy electromagnetic signal sample features.

[0109] It can be understood that, according to the joint Gaussian distribution, forward diffusion processing is performed on the electromagnetic signal sample features, which means adding noise sampled from the joint Gaussian distribution in each time step. As the diffusion steps increase, the noise level of the electromagnetic signal sample features gradually deepens, and the noisy electromagnetic signal sample features are obtained. The local features of the noisy electromagnetic signal sample features are covered by the added noise, which can be further used to gradually remove the noise in the reverse diffusion processing process.

[0110] The forward diffusion data of the electromagnetic signal sample features refers to a series of data generated in the process of forward diffusion of the electromagnetic signal sample features, including the noise sampled from the joint Gaussian distribution added at each time step (called forward noise), and also includes the intermediate state of the electromagnetic signal sample features after the noise is added at each time step (called the electromagnetic signal sample forward feature).

[0111] In some embodiments, adding a standard Gaussian-based noise value to the observations at step T can be expressed as a Markov chain:

[0112]

[0113] Among them, q(X 1:T |X0) means that under the given initial state X0, the entire time series X 1:T The joint probability distribution of q(X t |X t-1 ) means that given the previous state X t-1 Under the condition, the current state is X t The conditional probability of ; ∏ represents the product, that is, the continuous multiplication of all items;

[0114] Establish a conditional distribution of the approximation error given a given observation, based on (x c ,x m )~P(x) can be approximately expressed as

[0115]

[0116] in, Indicates that at a given observation Missing value The probability of occurrence, t = 0 represents the initial value, represents an expected value, where q is an approximate distribution used to estimate the true distribution p. The expected value is calculated under the q distribution and consists of the sum of two log-likelihood ratios: Represents the entire time series X 0:T The probability and conditional probability q(X 1:T |X0), Indicates that under given conditions Lower sequence The probability of The logarithm of the ratio of

[0117] By derivation, we can finally get the conditional distribution of the approximate error:

[0118]

[0119] Where D KL Expressed as KL divergence calculation, E q Expressed as expected value calculation, where p θ is a probability filling model that aims to approximate the true conditional distribution P(X m |X c ).

[0120] Step 550: perform reverse diffusion processing on the noisy electromagnetic signal sample features according to the forward diffusion data of the electromagnetic signal sample features. In the reverse diffusion processing, determine the loss of the multivariate time series interpolation diffusion model, and optimize the parameters of the multivariate time series interpolation diffusion model based on the loss.

[0121] It can be understood that the reverse diffusion processing starts from the noisy electromagnetic signal sample characteristics at the end of the forward diffusion process. In each time step, the noise added in the forward diffusion process is predicted, and the predicted noise (called reverse noise) is removed from the current noisy data to obtain the intermediate state of the electromagnetic signal sample characteristics after noise removal at each time step (called the reverse characteristics of the electromagnetic signal samples).

[0122] In the reverse diffusion process, after removing the noise at each time step, the difference between the reverse noise and the forward noise can be calculated through the loss function to quantify the accuracy of the model prediction. Based on the evaluation result of the loss function, the parameters of the multivariate time series interpolation diffusion model are optimized through gradient descent or other optimization algorithms to minimize the loss function, thereby improving the prediction performance of the model.

[0123] In the above technical scheme, feature extraction is performed on the electromagnetic signal sample to obtain the electromagnetic signal sample feature, and the electromagnetic signal sample feature is subjected to frequency division processing to determine the main frequency signal feature and non-main frequency signal feature of the electromagnetic signal sample. According to the non-main frequency signal feature of the electromagnetic signal sample, a joint Gaussian distribution is determined. According to the joint Gaussian distribution, the electromagnetic signal sample feature is subjected to forward diffusion processing to obtain forward diffusion data of the electromagnetic signal sample feature. According to the forward diffusion data of the electromagnetic signal sample feature, the noisy electromagnetic signal sample feature is subjected to reverse diffusion processing. In the reverse diffusion processing, the loss of the multivariate time series interpolation diffusion model is determined, and the parameters of the model are optimized based on the loss, thereby realizing the training of the multivariate time series interpolation diffusion model, improving the accuracy and stability of the model in predicting the missing values ​​of the electromagnetic signal, and thereby improving the accuracy and stability of the electromagnetic signal correction. Object detection is performed based on the corrected electromagnetic signal, which can improve the accuracy and anti-interference of object detection.

