Object recognition method and device, terminal equipment and storage medium

Abstract

The invention discloses an object identification method and device, terminal equipment and a storage medium, and belongs to the technical field of signal detection, and the method comprises the steps: carrying out the detection of a target object through a ground penetrating radar, and obtaining an electromagnetic wave detection signal; performing short-time Fourier transform processing on the electromagnetic wave detection signal to generate a first time-frequency diagram; identifying each non-target frequency band corresponding to each fixed interference signal in the first time-frequency graph, and masking each non-target frequency band to obtain a second time-frequency graph; inputting the second time-frequency graph into a dynamic interference signal separation model to determine each dynamic interference signal in the second time-frequency graph, and outputting a third time-frequency graph in which each dynamic interference signal is removed; and generating a radar detection image of the target object according to the third time-frequency graph, and identifying position information of the target object. Therefore, by implementing the method and the device, the problem that the recognition of the target object in the radar detection image is influenced because the radar detection image obtained by detecting the target object by the ground penetrating radar in the prior art contains the interference signal waveform can be solved.

Description

Technical Field

[0001] The present invention relates to the technical field of signal detection, and particularly to an object recognition method, device, terminal device and storage medium. Background Art

[0002] Ground Penetrating Radar (GPR) can conveniently and efficiently perform non-destructive detection on underground media and other invisible media. Ground Penetrating Radar uses high-frequency electromagnetic waves for detection. When these electromagnetic waves propagate in underground media, they will reflect and refract when encountering the interfaces of different media. By receiving these reflected electromagnetic wave signals, Ground Penetrating Radar can generate images or cross-sectional views of underground structures, thereby realizing the detection, recognition and positioning of underground target objects.

[0003] When Ground Penetrating Radar is used for road detection, due to the diversity of underground targets and the complexity of the underground environment during the detection process, when Ground Penetrating Radar detects target objects, a large number of false reflection signals are generated due to interference from ground debris, which are highly similar to the echo characteristics of the target objects to be detected. These interferences make it difficult for non-professionals to distinguish real targets from noise through the intuitive waveforms in the radar detection images. Especially in urban dense areas or complex geological conditions, misjudgment is likely to occur. Summary of the Invention

[0004] Embodiments of the present invention provide an object recognition method, device, terminal device and storage medium, which can solve the problem that the radar detection image obtained by detecting target objects with existing Ground Penetrating Radar contains interference signal waveforms, affecting the recognition of target objects in the radar detection image. Improve the accuracy and convenience of obtaining the position information of target objects when Ground Penetrating Radar detects target objects.

[0005] An embodiment of the present invention provides an object recognition method, including:

[0006] Detecting a target object through a Ground Penetrating Radar to obtain an electromagnetic wave detection signal; wherein, the electromagnetic wave detection signal includes a target signal, a plurality of fixed interference signals and a plurality of dynamic interference signals;

[0007] Performing short-time Fourier transform processing on the electromagnetic wave detection signal to generate a first time-frequency diagram corresponding to the electromagnetic wave detection signal;

[0008] Identifying non-target frequency bands corresponding to the respective fixed interference signals in the first time-frequency diagram, and masking the respective non-target frequency bands to obtain a second time-frequency diagram with the respective non-target frequency bands removed;

[0009] Input the second time-frequency diagram into the dynamic interference signal separation model, so that the dynamic interference signal separation model extracts the energy sudden increase feature and frequency band offset feature in the second time-frequency diagram, determines each dynamic interference signal in the second time-frequency diagram according to the energy sudden increase feature and frequency band offset feature, and outputs a third time-frequency diagram with each dynamic interference signal removed;

[0010] Generate a radar detection image of the target object according to the third time-frequency diagram, and identify the position information of the target object according to the radar detection image.

[0011] Further, the construction of the dynamic interference signal separation model includes:

[0012] Obtain a number of time-frequency diagram samples containing dynamic interference signals; among them, each time-frequency diagram sample is labeled with an energy sudden increase feature, a frequency band offset feature, and a dynamic interference signal;

[0013] Construct an initial dynamic interference signal separation model, use the time-frequency diagram sample as the input of the initial dynamic interference signal separation model, and use the time-frequency diagram sample with the dynamic interference signal removed as the output. Iteratively train the initial dynamic interference signal separation model until the initial dynamic interference signal separation model converges, and generate a dynamic interference signal separation model; among them, in each iterative training process, the initial dynamic interference signal separation model extracts the energy sudden increase feature and frequency band offset feature in the time-frequency diagram sample, and determines the time-frequency diagram sample in the time-frequency diagram sample according to the energy sudden increase feature and frequency band offset feature in the time-frequency diagram sample.

