WO2020134847A1

WO2020134847A1 - Foreign object detection method, security check apparatus, and device having storage function

Info

Publication number: WO2020134847A1
Application number: PCT/CN2019/121763
Authority: WO
Inventors: 黄雄伟; 祁春超
Original assignee: 深圳市华讯方舟太赫兹科技有限公司; 华讯方舟科技有限公司
Priority date: 2018-12-29
Filing date: 2019-11-28
Publication date: 2020-07-02
Also published as: CN109814166B; CN109814166A

Abstract

Disclosed are a foreign object detection method, a security check apparatus, and a device having a storage function. The method comprises: acquiring a millimeter wave signal reflected by a subject being checked; inputting the reflected millimeter wave signal to a three-dimensional foreign object identification network, and acquiring an identification result of the three-dimensional foreign object identification network, thus detecting whether the subject being checked is carrying a foreign object, where the three-dimensional foreign object identification network is a pretrained three-dimensional convolutional neural network. By such means, the present application implements quick and accurate foreign object detection.

Description

Foreign object detection method, security inspection equipment and device with storage function

Technical field

This application relates to the technical field of security inspection, in particular to a foreign object detection method, security inspection equipment, and a device with a storage function.

Background technique

The use of millimeter wave technology to detect human body is a more effective detection method. In the current detection method, the transmitting antenna array generally radiates a millimeter wave signal to the detection object, and the receiving antenna array receives the reflected millimeter wave, and then obtains a set of echo data. During detection, the echo data is imaged first, and then the object detection algorithm is used to detect foreign objects in the picture. In this process, it not only takes time, but also loses the spatial information of the scanned object. In addition, the accuracy of foreign object detection relies heavily on imaging and picture processing technologies.

Technical solution

This application mainly provides a foreign object detection method, security inspection equipment, and a device with a storage function, which can realize more rapid and accurate foreign object detection.

In order to solve the above technical problems, a technical solution adopted by the present application is to provide a foreign object detection method, which includes: acquiring a millimeter wave signal reflected by a detected object; inputting the reflected millimeter wave signal into a three-dimensional foreign object recognition network, and acquiring a three-dimensional foreign object The recognition result of the recognition network to detect whether the inspected object carries foreign objects; among them, the three-dimensional foreign object recognition network is a pre-trained three-dimensional convolutional neural network.

In order to solve the above technical problems, another technical solution adopted by the present application is to provide a security inspection device including: a receiving antenna and a processor connected to each other; the processor uses the receiving antenna to obtain the millimeter wave signal reflected by the object to be inspected and execute Program to implement the foreign object detection method as described above.

In order to solve the above technical problems, another technical solution adopted by the present application is to provide a device with a storage function that stores an instruction, and when the instruction is executed, the foreign object detection method described above is implemented.

The beneficial effect of the present application is that it is different from the situation in the prior art. In the embodiments of the present application, after acquiring the millimeter wave signal reflected by the detected object, the reflected millimeter wave signal is directly input into a pre-trained three-dimensional foreign object recognition network Medium, that is, the recognition results of the three-dimensional foreign object recognition network can be obtained to achieve the purpose of detecting whether the detected object carries foreign objects, without the need for imaging processing and feature extraction and target detection of the picture, so that the spatial structure of the detected object can be retained to the maximum Information and detailed information to achieve faster and more accurate detection.

BRIEF DESCRIPTION

FIG. 1 is a schematic flowchart of a first embodiment of a foreign object detection method of the present application;

2 is a schematic diagram of a scenario in which the foreign object detection method of the present application is applied to a human body security detector;

3 is a schematic flowchart of a second embodiment of the foreign object detection method of the present application;

4 is a schematic diagram of feature extraction using a 3D convolution kernel;

FIG. 5 is a schematic flowchart of step S131 in FIG. 3;

FIG. 6 is a schematic diagram of feature extraction on data of multiple channels using the same 3D convolution kernel;

7 is a schematic flowchart of step S132 in FIG. 3;

8 is a schematic flowchart of a third embodiment of the foreign object detection method of the present application;

9 is a schematic structural view of an embodiment of the security inspection device of the present application;

FIG. 10 is a schematic structural diagram of an embodiment of a device with a storage function according to the present application.

