A partition wall detection imaging method, device, readable storage medium and detection equipment

Abstract

The present invention discloses a wall detection imaging method, device, readable storage medium and detection equipment, the method comprising: obtaining an electromagnetic wave detection signal detected by a detection radar, the electromagnetic wave detection signal at least comprising an echo of a target to be detected, an echo of an obstacle and clutter; inputting the electromagnetic wave detection signal into a pre-trained imaging model, so as to obtain a detection imaging according to the echo of the target to be detected through the pre-trained imaging model. The present invention solves the problem of low imaging accuracy of wall detection imaging in the prior art.

Description

Technical Field

[0001] The present invention relates to the field of detection technology, and in particular to a partition wall detection imaging method, device, readable storage medium and detection equipment. Background Art

[0002] In recent years, the demand for life detection, target positioning and imaging behind walls has always been a concern. At present, the solution based on radar system has become mature, but the detection technology for targets inside walls has developed slowly. However, the demand for detection of targets inside walls is becoming more and more urgent in many fields. Ultra-wideband radar can use the low-frequency penetration of electromagnetic waves to detect non-transparent media such as walls and wooden boards. It can receive target echoes and use a series of processing algorithms to realize the positioning, speed measurement and imaging of targets inside obstacles. Good wall penetration ability and strong anti-interference ability also make ultra-wideband radar widely used in target detection, positioning and imaging in indoor and closed environments. The whole process does not require contact, nor does it need to displace or destroy obstacles. It has unique advantages and broad application prospects in national defense security, military, construction and other aspects.

[0003] In the prior art, when processing detection data through radar detection reflection, since the material used for the wall is a non-uniformly distributed medium, and the target to be detected is located on the surface of the medium layer, this will cause the medium surface reflected clutter and the target echo to alias, increasing the difficulty of signal processing. At the same time, the intensity of the medium surface reflected clutter is greater than the target echo, causing the target echo to be completely submerged, resulting in poor imaging effect. Summary of the invention

[0004] In view of this, an object of the present invention is to provide a partition wall detection imaging method, device, readable storage medium and equipment, aiming to solve the problem of low accuracy of partition wall detection imaging in the prior art.

[0005] The present invention is implemented as follows: a partition wall detection imaging method, the method comprising:

[0006] Acquire an electromagnetic wave detection signal detected by a detection radar, wherein the electromagnetic wave detection signal at least includes an echo of a target to be detected, an echo of an obstacle, and clutter;

[0007] The electromagnetic wave detection signal is input into a pre-trained imaging model, so as to obtain detection imaging according to the echo of the target to be detected through the pre-trained imaging model.

[0008] Furthermore, in the above partition wall detection imaging method, the training process of the imaging model includes:

[0009] Acquiring a historical electromagnetic wave detection signal and an optical image corresponding to the historical electromagnetic wave detection signal;

[0010] Inputting the historical electromagnetic wave detection signal into a preset convolutional neural network for deep learning training;

[0011] The output target image is obtained until the similarity between the target image and the optical image is higher than a preset similarity threshold.

[0012] Furthermore, in the above-mentioned partition wall detection imaging method, the convolutional neural network includes an input layer, 10 convolution blocks and an output layer connected in sequence, each of the convolution blocks includes a first convolution block, a second convolution block and a third convolution block connected in sequence, the first convolution block includes 64 convolution kernels, the second convolution block includes 32 convolution kernels, the third convolution block includes 1 convolution kernel, and the size of the convolution kernel is 3*3.

[0013] Furthermore, in the above-mentioned partition wall detection imaging method, an activation function is provided after each of the first convolution blocks and the second convolution.

[0014] Furthermore, in the above partition wall detection imaging method, after the step of acquiring the historical electromagnetic wave detection signal and the optical image corresponding to the historical electromagnetic wave detection signal, the method further comprises:

[0015] The historical electromagnetic wave detection signal is subjected to dimensionality reduction processing, and the optical image is sequentially subjected to left-right mirroring, grayscale conversion, cropping compression, and binarization processing.

[0016] Furthermore, in the above-mentioned partition wall detection imaging method, before the step of acquiring the electromagnetic wave detection signal detected by the detection radar, wherein the electromagnetic wave detection signal at least includes the echo of the target to be detected, the echo of the obstacle and the clutter, the following step is further included:

[0017] Obtain background information of obstacles in the target to be detected; when detecting the target to be detected, perform a cancellation process according to the background information.