[0014] Further, the generating a radar detection image of the target object according to the third time-frequency diagram and identifying the position information of the target object according to the radar detection image includes:

[0015] Generate a radar detection image of the target object according to the third time-frequency diagram;

[0016] Convert the third time-frequency diagram into a binary representation to obtain the binary data of the third time-frequency diagram;

[0017] Use the connected component labeling algorithm to process the binary data to determine several connected components of the binary data;

[0018] Determine the position feature information of the target object on each connected component according to each connected component;

[0019] Determine the position information of the target object on the radar detection image according to the position feature information on each connected component.

[0020] Further, the performing short-time Fourier transform processing on the electromagnetic wave detection signal to generate the first time-frequency diagram corresponding to the electromagnetic wave detection signal includes:

[0021] Divide the electromagnetic wave detection signal into a plurality of first signal time segments, and apply a preset window function to each first signal time segment for processing to obtain each second signal time segment;

[0022] Use the fast Fourier transform algorithm to process each second signal time segment to obtain the spectrum of each second signal time segment;

[0023] Combine the spectra of each second signal time segment to obtain the first time-frequency diagram.

[0024] Based on the foregoing method item embodiments, the present invention correspondingly provides device item embodiments;

[0025] An embodiment of the present invention correspondingly provides an object recognition device, including: a detection signal acquisition module, a fixed interference signal removal module, a dynamic interference signal removal module, and a target object recognition module;

[0026] The detection signal acquisition module is used to detect a target object through a ground penetrating radar to obtain an electromagnetic wave detection signal; wherein, the electromagnetic wave detection signal includes a target signal, a plurality of fixed interference signals, and a plurality of dynamic interference signals;

[0027] The fixed interference signal removal module is used to perform short-time Fourier transform processing on the electromagnetic wave detection signal to generate a first time-frequency diagram corresponding to the electromagnetic wave detection signal; identify each non-target frequency band corresponding to each fixed interference signal in the first time-frequency diagram, and mask each non-target frequency band to obtain a second time-frequency diagram with each non-target frequency band removed;

[0028] The dynamic interference signal removal module is used to input the second time-frequency diagram into a dynamic interference signal separation model, so that the dynamic interference signal separation model extracts the energy sudden increase feature and frequency band offset feature in the second time-frequency diagram, determines each dynamic interference signal in the second time-frequency diagram according to the energy sudden increase feature and frequency band offset feature, and outputs a third time-frequency diagram with each dynamic interference signal removed;

[0029] The target object recognition module is used to generate a radar detection image of the target object according to the third time-frequency diagram, and identify the position information of the target object according to the radar detection image.

[0030] Further, it further includes a model construction module;

[0031] The model construction module is used to obtain a plurality of time-frequency diagram samples containing dynamic interference signals; wherein, each time-frequency diagram sample is labeled with an energy sudden increase feature, a frequency band offset feature, and a dynamic interference signal;

[0032] Construct an initial dynamic interference signal separation model, using the time-frequency diagram samples as the input of the initial dynamic interference signal separation model and the time-frequency diagram samples with dynamic interference signals removed as the output, and iteratively train the initial dynamic interference signal separation model until the initial dynamic interference signal separation model converges to generate a dynamic interference signal separation model; wherein, in each iterative training process, the initial dynamic interference signal separation model extracts the energy sudden increase feature and frequency band offset feature in the time-frequency diagram samples, and determines the time-frequency diagram samples in the time-frequency diagram samples according to the energy sudden increase feature and frequency band offset feature in the time-frequency diagram samples.

[0033] Further, the generating a radar detection image of the target object according to the third time-frequency diagram and identifying the position information of the target object according to the radar detection image includes:

[0034] Generate a radar detection image of the target object according to the third time-frequency diagram;

[0035] Convert the third time-frequency diagram into a binary representation to obtain the binary data of the third time-frequency diagram;

[0036] Process the binary data using a connected component labeling algorithm to determine several connected components of the binary data;

[0037] Determine the position feature information of the target object on each connected component according to each connected component;

[0038] Determine the position information of the target object on the radar detection image according to the position feature information on each connected component.