Embodiments of the invention

The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without making creative efforts fall within the protection scope of the present application.

As shown in FIG. 1, the first embodiment of the foreign object detection method of the present application includes:

S12: Obtain the millimeter wave signal reflected by the detected object.

Among them, the detected object may be a human body or an object. Because the electromagnetic radiation power of millimeter waves (such as terahertz waves) is small, millimeter wave human body security instruments are widely used in human body security inspection. In the embodiment of the present application, the object to be tested is described by taking a human body as an example.

If the subject to be inspected carries foreign objects, such as metal knives, powders or liquids, etc., due to the different degree of absorption of the millimeter waves by the foreign materials of different materials, and the degree of absorption of the millimeter waves by the foreign objects and human skin or clothing, they are The millimeter wave signal reflected by the inspection object can reflect whether the inspection object carries foreign objects.

Optionally, as shown in FIG. 1, before step S12, it further includes:

S11: Send a millimeter wave signal to the subject.

Before receiving the echo signal reflected by the detected object, that is, the reflected millimeter wave signal, it is necessary to first transmit the millimeter wave signal to the detected object.

Specifically, in an application example, as shown in FIG. 2, the human body security instrument may be a cylindrical security instrument as shown in FIG. 2, the cylindrical security instrument is provided with a scanning antenna array, and the scanning antenna array 11 includes a transmitting antenna array and The receiving antenna array, the transmitting antenna array includes a plurality of transmitting antennas, the receiving antenna array includes a plurality of receiving antennas, and each transmitting antenna and each receiving antenna correspond one-to-one to transmit a millimeter wave transmission signal and receive an echo signal. When it is necessary to perform security inspection on the human body A, the scanning antenna array can be controlled to rotate. During the rotation, the transmitting antenna array transmits the millimeter wave signal to the human body A, and the receiving antenna array receives the millimeter wave signal (ie, the echo signal) reflected by the human body A.

Among them, for one scan result of each detected object, the received millimeter wave signal reflected by the detected object can be saved as a multi-dimensional matrix. For example, the transmitting antenna array radiates the millimeter wave signal to the detection object, and the receiving antenna array receives the reflected millimeter wave signal. For each scanned object, the reflected millimeter wave signal can be saved to obtain a 250*415*96 matrix .

S13: Input the reflected millimeter wave signal into the three-dimensional foreign object recognition network, and obtain the recognition result of the three-dimensional foreign object recognition network to detect whether the detected object carries foreign objects.

Wherein, the three-dimensional foreign object recognition network is a pre-trained three-dimensional convolutional neural network. The three-dimensional convolutional neural network can directly perform foreign object recognition based on the input millimeter wave signal, and its recognition result can directly display whether the detected object carries foreign objects.

Specifically, in an application example, after acquiring the millimeter wave signal reflected by the detected object, since the millimeter wave signal has information reflecting whether the detected object carries foreign objects, the millimeter wave signal is input to a pre-trained three-dimensional foreign object In the recognition network, the three-dimensional foreign object recognition network can process the input millimeter wave signal to identify whether the millimeter wave signal carries the characteristic information of the foreign body, if the millimeter wave signal carries the characteristic information of the foreign body, then The output of the recognition result will show that the inspected object carries foreign objects, otherwise it shows that the inspected object does not carry foreign objects. In addition, in other application examples, the three-dimensional foreign object recognition network can also recognize the positions of foreign objects at the same time, and simultaneously output the positions of the foreign objects in the recognition result, for example, the foreign body is located on the waist of the object to be detected in the recognition result.