[0124] In some embodiments of the present application, performing forward diffusion processing on the electromagnetic signal sample features according to the joint Gaussian distribution to obtain forward diffusion data of the electromagnetic signal sample features includes:

[0125] According to the current time step and the preset time step, the joint Gaussian distribution is sampled to determine the current step forward noise;

[0126] Adding the current step forward noise to the previous step electromagnetic signal sample forward feature to obtain the current step electromagnetic signal sample forward feature;

[0127] Gradually increase the current time step and repeat the above steps until the current time step is equal to the preset time step, and determine that the positive feature of the electromagnetic signal sample at the current time step is the noisy electromagnetic signal sample feature;

[0128] The forward diffusion data of the electromagnetic signal sample characteristics also includes each current step forward noise and each current step electromagnetic signal sample forward characteristics.

[0129] It is understandable that the current time step t refers to the step number of the forward diffusion currently in progress. Each time forward noise is added, the current time step is controlled to increase by 1, and this process will continue until the current time step is equal to the preset time step t=T. The preset time step T refers to the total number of times forward noise needs to be added during the entire forward noise addition process. After T times of forward noise addition, the noisy electromagnetic signal sample characteristics are obtained. It is understandable that the local characteristics of the noisy electromagnetic signal sample characteristics are covered by the added forward noise, which can be further used to gradually remove noise during the reverse diffusion process to complete and fit the electromagnetic signal samples, thereby solving problems such as numerical offset or missing values ​​that may exist in the electromagnetic signal samples.

[0130] The forward feature of the electromagnetic signal sample refers to the intermediate state feature after the electromagnetic signal sample feature gradually adds noise during the forward diffusion process. At the current time step, the forward noise of the current step is added to the forward feature of the electromagnetic signal sample in the previous step to obtain the forward feature of the electromagnetic signal sample in the current step.

[0131] The forward diffusion data of the electromagnetic signal sample feature includes the forward noise added at each time step and the electromagnetic signal sample forward feature generated thereby.

[0132] like Figure 6 As shown, taking the current time step as T as an example, the current time step is equal to the preset time step t=T, that is, the current step is the last time to add the current step forward noise, sample the joint Gaussian distribution (Joint Gaussian), determine the current step forward noise (not shown in the figure), and the positive feature of the electromagnetic signal sample in the previous step (observation value feature and missing value features ) Add the current step forward noise (Noising) to obtain the current step electromagnetic signal sample forward feature (observation value feature and missing value features ), as the sample feature of the noisy electromagnetic signal.

[0133] In some embodiments, when the current time step is 1, the electromagnetic signal sample feature is determined to be the positive feature of the electromagnetic signal sample in the previous step.

[0134] It can be understood that the current time step is 1, that is, the current step is the first time to add the current step forward noise. Since there is no positive feature of the electromagnetic signal sample in the previous step, the electromagnetic signal sample feature is determined to be the forward feature of the electromagnetic signal sample in the previous step, realizing the first noise addition to the electromagnetic signal sample feature in the forward diffusion processing process.

[0135] like Figure 6 As shown, the current time step is 1, the joint Gaussian distribution is sampled (Sampler), the current step forward noise (not shown in the figure) is determined, and the electromagnetic signal sample characteristics (observation value characteristics) are and missing value features ) Add the current step forward noise to obtain the current step electromagnetic signal sample forward feature (observation value feature and missing value features ).

[0136] In the above technical solution, when the current time step is 1, the electromagnetic signal sample feature is determined to be the forward feature of the electromagnetic signal sample in the previous step, thereby realizing the first noise addition to the electromagnetic signal sample feature during the forward diffusion process.

[0137] In the above technical scheme, the joint Gaussian distribution is sampled according to the current time step and the preset time step, the forward noise of the current step is determined, the forward noise of the current step is added to the forward feature of the electromagnetic signal sample of the previous step, and the forward feature of the electromagnetic signal sample of the current step is obtained. The current time step is gradually increased and the above steps are repeated until the current time step is equal to the preset time step. The forward feature of the electromagnetic signal sample of the current step is determined to be the noisy electromagnetic signal sample feature, and the forward diffusion processing of the electromagnetic signal sample feature is realized. The forward diffusion data of the electromagnetic signal sample feature is obtained, which can be used for reverse diffusion processing of the noisy electromagnetic signal sample feature, and the multivariate time series interpolation diffusion model is optimized in the reverse diffusion processing.