[0039] Further, the performing a short-time Fourier transform process on the electromagnetic wave detection signal to generate a first time-frequency diagram corresponding to the electromagnetic wave detection signal includes:

[0040] Divide the electromagnetic wave detection signal into several first signal time segments, and process each first signal time segment using a preset window function to obtain each second signal time segment;

[0041] Process each second signal time segment using a fast Fourier transform algorithm to obtain the spectrum of each second signal time segment;

[0042] Combine the spectra of each second signal time segment to obtain the first time-frequency diagram.

[0043] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements an object recognition method described in the above-mentioned invention embodiment.

[0044] Another embodiment of the present invention provides a storage medium, which includes a stored computer program. When the computer program runs, it controls the device where the storage medium is located to execute an object recognition method described in the above-mentioned embodiment of the present invention.

[0045] By implementing the present invention, the following beneficial effects are achieved:

[0046] The present invention provides an object recognition method, device, terminal device and storage medium. After obtaining an electromagnetic wave detection signal and obtaining an electromagnetic wave detection signal including a target signal, several fixed interference signals and several dynamic interference signals, the fixed interference signals in the electromagnetic wave detection signal are removed by short-time Fourier transform, and the dynamic interference signal separation model is further used to extract features corresponding to the dynamic interference signals such as energy sudden increase features and frequency band offset features, and the dynamic interference signals are removed, so that the finally generated radar detection image including the target object does not include fixed interference signals and dynamic interference signals. Furthermore, the position information of the target object can be conveniently and accurately recognized according to the radar detection image. This solves the problem that the radar detection image obtained by detecting the target object with a ground penetrating radar in the prior art contains interference signal waveforms, which affects the recognition of the target object in the radar detection image, and improves the accuracy and convenience of obtaining the position information of the target object when the ground penetrating radar detects the target object. It more efficiently assists the user of the ground penetrating radar to obtain the position information of the required target object. Description of the Drawings

[0047] Figure 1 is a schematic flow chart of an object recognition method provided by an embodiment of the present invention.

[0048] Figure 2 is a schematic structural diagram of an object recognition device provided by an embodiment of the present invention. Detailed Embodiments

[0049] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present invention.

[0050] To make the objectives, technical solutions and advantages of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are a part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the scope of protection of the present application.

[0051] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field to which this application belongs; the terms used herein are for the purpose of describing specific embodiments only and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above description of the drawings are intended to cover non-exclusive inclusion.

[0052] In the description of the embodiments of this application, technical terms such as "first" and "second" are only used to distinguish different objects and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity, specific order or primary-secondary relationship of the indicated technical features. In the description of the embodiments of this application, the meaning of "a plurality" is more than two, unless otherwise specifically defined.

[0053] Reference to "embodiment" herein means that a particular feature, structure or characteristic described in connection with the embodiment can be included in at least one embodiment of this application. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0054] In the description of the embodiments of this application, the term "and / or" is merely a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " herein generally represents an "or" relationship between the associated objects before and after.

[0055] In the description of the embodiments of this application, the term "a plurality" refers to more than two (including two). Similarly, "multiple groups" refers to more than two groups (including two groups), and "multiple pieces" refers to more than two pieces (including two pieces).

[0056] In the description of the embodiments of this application, unless otherwise clearly specified and limited, technical terms such as "installation", "connection", "connection", "fixation" and the like should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or integrated; it can also be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and can be the communication inside two elements or the interaction relationship between two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the embodiments of this application can be understood according to specific circumstances.

[0057] See Figure 1, to solve the problem that the radar detection image obtained by detecting a target object with a ground penetrating radar in the prior art contains interference signal waveforms, which affects the recognition of the target object in the radar detection image, an object recognition method provided by an embodiment of the present invention includes:

[0058] Step S1: Detect a target object with a ground penetrating radar to obtain an electromagnetic wave detection signal; wherein, the electromagnetic wave detection signal includes a target signal, a plurality of fixed interference signals, and a plurality of dynamic interference signals;

[0059] Step S2: Perform short-time Fourier transform processing on the electromagnetic wave detection signal to generate a first time-frequency diagram corresponding to the electromagnetic wave detection signal;

[0060] Step S3: Identify each non-target frequency band corresponding to each fixed interference signal in the first time-frequency diagram, and mask each non-target frequency band to obtain a second time-frequency diagram with each non-target frequency band removed;

[0061] Step S4: Input the second time-frequency diagram into a dynamic interference signal separation model, so that the dynamic interference signal separation model extracts the energy sudden increase feature and frequency band offset feature in the second time-frequency diagram, determines each dynamic interference signal in the second time-frequency diagram according to the energy sudden increase feature and frequency band offset feature, and outputs a third time-frequency diagram with each dynamic interference signal removed;

[0062] Step S5: Generate a radar detection image of the target object according to the third time-frequency diagram, and identify the position information of the target object according to the radar detection image.

