Bird detection and identification method and device based on ultra-wideband impulse radar

Abstract

The invention belongs to the technical field of intelligent detection, and discloses a bird detection and identification method and device based on an ultra wide band impulse radar, and the method comprises the steps that a display control module sends a work starting instruction to a control module; the control module sends a trigger signal to the excitation signal generator to enable the excitation signal generator to generate a pulse width modulation signal; the avalanche transistor Marx circuit generates a zero-order Gaussian pulse signal and sends the zero-order Gaussian pulse signal to the transmitting antenna array; the transmitting antenna array generates a detection signal; the receiving module sends the processed echo signal to the ADC module; the ADC module generates a digital signal and sends the digital signal to the control module; the control module generates a bird recognition result and sends the result to the display control module; and the display control module displays the bird recognition result. The high-power avalanche transistor Marx circuit is designed to serve as the pulse source of the ultra-wideband impulse radar, the signal-to-noise ratio of echo signals is effectively increased, a deep learning model is introduced to process echoes, and the bird target recognition capability with complex reflection characteristics is enhanced.

Description

Technical Field

[0001] This application relates to the field of intelligent detection technology, and particularly to a method and device for detecting and identifying birds based on an ultra-wideband impulse radar. Background Art

[0002] Existing ultra-wideband impulse radar systems are mainly developed for short-range detection scenarios such as urban street fighting, geological exploration, and biomedicine. When facing targets such as flying birds that are small in size and high in flight altitude, due to the limited emission power of the pulse source, the signal-to-noise ratio of the echo is very low and it is difficult to extract target features. In addition, existing ultra-wideband impulse radars usually adopt the architecture of FPGA+PC processors, with a complex system structure and difficult radar deployment, and it is impossible to conduct long-term monitoring of bird migration in the wild. Although some technologies have proposed applying the ZYNQ SoC integrating ARM and FPGA to the ultra-wideband impulse radar system to improve the system integration, its main function is still limited to traditional signal processing tasks and fails to combine a specifically developed deep learning algorithm module. This defect makes the radar unable to quickly and accurately detect and track flying birds in a complex natural environment, nor can it achieve accurate identification of bird species, and cannot meet the research needs of bird migration behavior. Summary of the Invention

[0003] This application provides a method and device for detecting and identifying birds based on an ultra-wideband impulse radar. By designing a high-power avalanche transistor Marx circuit as the pulse source of the ultra-wideband impulse radar, the signal-to-noise ratio of the echo signal of the ultra-wideband impulse radar is effectively improved. An ultra-wideband radar signal processing model is designed to enhance the feature extraction ability and real-time signal processing ability of the ultra-wideband impulse radar. Through the operation of the built-in model, efficient acquisition, processing, and target recognition of the ultra-wideband radar echo signal are realized, the overall performance and application value of the ultra-wideband impulse radar in the application scenarios of bird migration tracking and identification are improved, and the ability to identify bird targets with small size and complex reflection characteristics is enhanced.

[0004] In the first aspect, an embodiment of this application provides a method for detecting and identifying birds based on an ultra-wideband impulse radar, and the method includes:

[0005] The display and control module sends a start working instruction to the control module;

[0006] The control module receives the start working instruction and sends a trigger signal to the excitation signal generator;

[0007] The excitation signal generator receives the trigger signal and generates a pulse width modulation signal to send to the avalanche transistor Marx circuit;

[0008] The avalanche transistor Marx circuit receives the pulse width modulation signal, generates a zero-order Gaussian pulse signal and sends it to the transmitting antenna array;

[0009] The transmitting antenna array receives a zero-order Gaussian pulse signal, generates a detection signal and transmits it;

[0010] The receiving module receives and processes the echo signal, and sends the processed echo signal to the ADC module;

[0011] The ADC module converts the processed echo signal, generates a digital signal and sends it to the control module;

[0012] The control module receives and processes the digital signal, generates a bird recognition result and sends it to the display and control module;

[0013] The display and control module receives the bird recognition result and displays it.