[0138] In some embodiments of the present application, the forward diffusion data of the electromagnetic signal sample features is used to perform reverse diffusion processing on the noisy electromagnetic signal sample features. In the reverse diffusion processing, the loss of the multivariate time series interpolation diffusion model is determined, and the parameters of the multivariate time series interpolation diffusion model are optimized based on the loss, including:

[0139] Determine the current step reverse noise according to the current time step and the reverse characteristics of the electromagnetic signal sample of the current step;

[0140] According to the reverse noise of the current step, the reverse feature of the electromagnetic signal sample of the current step is removed to determine the reverse feature of the electromagnetic signal sample of the previous step;

[0141] Determining the current step loss of the multivariate time series interpolation diffusion model through an error estimation model;

[0142] Optimizing the parameters of the multivariate time series interpolation diffusion model based on the current step loss;

[0143] Gradually reduce the current time step and repeat the above steps until the current time step is 1.

[0144] During the reverse diffusion process, the current time step gradually decreases from the preset time step T. The current time step t refers to the step number of the reverse diffusion currently being performed. Each time the reverse noise removal is performed, the current time step is controlled to decrease by 1, and this process will continue until the current time step is 1. The preset time step T refers to the total number of times the reverse noise needs to be removed during the entire reverse noise removal process. It can be considered that after T times of reverse noise removal, the noise in the noised electromagnetic signal sample features has been almost completely removed.

[0145] The reverse feature of the electromagnetic signal sample refers to the intermediate state feature after the noise of the noisy electromagnetic signal sample feature is gradually removed during the reverse diffusion process. At the current time step, the reverse noise of the current step is removed from the reverse feature of the electromagnetic signal sample to obtain the reverse feature of the previous step of the electromagnetic signal sample.

[0146] It can be understood that each time the reverse noise removal is performed, the loss of the current step will be confirmed through the error estimation model, and the parameters of the multivariate time series interpolation diffusion model will be optimized according to the loss. The optimized parameters will be applied to the next reverse noise denoising, so that the reverse characteristics of the electromagnetic signal samples at each step are as close as possible to the corresponding forward characteristics, so as to achieve fine adjustment of the model.

[0147] like Figure 6 As shown in the figure, taking the current time step as 1 as an example, that is, the current step is the last time to remove the reverse noise (Denoising), according to the reverse feature of the electromagnetic signal sample in the current step (missing value feature estimation ) and the current time step t = 1, determine the current step reverse noise (not shown in the figure), and according to the current step reverse noise, perform the reverse feature of the current step electromagnetic signal sample Remove noise and determine the reverse features of the electromagnetic signal sample in the previous step (missing value feature estimation ), as a multivariate time series interpolation diffusion model, the electromagnetic signal sample feature estimation is obtained by forward diffusion and reverse diffusion processing according to the electromagnetic signal sample features. The forward diffusion processing refers to the process of gradually adding noise to the time series data (i.e., the electromagnetic signal sample features), and the reverse diffusion refers to the process of gradually removing noise from the noisy electromagnetic signal sample features obtained by the forward diffusion processing, and finally obtaining the electromagnetic signal sample feature estimation. It can be understood that for a trained model, the electromagnetic signal sample feature estimation can realize data completion and fitting of the electromagnetic signal samples, and solve the problems of numerical offset or missing values ​​that may exist in the electromagnetic signal samples.

[0148] In some embodiments, when the current time step is equal to the preset time step, the noisy electromagnetic signal sample feature is determined to be the inverse feature of the current electromagnetic signal sample.

[0149] It can be understood that the current time step is equal to the preset time step t=T, that is, the current step is the first time to remove the reverse noise. Since there is no reverse feature of the electromagnetic signal sample of the current step, the noisy electromagnetic signal sample feature is determined to be the reverse feature of the electromagnetic signal sample of the current step, thereby realizing the first noise removal of the noisy electromagnetic signal sample feature in the reverse diffusion processing process.

[0150] like Figure 6 As shown, the current time step is T, according to the inverse characteristics of the electromagnetic signal sample at the current step (missing value feature estimation ) and the current time step t = T, determine the current step reverse noise (not shown in the figure), and according to the current step reverse noise, perform the reverse feature of the current step electromagnetic signal sample Remove noise and determine the reverse features of the electromagnetic signal sample in the previous step (missing value feature estimation ).