[0018] Furthermore, in the above-mentioned partition wall detection imaging method, during the training process of the convolutional neural network, the current loss function of the convolutional neural network is calculated, and the network parameters of the convolutional neural network are adjusted according to the current loss function to iteratively update the weights of the convolutional neural network and then perform training.

[0019] Another object of the present invention is to provide a partition wall detection imaging device, the device comprising:

[0020] An acquisition module is used to acquire the electromagnetic wave detection signal detected by the detection radar, wherein the electromagnetic wave detection signal at least includes an echo of a target to be detected, an echo of an obstacle, and clutter;

[0021] An imaging module is used to input the electromagnetic wave detection signal into a pre-trained imaging model so as to obtain detection imaging according to the echo of the target to be detected through the pre-trained imaging model.

[0022] Furthermore, in the above partition wall detection imaging device, the system further comprises:

[0023] A training set acquisition module, used to acquire historical electromagnetic wave detection signals and optical images corresponding to the historical electromagnetic wave detection signals;

[0024] A training module, used for inputting the historical electromagnetic wave detection signal into a preset convolutional neural network for deep learning training;

[0025] The image acquisition module is used to acquire the output target image until the similarity between the target image and the optical image is higher than a preset similarity threshold.

[0026] Furthermore, in the above-mentioned partition wall detection imaging device, in the training module, the convolutional neural network includes 10 convolution blocks connected in sequence, each of the convolution blocks includes a first convolution block, a second convolution block and a third convolution block connected in sequence, the first convolution block includes 64 convolution kernels, the second convolution block includes 32 convolution kernels, the third convolution block includes 1 convolution kernel, and the size of the convolution kernel is 3*3.

[0027] Furthermore, in the above-mentioned partition wall detection imaging device, in the training module, each of the first convolution blocks and the second convolution blocks are provided with an activation function.

[0028] Furthermore, in the above partition wall detection imaging device, the system further comprises:

[0029] The processing module is used to perform dimensionality reduction processing on the historical electromagnetic wave detection signal, and to perform left-right mirroring, grayscale conversion, cropping compression and binarization processing on the optical image in sequence.

[0030] Furthermore, in the above partition wall detection imaging device, the system further comprises:

[0031] The cancellation module is used to obtain background information of obstacles in the target to be detected; when detecting the target to be detected, cancellation processing is performed according to the background information.

[0032] Furthermore, in the above-mentioned partition wall detection imaging device, in the training module, the current loss function of the convolutional neural network is calculated, and the network parameters of the convolutional neural network are adjusted according to the current loss function, so as to train the convolutional neural network after iteratively updating the weights of the convolutional neural network.

[0033] Another object of the present invention is to provide a readable storage medium having a computer program stored thereon, wherein the program implements the steps of the above method when executed by a processor.

[0034] Another object of the present invention is to provide a detection device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the program.

[0035] The present invention uses a pre-trained imaging model and inputs the electromagnetic wave detection signal detected by the radar into the imaging model. The model is used to separate the wall echo to highlight the outline of the target, and to suppress the shallow clutter before imaging to achieve a better imaging effect, thereby achieving high-precision imaging of radar wall detection. The problem of low imaging accuracy of radar wall detection in the prior art is solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A schematic diagram of the structure of a partition wall detection imaging system provided by an embodiment of the present invention;

[0037] Figure 2 It is a flow chart of the partition wall detection imaging method proposed in the first embodiment of the present invention;

[0038] Figure 3 It is a flow chart of the partition wall detection imaging method proposed in the second embodiment of the present invention;

[0039] Figure 4 A process diagram of optical image preprocessing in the partition wall detection imaging method proposed in the second embodiment of the present invention;

[0040] Figure 5 It is a structural diagram of a convolutional neural network in a partition wall detection imaging method proposed in one embodiment of the present invention;

[0041] Figure 6 It is a flow chart of the partition wall detection imaging method proposed in the third embodiment of the present invention;

[0042] Figure 7 4 is a structural block diagram of a partition wall detection imaging device in a fourth embodiment of the present invention.

[0043] The following specific implementation manner will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION

[0044] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. Several embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0045] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0046] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed types.