[0014] Furthermore, the control module includes an input data buffer module, a data preprocessing module, a target recognition and classification module, and an output data buffer module;

[0015] The control module receives and processes the digital signal, generates a bird recognition result, including:

[0016] The input data buffer module buffers the received digital signal and sends it to the data preprocessing module;

[0017] The data preprocessing module processes the digital signal, obtains target data and sends it to the target recognition and classification module;

[0018] The target recognition and classification module recognizes the target data, obtains a bird recognition result and sends it to the output data buffer module;

[0019] The output data buffer module buffers the bird recognition result and sends it to the display and control module.

[0020] Furthermore, the data preprocessing module includes a coherent accumulation unit, an accumulation and average background cancellation unit, a detection algorithm unit, a Fourier transform unit, and an image slicing unit;

[0021] The data preprocessing module processes the digital signal, obtains target data, including:

[0022] The coherent accumulation unit receives the detection signal, processes the detection signal according to the FPGA parallel accumulator, obtains a first result signal and sends it to the accumulation and average background cancellation unit;

[0023] The accumulation and average background cancellation unit receives the first result signal, filters out the static clutter of the first result signal, obtains a second result signal and sends it to the detection algorithm unit;

[0024] The detection algorithm unit receives the second result signal, detects the second result signal according to the OS CFAR algorithm, obtains the first target information, and sends it to the Fourier transform unit;

[0025] The Fourier transform unit receives the first target information, performs Doppler processing on the first target information based on the fast Fourier transform, obtains the second target information and sends it to the image slicing unit;

[0026] The image slicing unit receives the second target information, performs data slicing operations on the second target information, obtains the target data and sends it to the target recognition and classification module.

[0027] Furthermore, the target recognition and classification module includes a patch embedding unit, a flattening and linear mapping unit, an encoder unit, a channel fusion unit, and a multi-layer perception function unit;

[0028] The target recognition and classification module recognizes the target data and obtains the bird recognition result, including:

[0029] The patch embedding unit arranges and splices the received target data in a preset order and preset direction to obtain multi-dimensional data, and sends it to the flattening and linear mapping unit;

[0030] The flattening and linear mapping unit converts the received multi-dimensional data into a one-dimensional vector, performs a linear transformation on the one-dimensional vector using a fully connected layer, obtains the linear vector and sends it to the encoder unit;

[0031] The encoder unit analyzes the linear vector to obtain high-frequency components and low-frequency components; processes the high-frequency components to obtain the first feature, processes the low-frequency components to obtain the second feature; sends the first feature and the second feature to the channel fusion unit;

[0032] The channel fusion unit receives and combines the first feature and the second feature, obtains the fusion signal and sends it to the multi-layer perception function unit;

[0033] The multi-layer perception function unit receives and processes the fusion signal using a neural network to obtain the bird recognition result.

[0034] Furthermore, the encoder unit includes a global pooling layer;

[0035] The encoder unit analyzes the linear vector to obtain high-frequency components and low-frequency components, including:

[0036] The global pooling layer extracts the low-frequency components from the linear vector; subtracts the low-frequency components from the linear vector to obtain the high-frequency components.

[0037] Furthermore, the receiving module includes a receiving antenna, a low-noise amplifier unit, and a sampling unit; the receiving module receives and processes the echo signal, including:

[0038] The receiving antenna receives the echo signal and sends it to the low-noise amplifier unit;

[0039] The low-noise amplifier unit receives and processes the echo signal to obtain the first echo signal;

[0040] The sampling unit receives and samples the first echo signal to obtain a processed echo signal.

[0041] Furthermore, the repetition frequency of the pulse width modulation signal is 50 kHz, the amplitude is 6 V, and the duty cycle is 5%.

[0042] In a second aspect, an embodiment of the present application provides a bird detection and identification device based on an ultra-wideband impulse radar. The device includes:

[0043] A display and control module for sending a start work instruction to the control module; receiving and displaying the bird identification result;

[0044] A control module for receiving the start work instruction, sending a trigger signal to the excitation signal generator; receiving and processing the digital signal, generating a bird identification result and sending it to the display and control module;

[0045] A transmitting module includes an excitation signal generator, an avalanche transistor Marx circuit, and a transmitting antenna array; the excitation signal generator is used to receive the trigger signal, generate a pulse width modulation signal and send it to the avalanche transistor Marx circuit; the avalanche transistor Marx circuit is used to receive the pulse width modulation signal, generate a zero-order Gaussian pulse signal and send it to the transmitting antenna array; the transmitting antenna array is used to receive the zero-order Gaussian pulse signal, generate a detection signal and transmit it;

[0046] A receiving module for receiving and processing the echo signal, and sending the processed echo signal to the ADC module;

[0047] The ADC module is used to convert the processed echo signal, generate a digital signal and send it to the control module.

