Low-altitude slow small target detection method and system

Abstract

The invention provides a low-altitude slow and small target detection method and system, relates to the technical field of low-altitude target detection, and solves the technical problem that all-day all-weather high-precision low-false-alarm detection of a low-altitude slow and small target is difficult to realize in the prior art. The method specifically comprises the following steps: acquiring a detection data set; preprocessing the detection data set; performing target detection on the preprocessed detection data set to obtain a target detection result; the target detection result comprises a first detection result, a second detection result and a third detection result; performing multi-sensor data fusion on the target detection result, and continuously tracking the target through a target tracking algorithm to obtain a tracking result; based on the tracking result, classifying and identifying the tracking result through an image classification algorithm to obtain a classification and identification result; and based on the target detection result, the tracking result and the classification identification result, setting a corresponding early warning threshold for early warning, and taking emergency treatment measures. The method is used for low-altitude slow small target detection.

Description

Technical Field

[0001] This application relates to the field of low-altitude target detection technology, and in particular to a method and system for detecting low-altitude slow and small targets. Background Technology

[0002] Low-altitude, slow-moving, and small target detection is of great significance for the security and protection of critical areas such as airports and nuclear power plants. In existing technologies, detection methods mainly include radar detection, optical detection, acoustic detection, and combined detection methods that combine millimeter-wave radar with photoelectric sensors. Among them, radar determines the target's position and velocity based on the principle of electromagnetic wave reflection, optical equipment relies on the target's visible light or infrared thermal radiation characteristics for imaging and identification, acoustic sensors assist in judgment by capturing the sound wave signals generated by the target, and the combined scheme uses millimeter-wave radar to conduct a large-area scan to initially detect the target, and then uses photoelectric equipment for precise tracking and identification.

[0003] However, traditional radars have small radar cross-sections and weak echo signals for low-altitude, slow-moving, and small targets, making them susceptible to ground clutter and background noise interference, resulting in high false alarm rates and missed alarms. Optical detection is limited by weather and lighting conditions, with performance significantly degrading at night or in adverse weather. Acoustic detection has slow propagation speed and rapid attenuation, with limited detection range and susceptibility to environmental noise interference. Even the combination of millimeter-wave radar and photoelectric sensors lacks deep fusion of data from multiple sensors and advanced signal processing and target recognition algorithms, making it difficult to balance detection accuracy, stability, and all-weather applicability. Therefore, existing technologies face the technical challenge of achieving high-precision, low-false-alarm detection of low-altitude, slow-moving, and small targets at all times and in all weather conditions. Summary of the Invention

[0004] This application provides a method and system for detecting low-altitude, slow-moving, and small targets, which solves the technical problem that existing technologies struggle to achieve high-precision, low-false-alarm detection of low-altitude, slow-moving, and small targets around the clock and in all weather conditions.

[0005] To achieve the above objectives, this application adopts the following technical solution: Firstly, a method for detecting low-altitude, slow-moving, small targets is provided, comprising: acquiring a detection dataset; the detection dataset includes first data, second data, and third data; the first data is acquired by millimeter-wave radar; the second data is acquired by an optoelectronic sensor; the third data is acquired by an acoustic sensor; the optoelectronic sensor includes an infrared thermal imager and a visible light camera; preprocessing the detection dataset; performing target detection on the preprocessed detection dataset to obtain target detection results; the target detection results include the first detection result, the second detection result, and the third detection result; fusing the first detection result, the second detection result, and the third detection result into multi-sensor data, and continuously tracking the target using a target tracking algorithm to obtain tracking results; classifying and recognizing the tracking results using an image classification algorithm based on the tracking results to obtain classification and recognition results; and setting corresponding warning thresholds based on the target detection results, tracking results, and classification and recognition results to issue warnings and take emergency response measures.

[0006] In conjunction with the first aspect mentioned above, in one possible implementation, the detection dataset is preprocessed, including: amplifying the first data, filtering out interference signals using a bandpass filter, and extracting effective radar signals including target distance and velocity information; grayscale processing the second data, removing image noise using median filtering or Gaussian filtering algorithms, and enhancing the target edge contour using the Sobel operator or Canny operator; filtering the third data using an adaptive bandstop filter with an adjustable center frequency, separating mixed sound sources through independent component analysis, extracting the target acoustic wave characteristic frequency bands, and performing normalization processing.