[0063] Regarding step S1, based on its ability to perform non-destructive detection of underground invisible media, ground penetrating radar has been widely used in scenarios such as roads, bridges, tunnels, and pipelines that require non-destructive measurement. Taking a road as an example, taking the underground pipeline in the road as the target object, when detecting the position of the underground pipeline in the road, the position information of the underground pipeline can be determined based on the ground penetrating radar, thereby assisting underground construction. However, the underground environment is complex. When the ground penetrating radar performs detection, it will be affected by underground media, external environmental interference signals, and ground debris. These influencing factors will also receive the electromagnetic wave detection signal sent by the ground penetrating radar during the detection of the underground pipeline by the ground penetrating radar, and reflect the corresponding feedback signal to the ground penetrating radar according to the electromagnetic wave signal sent by the ground penetrating radar. These feedback signals are the interference signals that affect the detection of the underground pipeline position. That is, when detecting the underground pipeline with a ground penetrating radar, the obtained electromagnetic wave detection signal will include the target signal corresponding to the underground pipeline, as well as the interference signals brought by the remaining reflections.

[0064] For these interference signals, they can be divided into fixed interference signals and dynamic interference signals. Among them, for the interference signals generated by ground debris such as metal manhole covers and gravel, the reflected signals have fixed frequency characteristics, and these signals are regarded as fixed interference signals. External environmental interference signals, such as the interference signals generated by the ground penetrating radar equipment itself, are also quantifiable and can be regarded as fixed interference signals. For the interference signals caused by geological features of underground media such as underground soil humidity, groundwater flow, and underground rock layers, they do not have fixed frequency characteristics or regular changes, and the interference signals generated by their reflection are regarded as dynamic interference signals.

[0065] For step S2, perform short-time Fourier transform processing on the electromagnetic wave detection signal obtained in step S1 to convert it into a first time-frequency diagram to assist in removing fixed interference signals in step S3.

[0066] In a preferred embodiment, the performing short-time Fourier transform processing on the electromagnetic wave detection signal to generate a first time-frequency diagram corresponding to the electromagnetic wave detection signal includes: dividing the electromagnetic wave detection signal into several first signal time segments, and processing each first signal time segment with a preset window function to obtain each second signal time segment; using the fast Fourier transform algorithm to process each second signal time segment to obtain the spectrum of each second signal time segment; combining the spectra of each second signal time segment to obtain the first time-frequency diagram.

[0067] Specifically, the short-time Fourier transform (STFT) is a method that combines time-domain analysis and frequency-domain analysis. STFT can not only reflect the components of each frequency in the electromagnetic wave detection signal but also reflect the law of change of each frequency component over time. The processing principle formula of STFT can be expressed as:

[0068]

[0069] x(t) ∈ L 2 (R) w(t) ∈ L 2 (R) t × w(t) ∈ L 2 (R);

[0070] Among them, x(t) represents the input electromagnetic wave detection signal; w(t) represents the preset window function; STFT x (t, Ω) represents the short-time Fourier transform of x(t) at time t and frequency Ω; τ represents the integration variable; e represents the base of the natural logarithm; j is the imaginary unit; x(t) ∈ L 2 (R), w(t) ∈ L 2 (R) and t × w(t) ∈ L 2(R) indicates that x(t), w(t), and t×w(t) are square integrable.

[0071] Through the above Fourier transform algorithm, the electromagnetic wave detection signal can be segmented into several first signal time segments, and each of the segmented first signal time segments is a plurality of short-time windows that can overlap. Applying a preset window function to each first signal time segment for processing, each second signal time segment is obtained. Using the fast Fourier transform algorithm FFT to process each second signal time segment, the spectrum of each second signal time segment is obtained, and the spectra of each second signal time segment are combined to obtain the first time-frequency diagram.

[0072] Preferably, in practical applications, in addition to the short-time Fourier transform algorithm, the discrete short-time Fourier transform algorithm can also be used for processing, and the processing principle formula of the discrete short-time Fourier algorithm can be expressed as:

[0073]

[0074] where x(n) represents the discrete-time signal, corresponding to the electromagnetic wave detection signal; w(n) is the window function, used to localize the signal in time; n represents the current time point; e -jΩm is the complex exponential function, used to transform the signal from the time domain to the frequency domain; Ω represents the continuous angular frequency; STFT x (n,Ω) represents the discrete short-time Fourier transform of x(n) at the current time point n and frequency Ω; m is the time index, representing the time point in the discrete-time signal x(m).