[0151] In the above technical solution, when the current time step is equal to the preset time step, the noisy electromagnetic signal sample characteristics are determined as the reverse characteristics of the current electromagnetic signal sample, thereby realizing the first noise removal of the noisy electromagnetic signal sample characteristics during the reverse diffusion processing.

[0152] In some embodiments, according to the reverse noise of the current step, removing noise from the reverse feature of the electromagnetic signal sample of the current step to determine the reverse feature of the electromagnetic signal sample of the previous step includes:

[0153] According to the reverse noise of the current step, the mean value of the reverse feature of the electromagnetic signal sample of the previous step is determined;

[0154] Randomly sample the positive features (observation value features and missing value features) of the electromagnetic signal sample in the previous step to obtain the sampling noise of the current step;

[0155] The reverse feature of the electromagnetic signal sample in the previous step is determined (missing value feature estimation) according to the mean value of the reverse feature of the electromagnetic signal sample in the previous step and the sampling noise in the current step.

[0156] In the above technical scheme, the current step reverse noise is determined according to the current time step and the reverse characteristics of the current step electromagnetic signal sample, the previous step electromagnetic signal sample reverse characteristics are determined according to the reverse characteristics of the current step electromagnetic signal sample and the current step reverse noise, the current step loss of the multivariate time series interpolation diffusion model is determined through the error estimation model, the parameters of the multivariate time series interpolation diffusion model are optimized based on the current step loss, the current time step is gradually reduced and the above steps are repeated until the current time step is 1, the reverse diffusion processing of the noisy electromagnetic signal sample characteristics is realized, and the multivariate time series interpolation diffusion model is optimized in the reverse diffusion processing, the accuracy and stability of the model in predicting the missing values ​​of the electromagnetic signal are improved, and then the accuracy and stability of the electromagnetic signal correction are improved, and the object detection is performed based on the corrected electromagnetic signal, which can improve the accuracy and anti-interference of the object detection.

[0157] In some embodiments of the present application, determining the current step loss of the multivariate time series interpolation diffusion model through an error estimation model includes:

[0158] determining a first current step loss according to the current step reverse noise and the current step forward noise;

[0159] Determining a second current step loss according to the reverse feature of the electromagnetic signal sample of the previous step and the forward feature of the electromagnetic signal sample of the previous step;

[0160] The current step loss of the multivariate time series interpolation diffusion model is determined according to the first current step loss and the second current step loss.

[0161] According to the current step reverse noise and the current step forward noise, the first current step loss is determined. The first current loss can measure the accuracy of the model's prediction of noise in the reverse process, that is, the difference between the reverse noise predicted by the model and the actual added forward noise; according to the previous step electromagnetic signal sample reverse feature and the previous step electromagnetic signal sample forward feature, the second current step loss is determined. The second current step loss can measure the accuracy of the model's prediction of features in the reverse process, that is, the difference between the reverse feature predicted by the model and the actual forward feature.

[0162] In the reverse diffusion process, each time the reverse noise is removed, the first current step loss and the second current step loss of the current step are confirmed through the error estimation model, and the parameters of the multivariate time series interpolation diffusion model are optimized according to the loss, and the optimized parameters are applied to the next reverse noise removal, so that the reverse characteristics of the electromagnetic signal samples in each step are as close as possible to the corresponding forward characteristics, so as to achieve fine adjustment of the model. It can be understood that the error estimation model is mainly used to ensure the correct convergence direction in the reverse diffusion process to avoid ineffective convergence due to excessive errors.

[0163] In the above technical scheme, the first current step loss is determined according to the current step reverse noise and the current step forward noise, the second current step loss is determined according to the previous step electromagnetic signal sample reverse feature and the previous step electromagnetic signal sample forward feature, the current step loss of the multivariate time series interpolation diffusion model is determined according to the first current step loss and the second current step loss, the parameters of the multivariate time series interpolation diffusion model are optimized based on the current step loss, the accuracy and stability of the model in predicting the missing values ​​of the electromagnetic signal are improved, the accuracy and stability of the electromagnetic signal correction are improved, and object detection is performed based on the corrected electromagnetic signal, thereby improving the accuracy and anti-interference of object detection.