[0048] Furthermore, the control module includes an input data buffer module, a data preprocessing module, a target recognition and classification module, and an output data buffer module;

[0049] The input data buffer module is used to buffer the received digital signal and send it to the data preprocessing module;

[0050] The data preprocessing module is used to process the digital signal to obtain target data and send it to the target recognition and classification module;

[0051] The target recognition and classification module is used to identify the target data, obtain the bird identification result and send it to the output data buffer module;

[0052] The output data buffer module is used to buffer the bird identification result and send it to the display and control module.

[0053] Further, the data preprocessing module includes a coherent accumulation unit, an accumulated average background cancellation unit, a detection algorithm unit, a Fourier transform unit, and an image slicing unit;

[0054] The coherent accumulation unit is configured to receive the detection signal, process the detection signal according to the FPGA parallel accumulator, obtain a first result signal, and send it to the accumulated average background cancellation unit;

[0055] The accumulated average background cancellation unit is configured to receive the first result signal, filter out the static clutter of the first result signal, obtain a second result signal, and send it to the detection algorithm unit;

[0056] The detection algorithm unit is configured to receive the second result signal, detect the second result signal according to the OS CFAR algorithm, obtain first target information, and send it to the Fourier transform unit;

[0057] The Fourier transform unit is configured to receive the first target information, perform Doppler processing on the first target information based on the fast Fourier transform, obtain second target information, and send it to the image slicing unit;

[0058] The image slicing unit is configured to receive the second target information, perform data slicing on the second target information, obtain target data, and send it to the target recognition and classification module.

[0059] In summary, compared with the prior art, the beneficial effects brought by the technical solution provided by the embodiment of the present application at least include:

[0060] A method for detecting and identifying birds based on an ultra-wideband impulse radar provided by an embodiment of the present application. By adding a high-power avalanche transistor Marx circuit as the pulse source of the ultra-wideband impulse radar, the signal-to-noise ratio of the echo signal of the ultra-wideband impulse radar is effectively improved. An ultra-wideband radar signal processing model is designed to enhance the feature extraction ability and real-time signal processing ability of the ultra-wideband impulse radar. Through the operation of the built-in model, efficient acquisition, processing, and target recognition of the ultra-wideband radar echo signal are realized, improving the overall performance and application value of the ultra-wideband impulse radar in the application scenarios of bird migration tracking and identification, and enhancing the ability to identify bird targets with small volume and complex reflection characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a flowchart of a method for detecting and identifying birds based on an ultra-wideband impulse radar provided by an exemplary embodiment of the present application.

[0062] Figure 2 It is a structural diagram of a device for detecting and identifying birds based on an ultra-wideband impulse radar provided by an exemplary embodiment of the present application.

[0063] Figure 3The structural diagram of a bird detection and recognition device based on an ultra-wideband impulse radar provided for another exemplary embodiment of the present application.

[0064] Figure 4 The structural diagram of a bird detection and recognition device based on an ultra-wideband impulse radar provided for yet another exemplary embodiment of the present application. Detailed implementation manners

[0065] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments.

[0066] All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0067] Please refer to Figure 1 , the embodiments of the present application provide a bird detection and recognition method based on an ultra-wideband impulse radar, and the method specifically includes the following steps:

[0068] Step S1, the display and control module sends a start work instruction to the control module.

[0069] Among them, the display and control module sends the start work instruction to the PS end of the control module through the RS485 interface, transmits the command to the PL port through the AXI bus in the control module, and generates a trigger signal by the GPIO port at the PL end.

[0070] Step S2, the control module receives the start work instruction and sends a trigger signal to the excitation signal generator.

[0071] Among them, the control module is a ZYNQ platform. On the processor system (PS) side, the ARM processor runs an operating system, is responsible for executing tasks such as echo data reading, preprocessing, and target display, and realizes precise control of the data processing IP core at the programmable logic (PL) side; while on the PL side, it focuses on the signal transmission and reception drive control of the radar system, and at the same time designs a hardware IP core of the UWB-Mamba network using the Verilog hardware description language to improve the processing efficiency and real-time performance of the system.