[0075] To discretize the continuous frequency, a discrete frequency index r is defined, where r = 0, 1,..., N - 1, N is the number of sampling points, and the discrete angular frequency Ω corresponding to the discrete frequency index r is defined r = 2πr / N. Substituting the above continuous sampling angular frequency Ω with the discrete angular frequency Ω r , the following formula can be obtained:

[0076]

[0077] Let k = n - m, m = n - k in the above formula, where the value range of m is from to the value range of k is from to It can be further transformed to obtain:

[0078]

[0079] Let then there is:

[0080] The first time-frequency diagram after discrete short-time Fourier transform processing can be obtained through the above processing.

[0081] For step S3, identify each non-target frequency band corresponding to the fixed interference signal in the obtained first time-frequency diagram. For the non-target frequency bands corresponding to the fixed signal, the non-target frequency bands corresponding to each fixed signal can be obtained in advance, and compared with each frequency band in the first time-frequency diagram, so as to determine each non-target frequency band corresponding to each fixed interference signal, and set the elements corresponding to each non-target frequency band in the first time-frequency diagram to zero or a minimum value to mask the fixed interference signals corresponding to these non-target frequency bands, thereby obtaining the second time-frequency diagram.

[0082] For step S4, input the second time-frequency diagram into the dynamic interference signal separation model. The dynamic interference signal separation model extracts the energy sudden increase feature and the frequency band shift feature in the second time-frequency diagram. Among them, the energy sudden increase feature includes the energy sudden increase amplitude, the duration, and the position, and the frequency band shift feature includes the frequency band shift size and the frequency band shift direction. Through these features, each dynamic interference signal with abnormal changes in the second time-frequency diagram is determined. After classifying and filtering each dynamic interference signal, the third time-frequency diagram without interference signals can be obtained.

[0083] In a preferred embodiment, the construction of the dynamic interference signal separation model includes: obtaining a plurality of time-frequency diagram samples containing dynamic interference signals; wherein, each time-frequency diagram sample is labeled with the energy sudden increase feature, the frequency band shift feature, and the dynamic interference signal; constructing an initial dynamic interference signal separation model, using the time-frequency diagram sample as the input of the initial dynamic interference signal separation model, and the time-frequency diagram sample after removing the dynamic interference signal as the output, and performing iterative training on the initial dynamic interference signal separation model until the initial dynamic interference signal separation model converges, generating the dynamic interference signal separation model; wherein, in each iterative training process, the initial dynamic interference signal separation model extracts the energy sudden increase feature and the frequency band shift feature in the time-frequency diagram sample, and determines the time-frequency diagram sample in the time-frequency diagram sample according to the energy sudden increase feature and the frequency band shift feature in the time-frequency diagram sample.

[0084] Specifically, the initial dynamic interference signal separation model adopts a convolutional neural network model (CNN model) in the present invention. The condition for the initial dynamic interference signal separation model to converge is that the model iteration is lower than the loss threshold, and the loss threshold is taken as 0.001 in the present invention. In each iterative training, the initial dynamic interference signal separation model outputs a predicted time-frequency diagram sample after removing the dynamic interference signal according to the input time-frequency diagram sample, compares the predicted time-frequency diagram sample with the time-frequency diagram sample after removing the labeled dynamic interference signal in the input, and then the prediction error can be determined. According to the prediction error, the current iteration loss is determined, and the adjustment of the model parameters of the initial dynamic interference signal separation model is guided according to the current iteration loss.

[0085] For step S5, generate a radar detection image of the target object based on the third time-frequency diagram, and identify the position information of the target object according to the radar detection image.

[0086] In a preferred embodiment, the generating a radar detection image of the target object based on the third time-frequency diagram and identifying the position information of the target object according to the radar detection image includes: generating a radar detection image of the target object based on the third time-frequency diagram; converting the third time-frequency diagram into a binary representation to obtain the binary data of the third time-frequency diagram; processing the binary data using a connected component labeling algorithm to determine several connected components of the binary data; determining the position feature information of the target object on each connected component according to each connected component; and determining the position information of the target object on the radar detection image according to the position feature information on each connected component.

[0087] Specifically, when generating a radar detection image of the target object based on the third time-frequency diagram, the position of the target object in the radar detection image has not been framed in the radar detection image at this time. To accurately identify the position information of the target object, it is necessary to convert the third time-frequency diagram into a binary representation to obtain the binary data of the third time-frequency diagram. Converting to binary data is to support the processing of the connected component labeling algorithm. The binary data conversion method can use the Otsu algorithm.