[0007] In conjunction with the first aspect mentioned above, in one possible implementation, target detection is performed on the preprocessed detection dataset to obtain target detection results, including: using the CFAR algorithm to perform threshold decision on the preprocessed first data, and then obtaining the first detection result through Doppler frequency shift analysis; using a deep learning-based target detection algorithm to perform target detection and classification on the preprocessed second data to obtain the second detection result; and using energy detection and feature frequency analysis on the preprocessed third data to determine the target acoustic wave features and obtain the third detection result.

[0008] In conjunction with the first aspect mentioned above, in one possible implementation, the first detection result includes the target's distance information, velocity information, and location information of the suspected target area; the second detection result includes the target's coordinate position in the image and target category information; and the third detection result includes the determination result of the presence of the target's acoustic features, the feature frequency matching degree, and the signal occurrence time and location information.

[0009] In conjunction with the first aspect mentioned above, in one possible implementation, the first, second, and third detection results are fused using multi-sensor data. A target tracking algorithm is then used to continuously track the target, yielding the tracking result. This includes: setting a confidence threshold; determining whether the confidence level of the first detection result exceeds the threshold; if so, setting a sensor data synchronization window to synchronize the three types of detection results; using a Kalman filter algorithm to jointly estimate the target state of the synchronized detection results, obtaining the target's optimal position and velocity information; employing a joint probabilistic data association filtering algorithm and a multi-model adaptive tracking algorithm to continuously track the target based on the optimal position and velocity information, and outputting the tracking result. The tracking result includes the target's real-time position and trajectory.

[0010] In conjunction with the first aspect mentioned above, in one possible implementation, the joint probability data association filtering algorithm is used to calculate the association probability between the target and the measurement data of each sensor; the multi-model adaptive tracking algorithm establishes multiple dynamic models for different motion modes of the target and adaptively switches models according to the actual motion of the target.

[0011] In conjunction with the first aspect mentioned above, in one possible implementation, based on the tracking results, the tracking results are classified and identified using an image classification algorithm to obtain the classification and identification results. This includes: extracting a high-resolution image of the target acquired by a photoelectric sensor from the tracking results; inputting the high-resolution image of the target into a pre-trained deep learning image classification algorithm; the deep learning image classification algorithm is either the ResNet algorithm or the DenseNet algorithm; obtaining the target category probability output by the deep learning image classification algorithm; determining the specific type of the target based on the target category probability; and obtaining the classification and identification results.

[0012] In conjunction with the first aspect mentioned above, in one possible implementation, based on the target detection results, tracking results, and classification and identification results, a corresponding early warning threshold is set for early warning, and emergency response measures are taken, including: calculating the target threat index and setting a target threat index threshold; when the target threat index is greater than the target threat index threshold and the target enters the GIS electronic fence area, an early warning is triggered; the early warning includes audible and visual alarms, screen pop-ups, SMS messages, and application push notifications, and simultaneously transmits the target location, speed, type, and trajectory information; after the early warning is triggered, the interception device is activated, and a dispatch information is issued; the dispatch information is used to dispatch personnel to the target location for handling.

[0013] In conjunction with the first aspect mentioned above, one possible implementation is that the deep learning-based object detection algorithm is either the YOLO series algorithm or the Faster R-CNN algorithm.

[0014] Secondly, a low-altitude, slow-moving, small target detection system is provided, comprising: a millimeter-wave radar, an optoelectronic sensor, an acoustic sensor, and electronic equipment; the millimeter-wave radar is used to collect first data; the optoelectronic sensor, including an infrared thermal imager and a visible light camera, is used to collect second data; the acoustic sensor is used to collect third data; the electronic equipment is used to acquire a detection dataset; the detection dataset includes the first data, the second data, and the third data; the detection dataset is preprocessed; target detection is performed on the preprocessed detection dataset to obtain target detection results; the target detection results include the first detection result, the second detection result, and the third detection result; the first detection result, the second detection result, and the third detection result are fused using multi-sensor data, and the target is continuously tracked using a target tracking algorithm to obtain tracking results; based on the tracking results, the tracking results are classified and identified using an image classification algorithm to obtain classification and identification results; based on the target detection results, the tracking results, and the classification and identification results, corresponding warning thresholds are set for warning and emergency response measures are taken.