[0072] The basic hardware components of the ultra-wideband impulse radar based on ZYNQ mainly include a pulse source, a transmitting antenna array, a broadband receiver, a signal acquisition board, a ZYNQ signal processing carrier board, a DDR4 cache module, and a display and control module. The effective configuration of each module enables the system of the ultra-wideband impulse radar to operate stably and accurately continuously.

[0073] Step S3, the excitation signal generator receives the trigger signal, generates a pulse width modulation signal and sends it to the avalanche transistor Marx circuit.

[0074] Step S4, the avalanche transistor Marx circuit receives the pulse width modulation signal, generates a zero-order Gaussian pulse signal and sends it to the transmitting antenna array.

[0075] Among them, the repetition frequency of the pulse width modulation signal is 50 kHz, the amplitude is 6 V, and the duty cycle is 5%. In this embodiment, a high-power avalanche transistor Marx circuit is selected as the pulse source of the ultra-wideband impulse radar. Combined with the coherent accumulation algorithm, the signal-to-noise ratio of the echo signal of the ultra-wideband impulse radar is effectively improved, and the detection range is increased from about 10 meters of the existing ultra-wideband impulse radar to about 2400 meters, effectively enhancing the ability of the ultra-wideband impulse radar system to identify bird targets with small size and complex reflection characteristics.

[0076] Step S5, the transmitting antenna array receives the zero-order Gaussian pulse signal, generates a detection signal and transmits it.

[0077] Among them, the zero-order Gaussian pulse signal is an extremely short high-power zero-order Gaussian pulse signal with the same repetition frequency as the PWM excitation signal, ensuring the unity and stability of the signal frequency; the detection signal is a first-order Gaussian pulse signal with a pulse width of 2 ns generated after passing through the Vivaldi transmitting antenna array.

[0078] Step S6, the receiving module receives and processes the echo signal, and sends the processed echo signal to the ADC module.

[0079] Among them, the receiving module includes a receiving antenna, a low-noise amplifier unit and a sampling unit; the receiving module receives and processes the echo signal, including: the receiving antenna receives the echo signal and sends it to the low-noise amplifier unit; the low-noise amplifier unit receives and processes the echo signal to obtain a first echo signal; the sampling unit receives and samples the first echo signal to obtain the processed echo signal.

[0080] After the receiving antenna receives the echo signal, it is amplified by the low-noise amplifier unit and then passes through a band-pass filter. The sampling unit controls the sampling time point of the signal, opens the receiving channel (corresponding to the range bin of 0.3 - 2400 m) within the time period of 2 ns - 16 μs after the signal is transmitted, and uses the rising edge trigger method to collect, quantize and encode and frame the echo signal with the ADC module.

[0081] Step S7, the ADC module converts the processed echo signal, generates a digital signal and sends it to the control module.

[0082] Step S8, the control module receives and processes the digital signal, generates a bird recognition result and sends it to the display and control module.

[0083] The control module includes an input data cache module, a data preprocessing module, a target recognition and classification module, and an output data cache module; the input data cache module caches the received digital signals and sends them to the data preprocessing module; the data preprocessing module processes the digital signals to obtain target data and sends it to the target recognition and classification module; the target recognition and classification module recognizes the target data, obtains the bird recognition result and sends it to the output data cache module; the output data cache module caches the bird recognition result and sends it to the display and control module.

[0084] Among them, the digital signals are input into the control module through the JESD204B protocol interface, the data is cached in the DDR4 cache module at the PL end of the control module, after preprocessing the data, the target recognition and classification module is used to identify the bird target of the signal; the processed data is transmitted to the PS end through the AXI4 high-speed internal interface for operations such as track formation, and finally the detected bird species type and movement track are displayed on the display and control module.

[0085] Step S9, the display and control module receives the bird recognition result and displays it.

[0086] In the above-mentioned embodiment, a bird detection and recognition method based on an ultra-wideband impulse radar is provided. The above method can effectively improve the signal-to-noise ratio of the echo signal of the ultra-wideband impulse radar by adding a high-power avalanche transistor Marx circuit as the pulse source of the ultra-wideband impulse radar, strengthen the feature extraction ability and real-time signal processing ability of the ultra-wideband impulse radar, accelerate the operation of the built-in model, realize efficient acquisition, processing and target recognition of the ultra-wideband radar echo signal, improve the overall performance and application value of the ultra-wideband impulse radar in the application scenarios of bird migration tracking and recognition, and enhance the ability to recognize bird targets with small volume and complex reflection characteristics.