In this embodiment, after acquiring the millimeter wave signal reflected by the detected object, the reflected millimeter wave signal is directly input into the pre-trained three-dimensional foreign object recognition network, that is, the recognition result of the three-dimensional foreign object recognition network can be obtained to realize the detection of the detected object The purpose of whether the object carries foreign objects does not require imaging processing and feature extraction and target detection of the picture, so that the spatial structure information and detailed information of the detected object can be retained to the maximum extent, and because the three-dimensional foreign object recognition network uses a large number of samples in advance The three-dimensional convolutional neural network trained on the data can directly output the recognition results, thereby helping to achieve faster and more accurate detection.

As shown in FIG. 3, the second embodiment of the foreign object detection method of the present application is based on the first embodiment of the foreign object detection method of the present application, and further defining step S13 includes:

S131: Perform feature extraction on the reflected millimeter wave signal to obtain at least one feature information of the detected object.

Among them, the reflected millimeter wave signal carries the outline of the object to be inspected, and the outline, material, and other characteristics of the object carried by the object to be inspected. The three-dimensional foreign object recognition network has a feature extraction layer, which can perform feature extraction on the input millimeter wave signal, and it can advance at least one kind of feature information, such as material features.

Optionally, when extracting features, a 3D convolution kernel can be used for feature extraction.

Specifically, at least one three-dimensional space convolution kernel may be used to perform feature extraction on the reflected millimeter wave signal to obtain at least one feature information of the detected object. For example, as shown in FIG. 4, the millimeter wave signal reflected by each detected object can form a multi-dimensional matrix 401 (such as a 250*415*96 three-dimensional matrix), and each three-dimensional space convolution kernel (such as a 3*3*3 convolution) Product kernel), the convolution box 402 can slide through the entire matrix 401 from left to right, from top to bottom, and from front to back in sequence. Without padding and a step size of 1, you will get a 248*413*94 The matrix 403 is a kind of characteristic information.

A convolution kernel uses a set of weights to extract features in the same way when "scanning" the matrix, and the result is a feature. Using multiple convolution kernels can extract a variety of features, so as to achieve the purpose of extracting at least one feature information of the detected object.

Optionally, before extracting features, the acquired millimeter wave signal may have multiple channels of data, and for each channel, the same 3D convolution kernel may be used for feature extraction. Specifically, as shown in FIG. 5, step S131 includes:

S1311: Use the same three-dimensional space convolution kernel to perform feature extraction on different channel data of the reflected millimeter wave signal to obtain feature information of different channels.

The reflected millimeter wave signal may include different channel data such as the abscissa gradient and the ordinate gradient. The three-dimensional foreign object recognition network may further include a channel extraction layer before the feature extraction layer (such as a convolution layer). The channel extraction layer first extracts the data of different channels from the reflected millimeter wave signal, and then the data of the different channels Enter the feature extraction layer in sequence for feature extraction.

S1312: Synthesize the feature information of different channels to obtain the feature information of the detected object corresponding to the three-dimensional space convolution kernel.

Specifically, in an application example, as shown in FIG. 6, three channels of data can be extracted from the reflected millimeter wave signal, as shown in FIG. 6 by three

matrices

601, 602, and 603. For three matrix data, You can use the same 3D convolution kernel 604 to extract features for each matrix, that is, use the convolution frame of the 3D convolution kernel with the same weight to slide through each matrix from left to right, from top to bottom, and from front to back To obtain three feature information corresponding to the three channels. Since each 3D convolution kernel can only obtain one feature information of the detected object, when three feature information of three channels are obtained, the three feature information can be synthesized, for example, the three feature information are directly superimposed, that is After the elements at the same position in the three feature information are added, the feature information (feature matrix 610) of the detected object corresponding to the 3D convolution kernel can be obtained. Of course, in other application examples, when synthesizing the feature information of different channels, it is also possible to perform synthesis by weighted superposition or directly summing and adding offset data.

Before performing feature extraction on the reflected millimeter wave signal, downsampling can also be performed first to reduce the amount of data and then perform feature extraction, which can reduce the amount of data processing for feature extraction, speed up feature extraction, and ultimately accelerate the speed of foreign object recognition.