[0047] In recent years, the demand for life detection, target positioning and imaging behind walls has always been a concern. At present, the solution based on radar system has become mature, but the detection technology for targets inside walls has developed slowly. However, the demand for detection of targets inside walls is becoming more and more urgent in many fields. Ultra-wideband radar can use the low-frequency penetration of electromagnetic waves to detect non-transparent media such as walls and wooden boards. It can receive target echoes and use a series of processing algorithms to realize the positioning, speed measurement and imaging of targets inside obstacles. Good wall penetration ability and strong anti-interference ability also make ultra-wideband radar widely used in target detection, positioning and imaging in indoor and closed environments. The whole process does not require contact, nor does it need to displace or destroy obstacles. It has unique advantages and broad application prospects in national defense security, military, construction and other aspects.

[0048] In the prior art, when processing detection data through radar detection reflection, since the material used for the wall is a non-uniformly distributed medium, and the target to be detected is located on the surface of the medium layer, this will cause the medium surface reflected clutter and the target echo to alias, increasing the difficulty of signal processing. At the same time, the intensity of the medium surface reflected clutter is greater than the target echo, causing the target echo to be completely submerged, resulting in poor imaging effect.

[0049] See also Figure 1 , which shows a radar detection imaging system proposed in one embodiment of the present invention, and the following embodiments can be applied to Figure 1The radar detection imaging system shown in the figure includes an embedded processing unit, an FMCW radar front end, an LCD display, a power module, a data acquisition module and a data preprocessing module; wherein:

[0050] The radar front end completes the modulation and demodulation of the FMCW signal, and completes the transmission and reception of the signal through the MIMO antenna array; the power supply supplies power to each module of the system; the embedded processing unit completes the acquisition and processing of the radar signal, and communicates with the radar using the USB2.0 protocol; the LCD display screen serves as the user UI interaction channel to realize functions such as calibration, start / stop detection, and interface clearing; at the software level, the software architecture completes the processing and analysis of data and images based on the Python language; the data acquisition module acquires radar signals; the data preprocessing module analyzes, preprocesses and performs target imaging on the acquired data.

[0051] It should be pointed out that Figure 1 The structure shown does not constitute a limitation on the radar detection and imaging system. In other embodiments, the radar detection and imaging system may include fewer or more components than those shown in the figure, or combine certain components, or arrange the components differently.

[0052] How to achieve high-precision imaging detected by radar will be described in detail below with reference to specific embodiments and drawings.

[0053] Embodiment 1

[0054] See also Figure 2 , which shows a partition wall detection imaging method in a first embodiment of the present invention, and the method includes steps S10 to S11.

[0055] Step S10, acquiring the electromagnetic wave detection signal detected by the detection radar, wherein the electromagnetic wave detection signal at least includes the echo of the target to be detected, the echo of the obstacle and the clutter.

[0056] Among them, since most of the wall is made of non-uniformly arranged materials, when the radar scans the wall, the echo it obtains includes the echo of the target being measured, the echo generated by the wall, and other clutter. When imaging, the echo generated by the wall and other clutter will affect the imaging accuracy. Specifically, the radar transmits a signal electromagnetic wave to the detection target. After detecting the target, the signal electromagnetic wave reflects the echo signal and is received by the radar, thereby obtaining the electromagnetic wave detection signal.

[0057] Step S11, inputting the electromagnetic wave detection signal into a pre-trained imaging model, so as to obtain detection imaging according to the echo of the target to be detected through the pre-trained imaging model.

[0058] Among them, through the pre-trained imaging model, the electromagnetic wave detection signal can be processed, the wall echo can be separated separately to highlight the outline of the target, and the shallow clutter can be suppressed to achieve better imaging effects and realize high-precision imaging of radar wall detection.

[0059] In addition, in some optional embodiments of the present invention, in order to further improve the imaging effect of radar detection, the step of obtaining the electromagnetic wave detection signal detected by the detection radar, wherein the electromagnetic wave detection signal at least includes the echo of the target to be detected, the echo of the obstacle and the clutter, further includes:

[0060] Obtain background information of obstacles in the target to be detected; when detecting the target to be detected, perform a cancellation process according to the background information.