[0087] Please refer to Figure 2 , in some embodiments, the data preprocessing module includes a coherent accumulation unit, an accumulation average background cancellation unit, a detection algorithm unit, a Fourier transform unit, and an image slicing unit.

[0088] The data preprocessing module processes the digital signals to obtain target data, including:

[0089] Step S821, the coherent accumulation unit receives the detection signal, processes the detection signal according to the FPGA parallel accumulator, obtains the first result signal and sends it to the accumulation average background cancellation unit.

[0090] Among them, the coherent accumulation unit uses the accumulator inside the FPGA to construct a parallel accumulator through the Verilog programming language for calculation, and can quickly perform coherent accumulation on the first result signal to improve the signal-to-noise ratio of the signal.

[0091] Step S822: The cumulative average background cancellation unit receives the first result signal, filters out the static clutter of the first result signal, obtains the second result signal and sends it to the detection algorithm unit.

[0092] Among them, relying on the parallel processing ability of the FPGA, the cumulative average background cancellation unit performs a shift subtraction operation on the first result signal in real time to filter out static clutter, providing a solid foundation for subsequent data processing.

[0093] Step S823: The detection algorithm unit receives the second result signal, detects the second result signal according to the OS CFAR algorithm, obtains the first target information, and sends it to the Fourier transform unit. Among them, the detection algorithm unit is the Figure 2 OSCFAR module in

[0094] Step S824: The Fourier transform unit receives the first target information, performs Doppler processing on the first target information based on the fast Fourier transform, obtains the second target information and sends it to the image slicing unit. Among them, the Fourier transform unit is the Figure 2 slow-time dimension FFT module in

[0095] Among them, the OS CFAR algorithm can detect the approximate position corresponding to the target. Then, using the fast Fourier transform (FFT) IP core on the FPGA, Doppler processing is performed on the data to extract velocity information to generate the second target information.

[0096] Step S825: The image slicing unit receives the second target information, performs a data slicing operation on the second target information, obtains the target data and sends it to the target recognition and classification module. Among them, the image slicing unit is the Figure 2 time-Doppler map slicing module in

[0097] Please refer to Figure 3 , in some embodiments, the target recognition and classification module includes a patch embedding unit, a flattening and linear mapping unit, an encoder unit, a channel fusion unit, and a multi-layer perception function unit.

[0098] Among them, the target recognition and classification module is a lightweight bird recognition and classification network based on the echo time Doppler (TD) map, which is specially designed for problems such as large amount of IR-UWB radar echo data, complex flying bird signal reflection model, high real-time requirement for detection and recognition, but limited computing resources of the ZYNQ platform. This module uses a high-low frequency decomposition method to identify signal features, fully combines the local detection ability of the convolutional neural network and the efficient long-distance dependence modeling ability of the state space model, and can effectively extract potential bird activity feature data from long-term radar target data and classify and recognize the target.

[0099] The weight parameters of each layer of the target recognition and classification module are first trained on the CPU platform. The training dataset is collected by this radar and divided into multiple categories according to distance and different bird species. Since the CPU platform uses floating-point numbers for training, when transplanted to the control module, the weights of the target recognition and classification module need to be fixed-point processed. Considering the recognition accuracy of the target recognition and classification module and the hardware resource occupancy of the control module, 11-bit fixed-point processing is used in this application. After the weight file is saved as a header file in the order of the target recognition and classification module, the weight file is loaded into the target recognition and classification module on the PL side through the ARM. The integrated training greatly improves the stability of the ultra-wideband impulse radar system and has a more accurate effect in recognizing small-volume flying birds.

[0100] The target recognition and classification module recognizes the target data and obtains the bird recognition result, including:

[0101] Step S831, the patch embedding unit arranges and splices the received target data in a preset order and a preset direction to obtain multi-dimensional data, and sends it to the flattening and linear mapping unit. The patch embedding unit can make full use of the target data and enhance the data edge features.