S132: Classify the at least one feature information to obtain the recognition result.

The three-dimensional foreign object recognition network further includes a classifier, which is pre-trained to classify the input feature information to distinguish whether the detected object carries foreign objects. The classifier may be a linear classifier. Each classifier can distinguish whether the detected object carries a certain kind of foreign object. The three-dimensional foreign object recognition network can include multiple classifiers to distinguish whether the detected object carries a variety of foreign objects.

Specifically, after acquiring a kind of feature information, for example, after acquiring a feature matrix, it is input into a classifier, and the classifier can classify the feature information to distinguish whether the feature information includes feature information of foreign objects, such as metal The feature information of the tool, if the feature information includes the feature information of the foreign object, the object to be inspected carries the foreign object in the output recognition result, and the type of the foreign object can also be output in the recognition result.

Optionally, preset foreign object feature information may be preset in each classifier, and the input feature information is compared with the preset foreign object feature information during classification, that is, classification may be performed. Specifically, as shown in FIG. 7, step S132 includes:

S1321: Compare the at least one characteristic information with the preset foreign object characteristic information.

The preset foreign object feature information may be foreign object feature information samples acquired during training or learning of the classifier, or may be foreign object feature information summarized or learned during training or learning of the classifier. The preset foreign object characteristic information includes characteristic information of various prohibited items such as metal knives, ceramic knives, lighters, alcohol, powder, or the characteristic information collected when various prohibited items are placed in different parts of the human body.

Specifically, in an application example, after inputting at least one type of feature information of the detected object to the classifier, the classifier may compare the input feature information with preset foreign object feature information, if the input feature information Includes preset foreign object feature information, or part of the feature information is consistent with the preset foreign object feature information or the difference is within the allowable error range, it means that the at least one feature information matches the preset foreign object feature information, otherwise the two are not match.

If the at least one feature information matches the preset foreign object feature information, step S1322 is executed, otherwise step S1323 is executed.

S1322: Divide the detected object into carrying foreign objects, and output the recognition result of the detected object carrying foreign objects.

S1323: Divide the inspected object into uncarried foreign objects, and output the recognition result of the inspected object without carried foreign objects.

Specifically, if at least one feature information of the input classifier matches the preset foreign object feature information, the classifier divides the detected object into carrying foreign objects, and the three-dimensional foreign object recognition network outputs that the detected object carries foreign objects Recognition result; otherwise, the classifier classifies the detected object as not carrying foreign objects, and the three-dimensional foreign object recognition network outputs the recognition result of the detected object without carrying foreign objects.

Optionally, in the case where the object to be inspected is classified as carrying a foreign object, the three-dimensional foreign object recognition network can also identify the location of the foreign object. Specifically, as shown in FIG. 7, after step S1322, the method further includes:

S1324: Identify the location of the foreign object to obtain the spatial location information of the foreign object.

The spatial position information of the foreign object may be relative spatial position information of the foreign object and the detected object, for example, the foreign object is inside the arm of the detected object.

Specifically, when the classifier divides the detected object into carrying foreign objects, it means that the feature information input to the classifier matches the preset foreign object feature information. At this time, the feature information can be consistent with the preset foreign object feature information Or the spatial feature information (such as spatial position information) corresponding to the partial feature information whose difference is within the allowable error range, to identify the position of the foreign object carried by the subject to obtain the spatial position information of the foreign object. For example, when the classifier performs classification, the preset foreign object feature information includes the location information of the foreign object. When the classifier classifies the input feature information, it can simultaneously identify the location of the foreign object.

Of course, in other embodiments, the part of the millimeter wave signal reflected by the detected object and the part of the feature may be obtained according to the part of the feature information that is consistent with the preset foreign object feature information or the difference between the two is within the allowable error range The data corresponding to the information to obtain the spatial position data of the corresponding data in the millimeter wave signal, and then the spatial position information of the foreign object.

S1325: Output the spatial position information of the foreign object in the recognition result.