[0164] In one embodiment of the present application, the error estimation model includes:

[0165] a step embedding module, for determining an embedding vector according to the current time step, so that the multivariate time series interpolation diffusion model determines the current time step;

[0166] A time embedding module, used to add time dimension information to the electromagnetic signal sample features;

[0167] An attention mechanism estimator is used to perform high-dimensional computation processing on the outputs of the step embedding module and the time embedding module according to the attention mechanism.

[0168] Figure 7 Schematic diagram of error estimation model provided by some embodiments of the present application. Figure 7 As shown in the figure, the error estimation model includes: input layer (Input), linear transformation layer (Linear), step embedding module (Step Embed), time embedding module (Temporal Embed), attention mechanism estimator (Self-Attention) and addition and normalization layer (Add&Norm).

[0169] It is understandable that, since there are large differences between the reverse features of electromagnetic signal samples and the forward features of electromagnetic signal samples at different time steps, the error estimation model needs to obtain the current time step t for accurate error estimation. The step embedding module maintains T learnable N-dimensional embedding vectors for all steps, that is, when the current time step t is given, the step embedding module outputs an N-dimensional vector.

[0170] It can be understood that the electromagnetic signal time series input can be regarded as a sequential information. The time embedding module is used to add the time dimension to the sequential information so that similar magnetic field signals before and after the sequential information can learn from each other and refer to each other, effectively making up for the missing data. Optionally, the fixed position encoding (Position Embedding) is given for compilation in a one-hot encoding manner, where the fixed encoding is calculated as follows:

[0171] TE(t,2i)=sin(t / 10000 2i / d )

[0172] TE(t,2i+1)=cos(t / 10000 2i / d )

[0173] Among them, t is the time step, i is the dimension, and TE(T,2I) represents the position encoding.

[0174] It can be understood that the attention mechanism estimator uses the attention mechanism to model the output data of the step embedding module and the temporal embedding module and perform high-dimensional calculation processing, so that the error estimation model can autonomously learn to fit the data through deep learning. In some embodiments, Z represents the vector layer of the attention layer, where SE represents Step Embedding and TE represents Temporal Embedding, which is expressed as:

[0175] Z0=X T +SE+TE

[0176] After mapping Z0 to a length of 512 dimensions, the self-attention mechanism is introduced for calculation, which is expressed as:

[0177]

[0178] Among them, d model Indicates the size of the dimension, currently 512, Z0 is the input mapping vector, W Q To generate the weight matrix of the query vector, W V To generate the weight matrix of the value vector, the softmax function is used to normalize the attention scores, and Att(Z0) refers to the output of the self-attention mechanism.

[0179] In the above technical scheme, the error estimation model includes a step embedding module, a time embedding module and an attention mechanism estimator. The step embedding module is used to determine the embedding vector according to the current time step so that the multivariate time series interpolation diffusion model can determine the current time step. The time embedding module is used to add time dimension information to the electromagnetic signal sample characteristics. The attention mechanism estimator is used to perform high-dimensional calculation processing on the outputs of the step embedding module and the time embedding module according to the attention mechanism. By maintaining the learnable embedding vector and adding time dimension information, the error estimation model can more accurately capture the characteristic differences of the electromagnetic signal time series at different time steps, improve the accuracy of error estimation, and thus improve the training effect of the multivariate time series interpolation diffusion model, improve the accuracy and stability of electromagnetic signal correction, and perform object detection based on the corrected electromagnetic signal, thereby improving the accuracy and anti-interference of object detection.

[0180] The object detection method provided in the embodiment of the present application can be executed by an object detection device. In the embodiment of the present application, an object detection device executing the object detection method is taken as an example to illustrate the object detection device provided in the embodiment of the present application.

[0181] Figure 8 Schematic diagram of the structure of the object detection device provided in some embodiments of the present application. Figure 8 As shown, the object detection device 800 includes:

[0182] An acquisition unit 801 is used to acquire electromagnetic signals of each sub-region to be detected;

[0183] A correction unit 802 is used to correct the electromagnetic signal of each sub-region to be detected to obtain a corrected electromagnetic signal of each sub-region to be detected;

[0184] The determination unit 803 is used to determine whether there is a target object in the area to be detected according to the calibrated electromagnetic signals of the sub-areas to be detected, and the area to be detected is composed of the sub-areas to be detected.