[0061] Specifically, the background echo matrix B is recorded by detecting the wall through radar, wherein the background echo matrix B contains wall clutter and other noise. Then, the original echo matrix C of the wall containing the detection target is detected by radar, wherein the original echo matrix C contains the echo of the detection target, wall clutter and other noise. By subtracting and canceling the two echo matrices, the required final target echo matrix D is obtained, thereby filtering out background clutter such as the wall. Among them, since the radar signal obtained by the radar front end includes: target information, wall noise and other Gaussian noise, the separation effect of the wall features is related to whether the extracted target information is complete and correct, and indirectly affects the final target imaging effect. Specifically, the wall background information is recorded through pre-scanning, and the wall background information is canceled according to the background information scanned last time when detecting the target to be detected, thereby realizing the feature separation of the wall and the target.

[0062] In summary, the wall detection imaging method in the above embodiment of the present invention uses a pre-trained imaging model and inputs the electromagnetic wave detection signal detected by the radar into the imaging model. The wall echo is separated by the model to highlight the outline of the target, and the shallow clutter is suppressed before imaging to achieve a better imaging effect, thereby realizing high-precision imaging of radar wall detection. The problem of low imaging accuracy of radar wall detection in the prior art is solved.

[0063] Embodiment 2

[0064] See also Figure 3 , which shows a partition wall detection imaging method in a second embodiment of the present invention, and the method includes steps S20 to S24.

[0065] Step S20, acquiring a historical electromagnetic wave detection signal and an optical image corresponding to the historical electromagnetic wave detection signal.

[0066] Among them, the optical image is a high-definition optical image of the corresponding target collected by an optical camera without obstruction behind a wall. Specifically, the historical electromagnetic wave detection signal and the optical image are the training set of the imaging model, among which the historical electromagnetic wave detection signal is the network input and the optical image is the network output.

[0067] In addition, in some preferred embodiments of the present invention, in order to improve the accuracy and imaging speed of the imaging model, the training set is preprocessed before the model training of the training set. Specifically, the historical electromagnetic wave detection signal is subjected to dimensionality reduction processing, and the optical image is sequentially subjected to left-right mirroring, grayscale, cropping, compression and binarization processing. By performing dimensionality reduction and other preprocessing on the radar data and the optical image respectively, the imaging speed of the wall detection is improved.

[0068] More specifically, Figure 4 As shown, when performing optical image processing, the original optical image is read, and then the optical image is horizontally flipped, that is, the image is mirrored left and right, and then the optical image is converted into a grayscale image, and then the grayscale image is cropped. In a specific implementation, the length and width of the optical image are cropped to preset multiples, such as 1 / 3 to 2 / 3, respectively, and then the cropped optical image is binarized, and finally the optical image is saved at a preset size, such as 51*51.

[0069] Among them, taking the processing of optical images by programming in Python as an example, the (Image.open()) function in the Python image library PIL is used to read the original optical image, and the (cv2.flip()) function in Python is used to complete the horizontal flip of the image, that is, to achieve the left and right mirroring of the image; the (image.convert()) function in the Python image library PIL is used to convert the original optical image into a grayscale image, each pixel is represented by 8 bits, 0 represents black, 255 represents white, and other numbers represent different grayscales. The (image.imcrop()) function in the Python image library PIL is used to crop the optical image to a length of 1 / 3 to 2 / 3 of the original image, and a width of 1 / 3 to 2 / 3 of the original width, and save it. The cv2.threshold(gray, 0, 255, cv2.THRESH_OTSU) function, that is, Otsu's binarization, automatically finds an optimal binarization threshold and completes the matrix binarization operation. The (image.resize()) function is used to convert the image to a size of 51*51 and save it.

[0070] Step S21, inputting the historical electromagnetic wave detection signal into a preset convolutional neural network for deep learning training.

[0071] It should be noted that, in addition to proposing a partition wall detection imaging method, the present invention also proposes a convolutional neural network architecture for partition wall detection imaging, such as Figure 5 As shown, the convolutional neural network includes an input layer, 10 convolution blocks (Lblock) and an output layer connected in sequence, each of the convolution blocks includes three sub-convolution blocks (Sblock) connected in sequence, namely the first convolution block, the second convolution block and the third convolution block, the first convolution block includes 64 convolution kernels, the second convolution block includes 32 convolution kernels, the third convolution block includes 1 convolution kernel, and the size of the convolution kernel is 3*3.