Specifically, after the spatial location information of the foreign object is obtained, the spatial location information of the foreign object can be output simultaneously in the recognition result output by the three-dimensional foreign object recognition network. For example, the recognition result may include: a ceramic cutter is carried inside the forearm of the subject to be inspected.

Among them, the recognition results of the three-dimensional foreign object recognition network can be presented in the form of text, voice or images.

In this embodiment, the pre-trained three-dimensional foreign object recognition network is used to extract and classify the reflected millimeter wave signals, that is, whether the detected object carries the foreign object recognition result can be obtained. If the foreign object is carried, it can also be directly obtained The location of the foreign object does not require imaging processing and feature extraction and target detection of the picture, so that the spatial structure information and detailed information of the detected object can be retained to the maximum extent, and because the three-dimensional foreign object recognition network is pre-utilized by a large number of sample data The trained three-dimensional convolutional neural network can directly output the recognition result, thereby helping to achieve faster and more accurate detection.

As shown in FIG. 8, the third embodiment of the foreign object detection method of the present application is based on the first embodiment of the foreign object detection method of the present application, and before further defining step S11, further includes:

S101: Collect sample data carried by a plurality of inspected objects and not carried foreign objects.

In the sample data, the plurality of objects to be inspected include human bodies of different genders. The objects to be inspected include at least one type. For example, the objects to be inspected can carry one or more foreign objects at a time; The position of the foreign object may also be different every time, for example, this time place the foreign object on the waist, and next time place the foreign object on the forearm.

S102: Train the sample data to obtain a three-dimensional foreign object recognition network.

Before using the three-dimensional foreign object recognition network for foreign object recognition, the three-dimensional foreign object recognition network needs to be trained and learned to obtain the parameters of the recognition model used by the three-dimensional foreign object recognition network, such as the weight value of the 3D convolution kernel and the parameters of the classifier Wait.

Specifically, before training the three-dimensional foreign object recognition network, it is necessary to collect sample data required for training. At this time, sample data carried by a plurality of inspected objects and not carried foreign objects can be collected, and the sample data is input to random initialization After the three-dimensional foreign object recognition network, the three-dimensional foreign object recognition network can be trained, for example, using an online BP (back propagation) algorithm, and the final training results in recognition accuracy that meets the requirements (such as more than 90%) Three-dimensional foreign body recognition network.

In other embodiments, in the process of using the three-dimensional foreign object recognition network to recognize the input feature information, the model parameters of the three-dimensional foreign object recognition network can also be trained and adjusted in real time based on the feedback of the recognition results and the real results to continuously improve the three-dimensional Foreign object recognition network to improve the accuracy of foreign object recognition.

As shown in FIG. 9, in an embodiment of a security inspection device of the present application, the security inspection device 80 includes: a receiving antenna 801 and a processor 802 connected to each other.

The receiving antenna 801 may be a receiving antenna array, including a plurality of receiving antenna units, and may also include a switch array, each receiving antenna unit is connected to a switch in the switch array, and the switch array may control the working state of the receiving antenna 801 , For receiving the millimeter wave signal reflected by the detected object.

The processor 802 may be a central processing unit (Central Processing Unit, CPU), or other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application-specific integrated circuits (Application Specific Integrated Circuit, ASIC), off-the-shelf Programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The processor 802 uses the receiving antenna 801 to acquire the millimeter wave signal reflected by the object to be inspected, and executes a program to implement any one of the first to third embodiments of the foreign object detection method of the present application or a combination thereof that does not conflict method.

Optionally, the security inspection device 80 further includes: a transmitting antenna 803 connected to the processor 802, for transmitting a millimeter wave signal to the object to be inspected.

The transmitting antenna 803 may be a transmitting antenna array. The transmitting antenna array includes a plurality of transmitting antenna elements, and each transmitting antenna element corresponds to each receiving antenna element to transmit a millimeter wave transmission signal and receive an echo signal.

In other embodiments, the security inspection device 80 may further include a memory 804 for storing data and program instructions required by the processor 802.