[0185] Optionally, the determination unit 803 is used to:

[0186] Screening the corrected electromagnetic signals of each sub-area to be detected according to the allowable value, and retaining the stable electromagnetic signals of each sub-area to be detected;

[0187] Whether there is a target object in the to-be-detected area is determined according to the characteristic values ​​of the stable electromagnetic signals of each to-be-detected sub-area and its adjacent to-be-detected sub-area and the threshold range corresponding to the target object.

[0188] Optionally, the correction unit 802 is used for:

[0189] Inputting the electromagnetic signals of each sub-region to be detected into a multivariate time series interpolation diffusion model to obtain a corrected electromagnetic signal of each sub-region to be detected;

[0190] The multivariate time series interpolation diffusion model is obtained based on electromagnetic signal sample training.

[0191] Optionally, the multivariate time series interpolation diffusion model is trained by the following steps:

[0192] Extracting features from the electromagnetic signal sample to obtain features of the electromagnetic signal sample;

[0193] Performing frequency division processing on the electromagnetic signal sample characteristics to determine the main frequency signal characteristics of the electromagnetic signal sample and the non-main frequency signal characteristics of the electromagnetic signal sample;

[0194] Determining a joint Gaussian distribution according to the non-main frequency signal characteristics of the electromagnetic signal sample;

[0195] According to the joint Gaussian distribution, forward diffusion processing is performed on the electromagnetic signal sample features to obtain forward diffusion data of the electromagnetic signal sample features, wherein the forward diffusion data of the electromagnetic signal sample features includes the noisy electromagnetic signal sample features;

[0196] According to the forward diffusion data of the electromagnetic signal sample characteristics, the noisy electromagnetic signal sample characteristics are subjected to reverse diffusion processing. In the reverse diffusion processing, the loss of the multivariate time series interpolation diffusion model is determined, and the parameters of the multivariate time series interpolation diffusion model are optimized based on the loss.

[0197] Optionally, performing forward diffusion processing on the electromagnetic signal sample features according to the joint Gaussian distribution to obtain forward diffusion data of the electromagnetic signal sample features includes:

[0198] According to the current time step and the preset time step, the joint Gaussian distribution is sampled to determine the current step forward noise;

[0199] Adding the current step forward noise to the previous step electromagnetic signal sample forward feature to obtain the current step electromagnetic signal sample forward feature;

[0200] Gradually increase the current time step and repeat the above steps until the current time step is equal to the preset time step, and determine that the positive feature of the electromagnetic signal sample at the current time step is the noisy electromagnetic signal sample feature;

[0201] The forward diffusion data of the electromagnetic signal sample characteristics also includes each current step forward noise and each current step electromagnetic signal sample forward characteristics.

[0202] Optionally, when the current time step is 1, the electromagnetic signal sample feature is determined to be the positive feature of the electromagnetic signal sample in the previous step.

[0203] Optionally, the forward diffusion data of the electromagnetic signal sample features is used to perform reverse diffusion processing on the noisy electromagnetic signal sample features, and in the reverse diffusion processing, the loss of the multivariate time series interpolation diffusion model is determined, and the parameters of the multivariate time series interpolation diffusion model are optimized based on the loss, including:

[0204] Determine the current step reverse noise according to the current time step and the reverse characteristics of the electromagnetic signal sample of the current step;

[0205] According to the reverse noise of the current step, the reverse feature of the electromagnetic signal sample of the current step is removed to determine the reverse feature of the electromagnetic signal sample of the previous step;

[0206] Determining the current step loss of the multivariate time series interpolation diffusion model through an error estimation model;

[0207] Optimizing the parameters of the multivariate time series interpolation diffusion model based on the current step loss;

[0208] Gradually reduce the current time step and repeat the above steps until the current time step is 1.

[0209] Optionally, when the current time step is equal to the preset time step, the noisy electromagnetic signal sample feature is determined to be the inverse feature of the current step electromagnetic signal sample.

[0210] Optionally, determining the current step loss of the multivariate time series interpolation diffusion model through an error estimation model includes:

[0211] determining a first current step loss according to the current step reverse noise and the current step forward noise;

[0212] Determining a second current step loss according to the reverse feature of the electromagnetic signal sample of the previous step and the forward feature of the electromagnetic signal sample of the previous step;

[0213] The current step loss of the multivariate time series interpolation diffusion model is determined according to the first current step loss and the second current step loss.