[0102] Step S832, the flattening and linear mapping unit converts the received multi-dimensional data into a one-dimensional vector, and performs a linear transformation on the one-dimensional vector using a fully connected layer to obtain a linear vector and sends it to the encoder unit. The flattening and linear mapping unit helps to capture the global features of the multi-dimensional data and prepares for the subsequent non-linear processing.

[0103] Step S833, the encoder unit analyzes the linear vector to obtain a high-frequency component and a low-frequency component; processes the high-frequency component to obtain a first feature, processes the low-frequency component to obtain a second feature; and sends the first feature and the second feature to the channel fusion unit. Among them, the encoder unit includes a global pooling layer; the global pooling layer extracts the low-frequency component in the linear vector; subtracts the low-frequency component from the linear vector to obtain the high-frequency component. Among them, the encoder unit is the Figure 3 HLF-SSM encoder in

[0104] Specifically, the encoder unit further includes a signal high-frequency information processing unit and a signal low-frequency information processing unit. The signal high-frequency information processing unit consists of a 1×1 ordinary convolution, two 3×3 convolution kernels, and two spatial channel reconstruction convolutions (SC convolutions). Using SC convolutions can consider both spatial and channel information at the same time, thereby effectively capturing the details and context information of the linear vector and improving the network's detection ability for small targets such as flying birds. The signal low-frequency information processing unit extracts the global information of the linear vector through a linear layer, cross convolution, and state space model (SSM), and can effectively improve the radar's long-time detection ability and multi-target recognition ability.

[0105] Step S834, the channel fusion unit receives and combines the first feature and the second feature to obtain a fusion signal and sends it to the multi-layer perception function unit; the channel fusion unit can enhance the richness and diversity of features.

[0106] Step S835, the multi-layer perception function unit receives and processes the fusion signal using a neural network to obtain a bird recognition result. The multi-layer perception function unit can learn and represent the non-linear relationship of data, convert the output of the neural network into a probability distribution, and thus determine the type of bird target carried in the signal. Among them, the multi-layer perception function unit is the Figure 3 multi-layer perceptron and Softmax module in

[0107] Please refer to Figure 4 , another embodiment of the present application provides a bird detection and recognition device based on an ultra-wideband impulse radar, and the device includes:

[0108] The display and control module is used to send a start working instruction to the control module; receive and display the bird recognition result.

[0109] The control module is used to receive the start working instruction, send a trigger signal to the excitation signal generator; receive and process the digital signal, generate a bird recognition result and send it to the display and control module.

[0110] The transmitting module includes an excitation signal generator, an avalanche transistor Marx circuit, and a transmitting antenna array; the excitation signal generator is used to receive the trigger signal, generate a pulse width modulation signal and send it to the avalanche transistor Marx circuit; the avalanche transistor Marx circuit is used to receive the pulse width modulation signal, generate a zero-order Gaussian pulse signal and send it to the transmitting antenna array; the transmitting antenna array is used to receive the zero-order Gaussian pulse signal, generate a detection signal and transmit it.

[0111] The receiving module is used to receive and process the echo signal, and send the processed echo signal to the ADC module.

[0112] The ADC module is used to convert the processed echo signal to generate a digital signal and send it to the control module.

[0113] In some embodiments, the control module includes an input data cache module, a data preprocessing module, a target recognition and classification module, and an output data cache module;

[0114] The input data cache module is used to cache the received digital signal and send it to the data preprocessing module.

[0115] The data preprocessing module is used to process the digital signal to obtain target data and send it to the target recognition and classification module.

[0116] The target recognition and classification module is used to recognize target data, obtain the bird recognition result and send it to the output data cache module; among them, the target recognition and classification module is the Figure 2 UWB-Mamba IP in

[0117] The output data cache module is used to cache the bird recognition result and send it to the display and control module.

[0118] In some embodiments, the data preprocessing module includes a coherent accumulation unit, an accumulation average background cancellation unit, a detection algorithm unit, a Fourier transform unit, and an image slicing unit.

[0119] The coherent accumulation unit is used to receive the detection signal, process the detection signal according to the FPGA parallel accumulator, obtain the first result signal and send it to the accumulation average background cancellation unit.

[0120] The accumulation average background cancellation unit is used to receive the first result signal, filter out the static clutter of the first result signal, obtain the second result signal and send it to the detection algorithm unit.