[0015] Thirdly, a low-altitude, slow-moving, small target detection device is provided, comprising: a communication unit and a processing unit; the communication unit is used to acquire a detection dataset; the detection dataset includes first data, second data, and third data; the first data is acquired by a millimeter-wave radar; the second data is acquired by a photoelectric sensor; the third data is acquired by an acoustic sensor; the photoelectric sensor includes an infrared thermal imager and a visible light camera; the processing unit is used to preprocess the detection dataset; to perform target detection on the preprocessed detection dataset to obtain target detection results; the target detection results include a first detection result, a second detection result, and a third detection result; to perform multi-sensor data fusion on the first detection result, the second detection result, and the third detection result, and to continuously track the target using a target tracking algorithm to obtain tracking results; based on the tracking results, to classify and identify the tracking results using an image classification algorithm to obtain classification and identification results; based on the target detection results, tracking results, and classification and identification results, to set corresponding warning thresholds for warning and to take emergency response measures.

[0016] Fourthly, this application provides a low-altitude, slow-moving, small target detection device, comprising: a processor and a storage medium; the storage medium includes instructions, and the processor is configured to execute the instructions to implement the method described in the first aspect and any possible implementation thereof. This low-altitude, slow-moving, small target detection device may be an electronic device or a chip within an electronic device.

[0017] Fifthly, this application provides a computer-readable storage medium storing instructions that, when executed on a low-altitude slow-moving target detection device, cause the low-altitude slow-moving target detection device to perform the methods described in the first aspect and any possible implementation thereof.

[0018] In a sixth aspect, this application provides a computer program product containing instructions that, when the computer program product is run on a low-altitude slow-small target detection device, cause the low-altitude slow-small target detection device to perform the methods described in the first aspect and any possible implementation thereof.

[0019] This application provides a method and system for detecting low-altitude, slow-moving, and small targets. It can achieve stable all-weather, all-day detection of low-altitude, slow-moving, and small targets by deeply fusing multi-source data from millimeter-wave radar, photoelectric sensors, and acoustic sensors, combined with precise target detection, tracking, and classification algorithms. This significantly reduces false alarm and missed alarm rates, improves target identification accuracy and tracking reliability, and provides rapid response through intelligent early warning and decision-making mechanisms, effectively meeting protection requirements.

[0020] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[0021] Figure 1 A system architecture diagram of a low-altitude slow-moving small target detection system provided in this application embodiment; Figure 2 A flowchart illustrating a low-altitude, slow-moving, small target detection method provided in this application embodiment; Figure 3 A flowchart illustrating another low-altitude, slow-moving, small target detection method provided in this application embodiment; Figure 4 A flowchart illustrating another low-altitude, slow-moving, small target detection method provided in this application embodiment; Figure 5 A flowchart illustrating another low-altitude, slow-moving, small target detection method provided in this application embodiment; Figure 6 A flowchart illustrating another low-altitude, slow-moving, small target detection method provided in this application embodiment; Figure 7This is a schematic diagram of the structure of a low-altitude slow-moving small target detection device provided in an embodiment of this application; Figure 8 This is a schematic diagram of the hardware structure of a low-altitude slow-moving small target detection device provided in an embodiment of this application. Detailed Implementation

[0022] In the description of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B. The "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" means one or more, and "multiple" means two or more. The terms "first," "second," etc., do not limit the quantity or order of execution, and "first," "second," etc., do not necessarily imply differences.

[0023] It should be noted that, in this application, the terms "exemplary" or "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0024] The low-altitude slow-moving small target detection method provided in this application embodiment can be applied to, for example... Figure 1 The low-altitude slow-moving small target detection system shown includes: millimeter-wave radar 101, photoelectric sensor 102, acoustic sensor 103, and electronic equipment 104.

[0025] The system includes a millimeter-wave radar 101 for scanning low-altitude areas, emitting millimeter waves and receiving reflected echoes from targets to collect first data; a photoelectric sensor 102 for capturing target thermal radiation and visible light images, performing real-time imaging, and collecting second data; an acoustic sensor 103 for collecting sound wave signals generated by the target, capturing characteristic frequency bands, and collecting third data; and an electronic device 104 for acquiring a detection dataset. The detection dataset includes the first, second, and third data. The first data is acquired by the millimeter-wave radar; the second data is acquired by the photoelectric sensor; and the third data is acquired by the acoustic sensor. The photoelectric sensor includes an infrared thermal imager and a visible light camera. The detection dataset is preprocessed. Target detection is performed on the preprocessed detection dataset to obtain target detection results. The target detection results include the first, second, and third detection results. The first, second, and third detection results are fused using multi-sensor data, and the target is continuously tracked using a target tracking algorithm to obtain tracking results. Based on the tracking results, an image classification algorithm is used to classify and identify the tracking results to obtain classification and identification results. Based on the target detection results, tracking results, and classification and identification results, corresponding warning thresholds are set for early warning, and emergency response measures are taken.