The storage 804 may be an internal storage unit of the security inspection device 80, such as a hard disk or a memory of the security inspection device 80. The memory 804 may also be an external storage device of the security inspection device 80, such as a plug-in hard disk equipped on the security inspection device 80, a smart memory card (Smart) Media (SMC), a secure digital (SD) card, and a flash memory card (Flash Card) etc. Further, the memory 804 may also include both the internal storage unit of the security inspection device 80 and the external storage device. The memory 804 may also be used to temporarily store data that has been or will be output.

Those skilled in the art may understand that FIG. 9 is only an example of the security inspection device 80, and does not constitute a limitation on the security inspection device 80, and may include more or fewer components than the illustration, or a combination of certain components, or different components. For example, security inspection equipment may also include input and output devices, network access devices, buses, etc.

In this embodiment, after acquiring the millimeter wave signal reflected by the object to be inspected, the security inspection device directly inputs the reflected millimeter wave signal into the pre-trained three-dimensional foreign object recognition network, that is, the recognition result of the three-dimensional foreign object recognition network can be obtained to realize detection The purpose of whether the detected object carries foreign objects does not require imaging processing and feature extraction and target detection of the picture, so that the spatial structure information and detailed information of the detected object can be retained to the maximum extent, and because the three-dimensional foreign object recognition network is pre-utilized A three-dimensional convolutional neural network trained with a large number of sample data can directly output the recognition results, thereby helping to achieve faster and more accurate detection.

As shown in FIG. 10, in an embodiment of a device with a storage function in the present application, an instruction 901 is stored inside the device with a storage function 90, and when the instruction 901 is executed, the first to third embodiments of the foreign object detection method of the present application The method provided by any one of the embodiments or a non-conflicting combination

The storage-capable device 90 may specifically be a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, etc. that can store program instructions The medium may also be a server that stores the program instructions. The server may send the stored program instructions to other devices to run, or it may run the stored program instructions by itself.

In an embodiment, the device 90 with a storage function may also be a memory as shown in FIG. 9.

In this embodiment, when the instruction stored in the device with the storage function is executed, by acquiring the millimeter wave signal reflected by the detected object, the reflected millimeter wave signal can be directly input into the pre-trained three-dimensional foreign object recognition network, that is, Obtain the recognition results of the three-dimensional foreign object recognition network to achieve the purpose of detecting whether the detected object carries foreign objects, without imaging processing and feature extraction and target detection of the picture, so that the spatial structure information and detailed information of the detected object can be retained to the maximum extent And, because the three-dimensional foreign object recognition network is a three-dimensional convolutional neural network trained with a large number of sample data in advance, it can directly output the recognition results, which in turn helps to achieve faster and more accurate detection.

Those skilled in the art can clearly understand that, for convenience and conciseness of description, only the above-mentioned division of each functional unit and module is used as an example for illustration. In practical applications, the above-mentioned functions may be allocated by different functional units, Module completion means that the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit. The above integrated unit may use hardware It can also be implemented in the form of software functional units. In addition, the specific names of each functional unit and module are only for the purpose of distinguishing each other, and are not used to limit the protection scope of the present application. For the specific working processes of the units and modules in the above system, reference may be made to the corresponding processes in the foregoing method embodiments, which will not be repeated here.

In the above embodiments, the description of each embodiment has its own emphasis. For a part that is not detailed or recorded in an embodiment, you can refer to the related descriptions of other embodiments.

Persons of ordinary skill in the art may realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application of the technical solution and design constraints. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

In the embodiments provided by the present invention, it should be understood that the disclosed device/terminal device and method may be implemented in other ways. For example, the device/terminal device embodiments described above are only schematic. For example, the division of the module or unit is only a logical function division, and in actual implementation, there may be another division manner, such as multiple units Or components can be combined or integrated into another system, or some features can be ignored or not implemented. In addition, the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms.

The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit. The above integrated unit can be implemented in the form of hardware or software function unit.