[0088] It should be noted that the principle of the connected component labeling algorithm is that if the data p has two horizontal neighbors and two vertical neighbors at the coordinates (x, y), and their coordinates are (x + 1, y), (x - 1, y), (x, y + 1), and (x, y - 1) respectively, the set of these 4 neighbors of p is denoted as N4(p). The coordinates of the four diagonal neighbors of p are (x + 1, y + 1), (x + 1, y - 1), (x - 1, y + 1), and (x - 1, y - 1) respectively, and the set of these 4 neighbors of p and N4(p) is denoted as N8(p). If q ∈ N4(p), then it is said that the data p and q are 4-adjacent. Similarly, if q ∈ N8(p), then it is said that p and q are 8-adjacent. If every two data on a path are 4-adjacent or 8-adjacent, then this path is a 4-connected path or an 8-connected path, and it is said that these two data are 4-connected or 8-connected. For any data p, the set of all data connected (4-connected or 8-connected) to it is called the connected component containing p.

[0089] Based on the above connected component labeling algorithm, the binary data is processed. The connected path rules are defined using 4-neighborhood (adjacent up, down, left, and right) or 8-neighborhood (including diagonal adjacency). The unlabeled pixels are traversed using breadth-first search (BFS) or depth-first search (DFS) to label the connected components of the pixels one by one, generating a labeled matrix with the connected components labeled. The features of the target object and the non-target object are distinguished through this labeled matrix, and then the convenient description of the target object is calculated, and the position of the target object is further determined, so as to obtain the position information of the target object on the radar detection image and complete the recognition of the position information of the target object.

[0090] Based on the above method embodiment, the present invention correspondingly provides an apparatus embodiment.

[0091] As Figure 2 shown, an embodiment of the present invention provides an object recognition apparatus, including: a detection signal acquisition module, a fixed interference signal removal module, a dynamic interference signal removal module, and a target object recognition module;

[0092] The detection signal acquisition module is used to detect a target object through a ground penetrating radar to obtain an electromagnetic wave detection signal; wherein, the electromagnetic wave detection signal includes a target signal, a plurality of fixed interference signals, and a plurality of dynamic interference signals;

[0093] The fixed interference signal removal module is used to perform short-time Fourier transform processing on the electromagnetic wave detection signal to generate a first time-frequency diagram corresponding to the electromagnetic wave detection signal; identify each non-target frequency band corresponding to each fixed interference signal in the first time-frequency diagram, and mask each non-target frequency band to obtain a second time-frequency diagram with each non-target frequency band removed;

[0094] The dynamic interference signal removal module is used to input the second time-frequency diagram into a dynamic interference signal separation model, so that the dynamic interference signal separation model extracts the energy sudden increase feature and frequency band offset feature in the second time-frequency diagram, determines each dynamic interference signal in the second time-frequency diagram according to the energy sudden increase feature and frequency band offset feature, and outputs a third time-frequency diagram with each dynamic interference signal removed;

[0095] The target object recognition module is used to generate a radar detection image of the target object according to the third time-frequency diagram and identify the position information of the target object according to the radar detection image.

[0096] In a preferred embodiment, a model construction module is further included;

[0097] The model construction module is used to obtain a plurality of time-frequency diagram samples containing dynamic interference signals; wherein, each time-frequency diagram sample is labeled with an energy sudden increase feature, a frequency band offset feature, and a dynamic interference signal;

[0098] Construct an initial dynamic interference signal separation model, use the time-frequency diagram sample as the input of the initial dynamic interference signal separation model, and the time-frequency diagram sample after removing the dynamic interference signal as the output. Iteratively train the initial dynamic interference signal separation model until the initial dynamic interference signal separation model converges to generate a dynamic interference signal separation model. Wherein, in each iterative training process, the initial dynamic interference signal separation model extracts the energy sudden increase feature and the frequency band offset feature in the time-frequency diagram sample, and determines the time-frequency diagram sample in the time-frequency diagram sample according to the energy sudden increase feature and the frequency band offset feature in the time-frequency diagram sample.