[0026] To address the technical challenge of achieving high-precision, low-false-alarm detection of low-altitude, slow-moving, and small targets around the clock and in all weather conditions in existing technologies, this application provides a method for detecting such targets. The method includes: acquiring detection data using millimeter-wave radar, photoelectric sensors, and acoustic sensors; preprocessing and detecting targets separately; fusing multi-sensor data and continuously tracking the target; identifying the target type based on the tracking results using an image classification algorithm; and finally setting an early warning threshold and implementing emergency response measures based on the detection, tracking, and classification results. This method enables stable, all-weather, all-time detection of low-altitude, slow-moving, and small targets, significantly reducing false alarm and missed alarm rates, improving target identification accuracy and tracking reliability, and providing rapid response through an intelligent early warning and decision-making mechanism, effectively meeting security protection requirements.

[0027] like Figure 2 As shown in the embodiments of this application, the low-altitude slow-moving small target detection method includes: S201. Obtain the probe dataset.

[0028] The detection dataset refers to a collection of multi-source sensor data used for target detection, which includes first data, second data, and third data.

[0029] In this embodiment, millimeter-wave radar, photoelectric sensors, and acoustic sensors respectively collect corresponding data: the millimeter-wave radar performs a wide-range scan search at low altitudes, capturing target motion characteristics using high range and angular resolution. It can select the 24GHz or 77GHz frequency band, covering a detection range of several hundred meters to several kilometers at a scan rate of tens of times per second, reducing ground clutter interference to collect the first data; the photoelectric sensors include an infrared thermal imager and a visible light camera. The infrared thermal imager utilizes the difference in thermal radiation between the target and the background, adaptable to cooled or uncooled detectors, with resolutions ranging from 320×240 to 6... The camera uses a 40×480 pixel resolution and a suitable lens focal length to capture thermal radiation images at night or in low-light environments. The visible light camera, with a resolution of over 10 million pixels and a frame rate of 25-30fps, combined with adjustable lens parameters and autofocus and exposure functions, captures high-resolution images during the day or in good lighting conditions, together forming the second set of data. The acoustic sensor uses a high-sensitivity microphone array, arranged in a linear array (0.5m spacing) or a circular array (1.2m diameter), and uses the Capon beamforming algorithm to enhance signal directionality, capturing the sound wave signals generated by the target to collect the third set of data.

[0030] As an example, in an airport deployment scenario, a millimeter-wave radar scans the low-altitude range, an optoelectronic sensor provides real-time imaging, and an acoustic sensor captures sound signals. The millimeter-wave radar is an Aptiv FLR7, which can operate in the 77GHz band with a scan rate of 30 scans per second; the infrared thermal imager is a FLIR Tau 2 uncooled microbolometer; the visible light camera is a DS-2CD6A26FWD-IZH with 12 megapixels and a frame rate of 30fps; and the acoustic sensor is a Hella acoustic sensor array with an 8-microphone circular array and a sampling rate of 48kHz.

[0031] Based on the above steps, comprehensive multi-dimensional data on the target can be collected, providing rich information support for subsequent exploration.

[0032] S202. Preprocess the probe dataset.

[0033] Preprocessing refers to the process of purifying and enhancing the raw collected data.

[0034] In this embodiment, the radar signal in the first data is amplified sequentially, and a bandpass filter is used to filter out high-frequency noise and low-frequency interference signals to extract effective radar signals including target distance and speed information; the second data is grayscaled, and image noise is removed by median filtering or Gaussian filtering algorithms, and the target edge contour is enhanced by Sobel or Canny operators; the third data is filtered by an adaptive bandstop filter with dynamically adjustable center frequency, and the mixed sound sources are separated by independent component analysis, the target sound wave characteristic frequency band is extracted and normalized.

[0035] It should be noted that the preprocessing operation needs to be matched to the characteristics of different sensor data and the interference needs to be removed in a targeted manner.

[0036] As an example, high-frequency noise is filtered out from radar signals collected from airports, median filtering is used to denoise image data, and environmental interference in the 50Hz-5kHz range is removed from acoustic signals.

[0037] Based on the above steps, data quality can be improved and the impact of invalid information on subsequent processing can be reduced.

[0038] S203. Perform target detection on the preprocessed detection dataset to obtain the target detection results.