If the integrated module/unit is implemented in the form of a software functional unit and sold or used as an independent product, it may be stored in a computer-readable storage medium. Based on this understanding, the present invention can implement all or part of the processes in the methods of the above embodiments, and can also be completed by a computer program instructing relevant hardware. The computer program can be stored in a computer-readable storage medium. When the program is executed by the processor, the steps of the foregoing method embodiments may be implemented. Wherein, the computer program includes computer program code, and the computer program code may be in a source code form, an object code form, an executable file, or some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM, Read-Only Memory) , Random Access Memory (RAM, Random Access Memory), electrical carrier signals, telecommunications signals and software distribution media, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in jurisdictions. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media Excluded are electrical carrier signals and telecommunications signals.

The above are only the embodiments of the present application, and do not limit the patent scope of the present application. Any changes to the equivalent structure or equivalent process made by the description and drawings of this application, or used directly or indirectly in other related technologies In the field, the same reason is included in the scope of patent protection of this application.

Claims

A foreign object detection method, characterized in that it includes:

Obtain the millimeter wave signal reflected by the detected object;

Input the reflected millimeter wave signal into a three-dimensional foreign object recognition network, and obtain the recognition result of the three-dimensional foreign object recognition network to detect whether the detected object carries foreign objects;

Wherein, the three-dimensional foreign object recognition network is a pre-trained three-dimensional convolutional neural network.
The method according to claim 1, wherein the inputting the reflected millimeter wave signal to a three-dimensional foreign object recognition network and obtaining the recognition result of the three-dimensional foreign object recognition network to detect whether the detected object carries Foreign objects include:

Performing feature extraction on the reflected millimeter wave signal to obtain at least one feature information of the detected object;

Classify the at least one feature information to obtain the recognition result.
The method according to claim 2, wherein the feature extraction of the reflected millimeter wave signal to obtain at least one feature information of the detected object includes:

At least one three-dimensional space convolution kernel is used to perform feature extraction on the reflected millimeter wave signal to obtain at least one feature information of the detected object.
The method according to claim 3, wherein the feature extraction of the reflected millimeter wave signal using at least one three-dimensional space convolution kernel to obtain at least one feature information of the detected object includes:

Performing feature extraction on different channel data of the reflected millimeter wave signal by using the same three-dimensional space convolution kernel to obtain feature information of the different channels;

Synthesizing the feature information of the different channels to obtain feature information of the detected object corresponding to the three-dimensional space convolution kernel.
The method according to claim 2, wherein the classifying the at least one feature information to obtain the recognition result comprises:

Comparing the at least one characteristic information with the preset foreign body characteristic information;

If the at least one feature information matches the preset foreign object feature information, the detected object is classified as carrying a foreign object, and a recognition result of the detected object carrying a foreign object is output;

Otherwise, the detected object is classified as not carrying foreign objects, and the recognition result of the detected object carrying no foreign objects is output.
The method according to claim 5, wherein after dividing the detected object as carrying a foreign object and outputting the recognition result of the detected object as carrying a foreign object, the method further comprises:

Identifying the location of the foreign object to obtain the spatial location information of the foreign object;

The spatial position information of the foreign object is output in the recognition result.
The method according to any one of claims 1 to 6, characterized in that, before acquiring the millimeter wave signal reflected by the detected object, the method includes:

A millimeter wave signal is emitted to the subject.
The method according to claim 7, characterized in that, before the millimeter wave signal is transmitted to the subject, the method includes:

Collect sample data of multiple objects with or without foreign objects;

Training the three-dimensional foreign object recognition network by using the sample data.
A security inspection device, characterized by comprising: a receiving antenna and a processor connected to each other;

The processor uses the receiving antenna to obtain the millimeter wave signal reflected by the detected object, and executes a program to implement the foreign object detection method according to any one of claims 1-8.
The security inspection device according to claim 9, further comprising: a transmitting antenna connected to the processor, for transmitting a millimeter wave signal to the inspected object.
A device with a storage function that stores instructions, characterized in that when the instructions are executed, the foreign object detection method according to any one of claims 1-8 is realized.