[0099] In a preferred embodiment, the generating a radar detection image of the target object according to the third time-frequency diagram and identifying the position information of the target object according to the radar detection image includes:

[0100] Generate a radar detection image of the target object according to the third time-frequency diagram;

[0101] Convert the third time-frequency diagram into a binary representation to obtain the binary data of the third time-frequency diagram;

[0102] Process the binary data using a connected component labeling algorithm to determine several connected components of the binary data;

[0103] Determine the position feature information of the target object on each connected component according to each connected component;

[0104] Determine the position information of the target object on the radar detection image according to the position feature information on each connected component.

[0105] In a preferred embodiment, the performing a short-time Fourier transform process on the electromagnetic wave detection signal to generate a first time-frequency diagram corresponding to the electromagnetic wave detection signal includes:

[0106] Divide the electromagnetic wave detection signal into several first signal time segments, and process each first signal time segment using a preset window function to obtain each second signal time segment;

[0107] Process each second signal time segment using a fast Fourier transform algorithm to obtain the spectrum of each second signal time segment;

[0108] Combine the spectra of each second signal time segment to obtain the first time-frequency diagram.

[0109] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the attached drawings of the device embodiments provided by the present invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art can understand and implement without creative efforts.

[0110] Those skilled in the art can clearly understand that for the convenience and simplicity, the specific working process of the device described above can refer to the corresponding process in the foregoing method embodiments, and will not be elaborated here.

[0111] Based on the above method item embodiments, the present invention correspondingly provides terminal device item embodiments.

[0112] An embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements an object recognition method described in any one of the present invention.

[0113] The terminal device can be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The terminal device may include, but is not limited to, a processor and a memory.

[0114] The so-called processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), 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 also be any conventional processor, etc. The processor is the control center of the terminal device, and connects various parts of the entire terminal device through various interfaces and lines.

[0115] The memory can be used to store the computer program. By running or executing the computer program stored in the memory and invoking the data stored in the memory, the processor realizes various functions of the terminal device. The memory mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include high-speed random access memory and can also include non-volatile memory, such as a hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage devices.

[0116] Based on the above method embodiment, the present invention correspondingly provides a storage medium embodiment.

[0117] An embodiment of the present invention provides a storage medium, which includes a stored computer program. When the computer program runs, it controls the device where the storage medium is located to execute an object recognition method described in any one of the present invention.

[0118] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of the above various method embodiments. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0119] The above is the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements are also regarded as the protection scope of the present invention.

Claims

1. An object recognition method, characterized in that, Including: Detecting a target object by a ground penetrating radar to obtain an electromagnetic wave detection signal; wherein, the electromagnetic wave detection signal includes a target signal, a plurality of fixed interference signals and a plurality of dynamic interference signals; Performing short-time Fourier transform processing on the electromagnetic wave detection signal to generate a first time-frequency diagram corresponding to the electromagnetic wave detection signal; Identifying each non-target frequency band corresponding to each fixed interference signal in the first time-frequency diagram, and masking each non-target frequency band to obtain a second time-frequency diagram with each non-target frequency band removed; Inputting the second time-frequency diagram into a dynamic interference signal separation model, so that the dynamic interference signal separation model extracts an energy sudden increase feature and a frequency band shift feature in the second time-frequency diagram, determines each dynamic interference signal in the second time-frequency diagram according to the energy sudden increase feature and the frequency band shift feature, and outputs a third time-frequency diagram with each dynamic interference signal removed; Generating a radar detection image of the target object according to the third time-frequency diagram, and identifying the position information of the target object according to the radar detection image.

2. The object recognition method according to claim 1, wherein, The construction of the dynamic interference signal separation model includes: Obtaining a plurality of time-frequency diagram samples containing dynamic interference signals; wherein, each time-frequency diagram sample is labeled with an energy sudden increase feature, a frequency band shift feature and a dynamic interference signal; Constructing an initial dynamic interference signal separation model, using the time-frequency diagram sample as the input of the initial dynamic interference signal separation model, and using the time-frequency diagram sample with the dynamic interference signal removed as the output, and performing iterative training on the initial dynamic interference signal separation model until the initial dynamic interference signal separation model converges, generating a dynamic interference signal separation model; wherein, in each iterative training process, the initial dynamic interference signal separation model extracts the energy sudden increase feature and the frequency band shift feature in the time-frequency diagram sample, and determines the time-frequency diagram sample in the time-frequency diagram sample according to the energy sudden increase feature and the frequency band shift feature in the time-frequency diagram sample.