[0214] Optionally, the error estimation model includes:

[0215] a step embedding module, for determining an embedding vector according to the current time step, so that the multivariate time series interpolation diffusion model determines the current time step;

[0216] A time embedding module, used to add time dimension information to the electromagnetic signal sample features;

[0217] An attention mechanism estimator is used to perform high-dimensional computation processing on the outputs of the step embedding module and the time embedding module according to the attention mechanism.

[0218] In the above technical scheme, the object detection device is used to obtain the electromagnetic signal of each sub-area to be detected by scanning, correct the electromagnetic signal of each sub-area to be detected, and obtain the corrected electromagnetic signal of each sub-area to be detected, and then determine whether there is a target object in the area to be detected composed of the sub-areas to be detected based on the corrected electromagnetic signal of each sub-area to be detected. By dividing the sub-areas to be detected and correcting the electromagnetic signal, object detection is performed based on the corrected electromagnetic signal, thereby improving the accuracy and anti-interference ability of object detection.

[0219] The article detection device in the embodiment of the present application can be an electronic device, or a component in the electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal, or other devices other than the terminal. Exemplary, the electronic device can be a mobile phone, a tablet computer, a laptop computer, a PDA, a car electronic device, a mobile Internet device (Mobile Internet Device, MID), an augmented reality (augmented reality, AR) / virtual reality (virtualreality, VR) device, a robot, a wearable device, an ultra-mobile personal computer (ultra-mobile personalcomputer, UMPC), a netbook or a personal digital assistant (personal digital assistant, PDA), etc., and can also be a server, a network attached storage (Network Attached Storage, NAS), a personal computer (personalcomputer, PC), a television (television, TV), a teller machine or a self-service machine, etc., and the embodiment of the present application is not specifically limited.

[0220] The object detection device in the embodiment of the present application may be a device having an operating system. The operating system may be a Microsoft (Windows) operating system, an Android (Android) operating system, an IOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present application.

[0221] The object detection device provided in the embodiment of the present application can implement each process implemented in the embodiment of the object detection method, and will not be described again here to avoid repetition.

[0222] In some embodiments, Fig. 9 As shown, the embodiment of the present application also provides a security door 900, including a processor 901, a memory 902, and a computer program stored in the memory 902 and executable on the processor 901. When the program is executed by the processor 901, each process of the above-mentioned article detection method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be described here.

[0223] An embodiment of the present application also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the various processes of the above-mentioned object detection method embodiment are implemented and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0224] The processor is the processor in the security door described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk.

[0225] An embodiment of the present application also provides a computer program product, including a computer program, which implements each process of the above-mentioned object detection method embodiment when executed by a processor.

[0226] The processor is the processor in the security door described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk.

[0227] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned object detection method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0228] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0229] It should be noted that, in this article, the term "comprises", "includes" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other equivalent elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the method and device in the embodiment of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in the opposite order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0230] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the relevant technology, can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0231] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

[0232] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to an equivalent embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0233] Although the embodiments of the present application have been shown and described, those skilled in the art will appreciate that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present application, and that the scope of the present application is defined by the claims and their equivalents.

Claims

1. A method for detecting an object, characterized in that: include: Acquire electromagnetic signals of each sub-area to be detected by scanning; Correcting the electromagnetic signals of the sub-regions to be detected to obtain corrected electromagnetic signals of the sub-regions to be detected; It is determined whether a target object exists in the area to be detected according to the corrected electromagnetic signals of the sub-areas to be detected, and the area to be detected is composed of the sub-areas to be detected.

2. The object detection method according to claim 1, characterized in that: The determining whether there is a target object in the area to be detected according to the corrected electromagnetic signals of each sub-area to be detected includes: Screening the corrected electromagnetic signals of each sub-area to be detected according to the allowable value, and retaining the stable electromagnetic signals of each sub-area to be detected; Whether there is a target object in the to-be-detected area is determined according to the characteristic values ​​of the stable electromagnetic signals of each to-be-detected sub-area and its adjacent to-be-detected sub-area and the threshold range corresponding to the target object.

3. The object detection method according to claim 1, characterized in that: The step of correcting the electromagnetic signals of the sub-regions to be detected to obtain the corrected electromagnetic signals of the sub-regions to be detected includes: Inputting the electromagnetic signals of each sub-region to be detected into a multivariate time series interpolation diffusion model to obtain a corrected electromagnetic signal of each sub-region to be detected; The multivariate time series interpolation diffusion model is obtained based on electromagnetic signal sample training.