[0039] The target detection result refers to the set of target-related information identified by each sensor through the detection algorithm, including the first detection result, the second detection result, and the third detection result.

[0040] In this embodiment, the first data is detected by a threshold decision combined with velocity analysis, the second data is detected and classified by a deep learning algorithm, and the target sound wave is judged based on energy and frequency characteristics of the third data.

[0041] As an example, in an airport scenario, target area determination is performed on radar preprocessing data, target coordinates are identified from photoelectric images, and feature frequency matching degree is analyzed from acoustic signals.

[0042] Based on the above steps, potential targets can be accurately identified, and effective target information can be initially screened.

[0043] S204. The first detection result, the second detection result and the third detection result are fused into multi-sensor data, and the target is continuously tracked by the target tracking algorithm to obtain the tracking result.

[0044] Multi-sensor data fusion refers to the collaborative processing of integrating detection results from different sensors.

[0045] In this embodiment, the three types of detection results are first synchronized, then the target state information is jointly estimated, and finally the target is continuously monitored through a tracking algorithm to output the real-time position and motion trajectory.

[0046] It should be noted that data synchronization requires controlling the time window to ensure the timeliness and consistency of the merged data.

[0047] Based on the above steps, the target can be tracked stably, improving the continuity and reliability of target monitoring.

[0048] S205. Based on the tracking results, the tracking results are classified and identified using an image classification algorithm to obtain the classification and identification results.

[0049] Among them, the classification and identification result refers to the determination result that clearly defines the specific type of the target, such as drones, birds, etc.

[0050] In this embodiment of the application, a high-resolution image of the target is extracted from the tracking results, input into a pre-trained deep learning classification algorithm, and the target type is determined based on the output probability.

[0051] It should be noted that the classification algorithm needs to be trained on data from multiple scenarios to ensure recognition accuracy in different environments.

[0052] As an example, in an airport scenario, images of tracked targets are extracted, and classification algorithms are used to distinguish between drones and birds.

[0053] Based on the above steps, the target type can be accurately distinguished, providing a basis for subsequent differentiated treatment.

[0054] S206. Based on the target detection results, tracking results, and classification and recognition results, set corresponding early warning thresholds to issue early warnings and take emergency response measures.

[0055] Among them, the early warning threshold refers to the critical condition that triggers an early warning, which is dynamically set based on multi-dimensional information about the target.

[0056] In this embodiment of the application, the target threat index is first calculated to determine whether the warning conditions are met, and then a warning signal is issued, and the interception device is activated or personnel are dispatched to handle the situation simultaneously.

[0057] Based on the above steps, we can respond quickly to potential threats and effectively reduce the risks posed by low-altitude, slow, and small targets.

[0058] Based on the above technical solutions, by deeply fusing multi-source data from millimeter-wave radar, photoelectric sensors, and acoustic sensors, and combining them with precise target detection, tracking, and classification algorithms, stable all-weather, all-day detection of low-altitude, slow-moving, and small targets can be achieved. This significantly reduces the false alarm rate and missed alarm rate, improves the accuracy of target identification and the reliability of tracking, and enables rapid response through intelligent early warning and decision-making mechanisms, effectively meeting protection requirements.

[0059] In one possible implementation of the embodiments of this application, combined with Figure 2 ,like Figure 3 As shown, the above S203 can be implemented through the following S301, S302 and S303, which are explained in detail below: S301. The CFAR algorithm is used to make a threshold decision on the preprocessed first data, and then the first detection result is obtained through Doppler frequency shift analysis.

[0060] Among them, the CFAR algorithm is a constant false alarm rate detection algorithm, which ensures detection stability in complex environments by adaptively adjusting the threshold; Doppler frequency shift is the signal frequency shift caused by target motion, which is positively correlated with radial velocity; the first detection result includes the target's distance information, velocity information, and the location information of the suspected target area.

[0061] In this embodiment, CFAR detection is first performed on the preprocessed radar signal: a preset number of reference cells are selected around each signal sampling point, the background noise power is estimated through the reference cells, and a decision threshold is calculated in combination with the set false alarm rate. If the signal amplitude of the current sampling point exceeds the threshold, it is marked as a suspected target area. Then, Doppler frequency shift analysis is performed on the multi-frame radar signal, and the signal is analyzed by Fast Fourier Transform (FFT) to calculate the radial velocity information of the target. The velocity information is combined with the position information of the suspected target area to filter out targets that meet the velocity characteristics of low-altitude slow and small targets. Finally, the first detection result containing target distance information, velocity information and suspected target area position information is output.