[0121] The detection algorithm unit is used to receive the second result signal, detect the second result signal according to the OS CFAR algorithm, obtain the first target information, and send it to the Fourier transform unit.

[0122] The Fourier transform unit is used to receive the first target information, perform Doppler processing on the first target information based on the fast Fourier transform, obtain the second target information and send it to the image slicing unit.

[0123] The image slicing unit is used to receive the second target information, perform data slicing on the second target information, obtain the target data and send it to the target recognition and classification module.

[0124] For the specific limitations of the bird detection and recognition device based on ultra-wideband impulse radar provided in this embodiment, reference can be made to the embodiments of the bird detection and recognition method based on ultra-wideband impulse radar in the above text, which will not be elaborated here. Each module in the above-mentioned bird detection and recognition device based on ultra-wideband impulse radar can be implemented in whole or in part by software, hardware, and their combination. The above-mentioned modules can be embedded in the processor in the computer device in hardware form or be independent of it, or be stored in the memory in the computer device in software form, so that the processor can call and execute the operations corresponding to the above-mentioned modules.

[0125] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, 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, it should be considered that the scope recorded in this specification.

[0126] The above-described embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several variations and improvements can still be made, and these all fall within the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A bird detection and recognition method based on ultra-wideband impulse radar, characterized in that The method comprises: The display and control module sends a start-work instruction to the control module; The control module receives the start-work instruction and sends a trigger signal to the excitation signal generator; The excitation signal generator receives the trigger signal, generates a pulse width modulation signal and sends it to the avalanche transistor Marx circuit; The avalanche transistor Marx circuit receives the pulse width modulation signal, generates a zero-order Gaussian pulse signal and sends it to the transmitting antenna array; The transmitting antenna array receives the zero-order Gaussian pulse signal, generates a detection signal and transmits it; The receiving module receives and processes the echo signal, and sends the processed echo signal to the ADC module; The ADC module converts the processed echo signal to generate a digital signal and sends it to the control module; The control module receives and processes the digital signal, generates a bird identification result and sends it to the display and control module; The display and control module receives and displays the bird identification result.

2. The method for detecting and identifying birds based on ultra-wideband impulse radar according to claim 1, characterized in that The control module includes an input data cache module, a data preprocessing module, a target recognition and classification module, and an output data cache module; The control module receives and processes the digital signal to generate a bird identification result, including: The input data buffer module buffers the received digital signal and sends it to the data preprocessing module; The data preprocessing module processes the digital signal to obtain target data and sends the target data to the target recognition and classification module; The target recognition and classification module recognizes the target data, obtains the bird recognition result and sends it to the output data buffer module; The output data buffer module buffers the bird recognition result and sends it to the display and control module.

3. The method for detecting and identifying birds based on ultra-wideband impulse radar according to claim 2, characterized in that The data preprocessing module includes a coherent accumulation unit, an accumulation average background cancellation unit, a detection algorithm unit, a Fourier transform unit and an image slicing unit; The data preprocessing module processes the digital signal to obtain target data, including: The coherent accumulation unit receives the detection signal, processes the detection signal according to the FPGA parallel accumulator, obtains a first result signal and sends it to the accumulation average background cancellation unit; The accumulation average background cancellation unit receives the first result signal, filters out static clutter in the first result signal, obtains a second result signal and sends it to the detection algorithm unit; The detection algorithm unit receives the second result signal, detects the second result signal according to the OS CFAR algorithm, obtains first target information, and sends it to the Fourier transform unit; The Fourier transform unit receives the first target information, performs Doppler processing on the first target information based on fast Fourier transform, obtains second target information and sends the second target information to the image slicing unit; The image slicing unit receives the second target information, performs a data slicing operation on the second target information, obtains target data, and sends the obtained target data to the target recognition and classification module.