[0072] In the convolutional neural network, each convolution operation is a fusion of local information and an extraction of features. In the convolution operation, each convolution kernel can only extract one image feature. Setting too few convolution kernels will lead to insufficient extracted image features and the training effect will not be ideal. Therefore, this network sets 10 convolution blocks and sets three sub-convolution blocks in each convolution block, namely the first convolution block, the second convolution block and the third convolution block. The first convolution block Sblock uses 64 convolution kernels, the second convolution block Sblock uses 32 convolution kernels, and the third convolution block Sblock uses 1 convolution kernel. Each convolution block Lblock sets the convolution kernel in the above manner, which is sufficient to extract enough image features.

[0073] In addition, in some preferred embodiments of the present invention, the convolutional neural network introduces the nonlinear function Relu as an activation function after the first and second Sblock convolution blocks. The output of each convolutional layer is a linear function of the upper layer input. It is easy to verify that no matter how many layers the neural network has, the output is a linear combination of the input. In this case, there is only a simple linear relationship between the layers, and each layer is equivalent to matrix multiplication, which will make it difficult to complete more complex tasks. Therefore, by using a nonlinear activation function, the nonlinear relationship between the layers of the neural network can be increased. In this way, more relevant features can be mined, and the extracted features are more representative. When the same effect is achieved, fewer neurons actually work in this method, and the generalization ability of the network is stronger.

[0074] Furthermore, in some preferred embodiments of the present invention, in order to improve the imaging accuracy of the imaging model, during model training, the weights and biases of the convolutional neural network are adjusted to perform denoising on the input electromagnetic wave detection signal. Since the input electromagnetic wave detection signal includes the target characteristics and noise of the target to be detected, wherein the noise includes the wall noise and the interference noise introduced during the transmission of the radar echo, specifically, the network weights and biases are adjusted to make the network output close to the target characteristics, as shown below:

[0075] X=R+N=R+∑(b i +g i );

[0076] f(X)=f(R+∑(b i +g i ))=R;

[0077] Among them, X is the network input, N is the noise, R is the target feature of the target to be detected, F(X) is the network output, and b i is the wall noise, g i Interference noise introduced during the transmission of radar echo.

[0078] Step S22, obtaining the output target image until the similarity between the target image and the optical image is higher than a preset similarity threshold.

[0079] Among them, the goal of introducing neural networks is to achieve high-definition imaging of radar signals. The radar data signal is used as the network input, assuming it is X; the preprocessed optical image is used as the network label, assuming it is Y; the network output result is obtained, assuming it is F(X). When the designed network should make the output imaging result F(X) (i.e., the target image) as close as possible to the network label Y, the closer the output result F(X) is to the network label Y, the better the performance of the network, and the corresponding imaging accuracy is also higher, thereby obtaining an imaging model. When the similarity between the target image and the optical image is higher than the similarity threshold, it means that the current imaging model meets the imaging accuracy requirements, and the higher the similarity, the higher the imaging accuracy of the imaging model.

[0080] By way of example and not limitation, when performing a similarity comparison, the average values ​​of each row of pixels in the image are calculated in turn, the average values ​​of each row of pixels are recorded, and the variance of all the average values ​​is calculated to obtain the characteristic value of the image. The variance can well reflect the fluctuation of the characteristics of each row of pixels, that is, it records the main information of the image. By comparing the difference in the variances of the two sets of data, the degree of closeness between the images can be reflected. The smaller the variance difference, the more similar the images are.

[0081] In addition, in some optional embodiments of the present invention, in order to improve the accuracy of image similarity comparison, before comparing the images, the images used for comparison are reduced to a preset size, for example, 8*8. By adjusting the images used for comparison to the same size, the details of the images are removed, and only basic information such as structure and light and dark is retained, and the image differences caused by different sizes and proportions are discarded, thereby improving the accuracy of the similarity comparison.

[0082] Step S23, acquiring the electromagnetic wave detection signal detected by the detection radar, wherein the electromagnetic wave detection signal at least includes the echo of the target to be detected, the echo of the obstacle and the clutter.

[0083] Step S24, inputting the electromagnetic wave detection signal into a pre-trained imaging model, so as to obtain detection imaging according to the echo of the target to be detected through the pre-trained imaging model.

[0084] In summary, the wall detection imaging method in the above embodiment of the present invention uses a pre-trained imaging model and inputs the electromagnetic wave detection signal detected by the radar into the imaging model. The wall echo is separated by the model to highlight the outline of the target, and the shallow clutter is suppressed before imaging to achieve a better imaging effect, thereby realizing high-precision imaging of radar wall detection. The problem of low imaging accuracy of radar wall detection in the prior art is solved.