3. The object recognition method according to claim 2, wherein The generating a radar detection image of the target object according to the third time-frequency diagram, and identifying the position information of the target object according to the radar detection image includes: Generating a radar detection image of the target object according to the third time-frequency diagram; Converting the third time-frequency diagram into a binary representation to obtain the binary data of the third time-frequency diagram; Processing the binary data by using a connected component labeling algorithm to determine a plurality of connected components of the binary data; Determining the position feature information of the target object on each connected component according to each connected component; Determining the position information of the target object on the radar detection image according to the position feature information on each connected component.

4. The object recognition method according to claim 3, characterized in that, The performing short-time Fourier transform processing on the electromagnetic wave detection signal to generate a first time-frequency diagram corresponding to the electromagnetic wave detection signal includes: Dividing the electromagnetic wave detection signal into a plurality of first signal time segments, and processing each first signal time segment by using a preset window function to obtain each second signal time segment; Processing each second signal time segment by using a fast Fourier transform algorithm to obtain the spectrum of each second signal time segment; Combining the spectra of each second signal time segment to obtain the first time-frequency diagram.

5. An object recognition device, characterized in that, Including: A detection signal acquisition module, a fixed interference signal removal module, a dynamic interference signal removal module, and a target object recognition module; The detection signal acquisition module is used to detect a target object through a ground penetrating radar to obtain an electromagnetic wave detection signal; wherein, the electromagnetic wave detection signal includes a target signal, a plurality of fixed interference signals, and a plurality of dynamic interference signals; The fixed interference signal removal module is used to perform short-time Fourier transform processing on the electromagnetic wave detection signal to generate a first time-frequency diagram corresponding to the electromagnetic wave detection signal; identify each non-target frequency band corresponding to each fixed interference signal in the first time-frequency diagram, and mask each non-target frequency band to obtain a second time-frequency diagram with each non-target frequency band removed; The dynamic interference signal removal module is used to input the second time-frequency diagram into a dynamic interference signal separation model, so that the dynamic interference signal separation model extracts the energy sudden increase feature and the frequency band offset feature in the second time-frequency diagram, determines each dynamic interference signal in the second time-frequency diagram according to the energy sudden increase feature and the frequency band offset feature, and outputs a third time-frequency diagram with each dynamic interference signal removed; The target object recognition module is used to generate a radar detection image of the target object according to the third time-frequency diagram, and identify the position information of the target object according to the radar detection image.

6. The object recognition device according to claim 5, characterized in that, It further includes a model construction module; The model construction module is used to obtain a plurality of time-frequency diagram samples containing dynamic interference signals; wherein, each time-frequency diagram sample is labeled with an energy sudden increase feature, a frequency band offset feature, and a dynamic interference signal; Construct an initial dynamic interference signal separation model, use the time-frequency diagram samples as the input of the initial dynamic interference signal separation model, and use the time-frequency diagram samples with dynamic interference signals removed as the output, and perform iterative training on the initial dynamic interference signal separation model until the initial dynamic interference signal separation model converges, and generate a dynamic interference signal separation model; wherein, in each iterative training process, the initial dynamic interference signal separation model extracts the energy sudden increase feature and the frequency band offset feature in the time-frequency diagram sample, and determines the time-frequency diagram sample in the time-frequency diagram sample according to the energy sudden increase feature and the frequency band offset feature in the time-frequency diagram sample.

7. The object recognition device according to claim 6, wherein The generating a radar detection image of the target object according to the third time-frequency diagram and identifying the position information of the target object according to the radar detection image includes: Generating a radar detection image of the target object according to the third time-frequency diagram; Converting the third time-frequency diagram into a binary representation to obtain the binary data of the third time-frequency diagram; Processing the binary data by using a connected component labeling algorithm to determine a plurality of connected components of the binary data; Determining the position feature information of the target object on each connected component according to each connected component; Determining the position information of the target object on the radar detection image according to the position feature information on each connected component.

8. The object recognition device according to claim 7, wherein The performing short-time Fourier transform processing on the electromagnetic wave detection signal to generate a first time-frequency diagram corresponding to the electromagnetic wave detection signal includes: Divide the electromagnetic wave detection signal into a plurality of first signal time segments, and apply a preset window function to each first signal time segment for processing to obtain each second signal time segment; Process each second signal time segment using the fast Fourier transform algorithm to obtain the frequency spectrum of each second signal time segment; Combine the frequency spectra of each second signal time segment to obtain the first time-frequency diagram.

9. A terminal device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements an object recognition method according to any one of claims 1 to 4.

10. A storage medium, characterized in that, The storage medium includes a stored computer program, wherein when the computer program runs, it controls the device where the storage medium is located to execute an object recognition method according to any one of claims 1 to 4.

Images