4. The method for detecting and identifying birds based on an ultra-wideband impulse radar according to claim 2, wherein The target recognition and classification module includes a patch embedding unit, a flattening and linear mapping unit, an encoder unit, a channel fusion unit and a multi-layer perception function unit; The target recognition and classification module recognizes the target data and obtains a bird recognition result, including: The patch embedding unit arranges and splices the received target data in a preset order and a preset direction to obtain multi-dimensional data, and sends it to the flattening and linear mapping unit; The flattening and linear mapping unit converts the received multi-dimensional data into a one-dimensional vector, and performs a linear transformation on the one-dimensional vector using a fully connected layer to obtain a linear vector and sends it to the encoder unit; The encoder unit analyzes the linear vector to obtain a high-frequency component and a low-frequency component; processes the high-frequency component to obtain a first feature, processes the low-frequency component to obtain a second feature; and sends the first feature and the second feature to the channel fusion unit; The channel fusion unit receives and combines the first feature and the second feature to obtain a fusion signal and sends it to the multi-layer perceptron function unit; The multi-layer perceptron function unit receives and processes the fusion signal using a neural network to obtain a bird recognition result.

5. The method for detecting and identifying birds based on an ultra-wideband impulse radar according to claim 4, wherein The encoder unit includes a global pooling layer; The encoder unit analyzes the linear vector to obtain a high-frequency component and a low-frequency component, including: The global pooling layer extracts the low-frequency component from the linear vector; Subtract the low-frequency component from the linear vector to obtain the high-frequency component.

6. The method for detecting and identifying birds based on an ultra-wideband impulse radar according to claim 1, characterized in that, The receiving module includes a receiving antenna, a low-noise amplifier unit, and a sampling unit; the receiving module receives and processes the echo signal, including: The receiving antenna receives the echo signal and sends it to the low-noise amplifier unit; The low-noise amplifier unit receives and processes the echo signal to obtain a first echo signal; The sampling unit receives and samples the first echo signal to obtain a processed echo signal.

7. The method for detecting and identifying birds based on ultra-wideband impulse radar according to claim 1, characterized in that, The repetition frequency of the pulse width modulation signal is 50 kHz, the amplitude is 6 V, and the duty cycle is 5%.

8. A bird detection and identification device based on an ultra-wideband impulse radar, characterized in that, The device includes: A display and control module, configured to send a start working instruction to the control module; receive and display the bird recognition result; A control module, configured to receive the start working instruction, send a trigger signal to the excitation signal generator; receive and process the digital signal, generate a bird recognition result and send it to the display and control module; A transmitting module, including an excitation signal generator, an avalanche transistor Marx circuit, and a transmitting antenna array; the excitation signal generator is configured to receive the trigger signal, generate a pulse width modulation signal and send it to the avalanche transistor Marx circuit; the avalanche transistor Marx circuit is configured to receive the pulse width modulation signal, generate a zero-order Gaussian pulse signal and send it to the transmitting antenna array; the transmitting antenna array is configured to receive the zero-order Gaussian pulse signal, generate a detection signal and transmit it; A receiving module, configured to receive and process the echo signal, and send the processed echo signal to the ADC module; The ADC module is configured to convert the processed echo signal to generate a digital signal and send it to the control module.

9. The bird detection and identification device based on ultra-wideband impulse radar according to claim 8, characterized in that, The control module includes an input data cache module, a data preprocessing module, a target recognition and classification module, and an output data cache module; The input data cache module is configured to cache the received digital signal and send it to the data preprocessing module; The data preprocessing module is configured to process the digital signal to obtain target data and send it to the target recognition and classification module; The target recognition and classification module is used to recognize the target data, obtain the bird recognition result and send it to the output data cache module; The output data cache module is used to cache the bird recognition result and send it to the display and control module.

10. The bird detection and identification device based on ultra-wideband impulse radar according to claim 9, characterized in that, The data preprocessing module includes a coherent accumulation unit, an accumulation average background cancellation unit, a detection algorithm unit, a Fourier transform unit and an image slicing unit; The coherent accumulation unit is used to receive the detection signal, process the detection signal according to the FPGA parallel accumulator, obtain the first result signal and send it to the accumulation average background cancellation unit; The accumulation average background cancellation unit is used to receive the first result signal, filter out the static clutter of the first result signal, obtain the second result signal and send it to the detection algorithm unit; The detection algorithm unit is used to receive the second result signal, detect the second result signal according to the OS CFAR algorithm, obtain the first target information, and send it to the Fourier transform unit; The Fourier transform unit is used to receive the first target information, perform Doppler processing on the first target information based on the fast Fourier transform, obtain the second target information and send it to the image slicing unit; The image slicing unit is used to receive the second target information, perform data slicing operation on the second target information, obtain the target data and send it to the target recognition and classification module.

Images