[0085] Embodiment 3

[0086] See also Figure 6 , which shows a partition wall detection imaging method proposed in a third embodiment of the present invention, and the method includes steps S30 to S35.

[0087] Step S30: acquiring a historical electromagnetic wave detection signal and an optical image corresponding to the historical electromagnetic wave detection signal.

[0088] Step S31, inputting the historical electromagnetic wave detection signal into a preset convolutional neural network for deep learning training.

[0089] Step S32, during the training process of the convolutional neural network, calculate the current loss function of the convolutional neural network, and adjust the network parameters of the convolutional neural network according to the current loss function to iteratively update the weights of the convolutional neural network and then perform training.

[0090] Among them, unlike the traditional stochastic gradient descent algorithm, the stochastic gradient descent algorithm maintains a single learning rate to update all the weights of the convolutional neural network, and the learning rate does not change during the training process. In this embodiment, the neural network weights can be iteratively updated based on the training data according to the loss function, and independent adaptive learning rates can be designed for different parameters, thereby improving the final imaging effect of the imaging model.

[0091] Specifically, the loss function is:

[0092]

[0093] Among them, m represents the total number of samples, T represents the label data, and f w (x i ) is the predicted value of the convolutional neural network.

[0094] Step S33, obtaining the output target image until the similarity between the target image and the optical image is higher than a preset similarity threshold.

[0095] Step S34, acquiring the electromagnetic wave detection signal detected by the detection radar, wherein the electromagnetic wave detection signal at least includes the echo of the target to be detected, the echo of the obstacle and the clutter.

[0096] Step S35, inputting the electromagnetic wave detection signal into a pre-trained imaging model, so as to obtain detection imaging according to the echo of the target to be detected through the pre-trained imaging model.

[0097] In summary, the wall detection imaging method in the above embodiment of the present invention uses a pre-trained imaging model and inputs the electromagnetic wave detection signal detected by the radar into the imaging model. The wall echo is separated by the model to highlight the outline of the target, and the shallow clutter is suppressed before imaging to achieve a better imaging effect, thereby realizing high-precision imaging of radar wall detection. The problem of low imaging accuracy of radar wall detection in the prior art is solved.

[0098] Embodiment 4

[0099] See also Figure 7 , shown is a partition wall detection imaging device proposed in a fourth embodiment of the present invention, the device comprising:

[0100] The acquisition module 100 is used to acquire the electromagnetic wave detection signal detected by the detection radar, wherein the electromagnetic wave detection signal at least includes the echo of the target to be detected, the echo of the obstacle and the clutter;

[0101] The imaging module 200 is used to input the electromagnetic wave detection signal into a pre-trained imaging model, so as to obtain detection imaging according to the echo of the target to be detected through the pre-trained imaging model.

[0102] Furthermore, in the above partition wall detection imaging device, the system further comprises:

[0103] A training set acquisition module, used to acquire historical electromagnetic wave detection signals and optical images corresponding to the historical electromagnetic wave detection signals;

[0104] A training module, used for inputting the historical electromagnetic wave detection signal into a preset convolutional neural network for deep learning training;

[0105] The image acquisition module is used to acquire the output target image until the similarity between the target image and the optical image is higher than a preset similarity threshold.

[0106] Furthermore, in the above-mentioned partition wall detection imaging device, in the training module, the convolutional neural network includes 10 convolution blocks connected in sequence, each of the convolution blocks includes a first convolution block, a second convolution block and a third convolution block connected in sequence, the first convolution block includes 64 convolution kernels, the second convolution block includes 32 convolution kernels, the third convolution block includes 1 convolution kernel, and the size of the convolution kernel is 3*3.

[0107] Furthermore, in the above-mentioned partition wall detection imaging device, in the training module, each of the first convolution blocks and the second convolution blocks are provided with an activation function.

[0108] Furthermore, in the above partition wall detection imaging device, the system further comprises:

[0109] The processing module is used to perform dimensionality reduction processing on the historical electromagnetic wave detection signal, and to perform left-right mirroring, grayscale conversion, cropping compression and binarization processing on the optical image in sequence.

[0110] Furthermore, in the above partition wall detection imaging device, the system further comprises:

[0111] The cancellation module is used to obtain background information of obstacles in the target to be detected; when detecting the target to be detected, cancellation processing is performed according to the background information.