4. The object detection method according to claim 3, characterized in that: The multivariate time series interpolation diffusion model is trained by the following steps: Extracting features from the electromagnetic signal sample to obtain features of the electromagnetic signal sample; Performing frequency division processing on the electromagnetic signal sample characteristics to determine the main frequency signal characteristics of the electromagnetic signal sample and the non-main frequency signal characteristics of the electromagnetic signal sample; Determining a joint Gaussian distribution according to the non-main frequency signal characteristics of the electromagnetic signal sample; According to the joint Gaussian distribution, forward diffusion processing is performed on the electromagnetic signal sample features to obtain forward diffusion data of the electromagnetic signal sample features, wherein the forward diffusion data of the electromagnetic signal sample features includes the noisy electromagnetic signal sample features; According to the forward diffusion data of the electromagnetic signal sample characteristics, the noisy electromagnetic signal sample characteristics are subjected to reverse diffusion processing. In the reverse diffusion processing, the loss of the multivariate time series interpolation diffusion model is determined, and the parameters of the multivariate time series interpolation diffusion model are optimized based on the loss.

5. The object detection method according to claim 4, characterized in that: The step of performing forward diffusion processing on the electromagnetic signal sample features according to the joint Gaussian distribution to obtain forward diffusion data of the electromagnetic signal sample features includes: According to the current time step and the preset time step, the joint Gaussian distribution is sampled to determine the current step forward noise; Adding the current step forward noise to the previous step electromagnetic signal sample forward feature to obtain the current step electromagnetic signal sample forward feature; Gradually increase the current time step and repeat the above steps until the current time step is equal to the preset time step, and determine that the positive feature of the electromagnetic signal sample at the current time step is the noisy electromagnetic signal sample feature; The forward diffusion data of the electromagnetic signal sample characteristics also includes each current step forward noise and each current step electromagnetic signal sample forward characteristics.

6. The object detection method according to claim 5, characterized in that: The method further comprises: performing reverse diffusion processing on the noisy electromagnetic signal sample features according to the forward diffusion data of the electromagnetic signal sample features, determining the loss of the multivariate time series interpolation diffusion model in the reverse diffusion processing, and optimizing the parameters of the multivariate time series interpolation diffusion model based on the loss, including: Determine the current step reverse noise according to the current time step and the reverse characteristics of the electromagnetic signal sample of the current step; According to the reverse noise of the current step, the reverse feature of the electromagnetic signal sample of the current step is removed to determine the reverse feature of the electromagnetic signal sample of the previous step; Determining the current step loss of the multivariate time series interpolation diffusion model through an error estimation model; Optimizing the parameters of the multivariate time series interpolation diffusion model based on the current step loss; Gradually reduce the current time step and repeat the above steps until the current time step is 1.

7. The object detection method according to claim 6, characterized in that: Determining the current step loss of the multivariate time series interpolation diffusion model through an error estimation model includes: determining a first current step loss according to the current step reverse noise and the current step forward noise; Determining a second current step loss according to the reverse feature of the electromagnetic signal sample of the previous step and the forward feature of the electromagnetic signal sample of the previous step; The current step loss of the multivariate time series interpolation diffusion model is determined according to the first current step loss and the second current step loss.

8. The object detection method according to any one of claims 6 to 7, characterized in that: The error estimation model includes: a step embedding module, for determining an embedding vector according to the current time step, so that the multivariate time series interpolation diffusion model determines the current time step; A time embedding module, used to add time dimension information to the electromagnetic signal sample features; An attention mechanism estimator is used to perform high-dimensional computation processing on the outputs of the step embedding module and the time embedding module according to the attention mechanism.

9. An object detection device, characterized in that: include: An acquisition unit, used for acquiring electromagnetic signals of each sub-area to be detected; A correction unit, used for correcting the electromagnetic signal of each sub-area to be detected to obtain a corrected electromagnetic signal of each sub-area to be detected; The distinguishing unit is used to determine whether there is a target object in the area to be detected according to the calibrated electromagnetic signals of each sub-area to be detected, and the area to be detected is composed of the sub-areas to be detected.

10. A security door, 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 object detection method according to any one of claims 1 to 8 is implemented.

Images