[0112] Furthermore, in the above-mentioned partition wall detection imaging device, in the training module, the current loss function of the convolutional neural network is calculated, and the network parameters of the convolutional neural network are adjusted according to the current loss function, so as to train the convolutional neural network after iteratively updating the weights of the convolutional neural network.

[0113] The functions or operation steps implemented when the above modules are executed are substantially the same as those in the above method embodiments, and will not be repeated here.

[0114] Embodiment 5

[0115] Another aspect of the present invention further provides a readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the steps of the method described in any one of the above embodiments 1 to 3 are implemented.

[0116] Embodiment 6

[0117] On the other hand, the present invention also provides a detection device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method described in any one of the above embodiments 1 to 3 when executing the program.

[0118] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0119] Those skilled in the art will appreciate that the logic and / or steps represented in the flowchart or otherwise described herein, for example, may be considered as an ordered list of executable instructions for implementing logical functions, and may be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For purposes of this specification, a "computer-readable medium" may be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0120] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0121] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or a combination thereof: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0122] In the description of this specification, the description with reference to the terms "one embodiment", "some 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 invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same 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.

[0123] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A partition wall detection imaging method, characterized in that: The method comprises: Acquire an electromagnetic wave detection signal detected by a detection radar, wherein the electromagnetic wave detection signal at least includes an echo of a target to be detected, an echo of an obstacle, and clutter; Inputting the electromagnetic wave detection signal into a pre-trained imaging model, so as to obtain detection imaging according to the echo of the target to be detected through the pre-trained imaging model; The training process of the imaging model includes: Acquiring a historical electromagnetic wave detection signal and an optical image corresponding to the historical electromagnetic wave detection signal; Inputting the historical electromagnetic wave detection signal into a preset convolutional neural network for deep learning training; Acquire the output target image until the similarity between the target image and the optical image is higher than a preset similarity threshold; The convolutional neural network includes an input layer, 10 convolution blocks and an output layer connected in sequence, each of the convolution blocks includes a first convolution block, a second convolution block and a third convolution block connected in sequence, the first convolution block includes 64 convolution kernels, the second convolution block includes 32 convolution kernels, the third convolution block includes 1 convolution kernel, and the size of the convolution kernel is 3*3; Among them, when training the model, the weights and biases of the convolutional neural network are adjusted to denoise the input electromagnetic wave detection signal. The expression of denoising is as follows: X=R+N=R+∑(b i +g i ) f(X)=f(R+∑(b i +g i ))=R; Among them, X is the network input, N is the noise, R is the target feature of the target to be detected, F(X) is the network output, and b i is the wall noise, g i Interference noise introduced during the transmission of radar echo.

2. The wall detection imaging method according to claim 1, characterized in that: An activation function is arranged after each of the first convolution blocks and the second convolution blocks.

3. The wall detection imaging method according to claim 1, characterized in that: After the step of acquiring the historical electromagnetic wave detection signal and the optical image corresponding to the historical electromagnetic wave detection signal, the following step further includes: The historical electromagnetic wave detection signal is subjected to dimensionality reduction processing, and the optical image is sequentially subjected to left-right mirroring, grayscale conversion, cropping compression, and binarization processing.

4. The wall detection imaging method according to claim 1, characterized in that: The step of obtaining the electromagnetic wave detection signal detected by the detection radar, wherein the electromagnetic wave detection signal at least includes the echo of the target to be detected, the echo of the obstacle and the clutter, further includes: Obtain background information of obstacles in the target to be detected; when detecting the target to be detected, perform a cancellation process according to the background information.

5. The wall detection imaging method according to claim 1, characterized in that: During the training process of the convolutional neural network, the current loss function of the convolutional neural network is calculated, and the network parameters of the convolutional neural network are adjusted according to the current loss function, so as to perform training after iteratively updating the weights of the convolutional neural network.

6. A partition wall detection imaging device, characterized in that: Used to implement the partition wall detection imaging method according to any one of claims 1 to 5, the device comprises: An acquisition module is used to acquire the electromagnetic wave detection signal detected by the detection radar, wherein the electromagnetic wave detection signal at least includes an echo of a target to be detected, an echo of an obstacle, and clutter; An imaging module is used to input the electromagnetic wave detection signal into a pre-trained imaging model so as to obtain detection imaging according to the echo of the target to be detected through the pre-trained imaging model.

7. A readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A detection device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of any one of the methods of claims 1 to 5 when executing the program.

Images