[0062] As an example, in low-altitude airport detection, 32 reference cells are selected to estimate noise, and a false alarm rate is set. FFT analysis was used to screen out suspected targets with speeds between 5 and 20 m / s.

[0063] Based on the above steps, ground clutter and background noise can be effectively suppressed, improving the radar's detection accuracy for low-altitude, slow-moving, and small targets, and reducing missed alarms and false alarms.

[0064] S302. A deep learning-based target detection algorithm is used to perform target detection and classification on the preprocessed second data to obtain the second detection result.

[0065] Among them, the deep learning target detection algorithm is a detection method based on learning target features from deep neural networks. It includes core steps such as feature extraction, candidate region generation and classification judgment. The second detection result includes the target's coordinate position in the image and the target category information.

[0066] In this embodiment, the YOLO series algorithm or Faster R-CNN algorithm is used as the deep learning object detection algorithm. The pre-processed infrared thermal imaging image sequence and visible light image sequence are input into the pre-trained model. The model extracts the appearance features such as shape, texture and color of the target through network structures such as convolutional layers and pooling layers, quickly scans the image to generate candidate target regions, performs feature matching and classification judgment on each candidate region, and outputs the coordinate position of the target in the image (represented in the form of bounding box) and target category information (such as drones, birds, etc.). The model is trained using a dataset containing 150,000 labeled images, covering 10 complex scenes such as rain, fog and night. The AdamW optimizer (learning rate 0.001, weight decay 0.01) is used to train the YOLOv7 model. CutMix data augmentation is used to improve the robustness of small target detection. During training, the network structure parameters, optimization algorithm and L1 / L2 regularization terms are adjusted to prevent model overfitting and improve generalization ability.

[0067] It should be noted that infrared and visible light images must be input into the model simultaneously to ensure the time consistency of the detection results.

[0068] Based on the above steps, it can adapt to different lighting and weather conditions, accurately output the location and category information of the target, and provide a foundation for subsequent classification and recognition.

[0069] S303. Perform energy detection and characteristic frequency analysis on the preprocessed third data to determine the target acoustic wave characteristics and obtain the third detection result. The third detection result includes the determination result of the existence of the target acoustic wave characteristics, the characteristic frequency matching degree, and the signal occurrence time and location information.

[0070] Among them, energy detection is a method to determine the existence of a target by calculating the intensity of the sound wave signal; characteristic frequency analysis is to extract specific frequency band signals related to the target for target feature matching.

[0071] In this embodiment, the energy of the preprocessed acoustic signal is first calculated, and a fixed energy threshold is set. If the signal energy exceeds the threshold, it is preliminarily determined that there is a potential target acoustic wave. Then, the characteristic frequency components of the signal are extracted and matched with a preset low-altitude slow-moving target acoustic wave spectrum library (which includes characteristic frequency ranges of different types of UAV motor sounds, bird flapping sounds, etc.). Finally, the existence determination result of the target acoustic wave characteristics, the characteristic frequency matching degree, and the time and location information of the signal appearance are recorded to form the third detection result.

[0072] As an example, acoustic data collected around the airport detected a signal with energy exceeding the threshold. After spectrum analysis, the signal matched the characteristic frequency of a certain type of consumer drone motor sound with a 92% accuracy. The recorded occurrence time was 14:23 and the location was on the east side of the airport area.

[0073] Based on the above steps, the existence of the target can be confirmed from the perspective of sound waves, enriching the information dimension of target detection and improving the reliability of multi-sensor fusion.

[0074] Based on the above technical solution, through targeted detection algorithms and multi-dimensional feature extraction, target detection is achieved from three dimensions: motion characteristics, image features, and sound wave features, respectively, outputting comprehensive and accurate detection results, providing high-quality data support for subsequent multi-sensor data fusion.

[0075] In one possible implementation of the embodiments of this application, combined with Figure 2 ,like Figure 4 As shown, the above S204 can be implemented through the following S401, S402 and S403, which are explained in detail below: S401. Set a confidence threshold and determine whether the confidence of the first detection result is greater than the confidence threshold. If it is, set a sensor data synchronization window to synchronize the three types of detection results.

[0076] Among them, the confidence threshold is the critical value for screening valid initial detection results, and the sensor data synchronization window is the time range for ensuring the temporal consistency of multi-source detection data.

[0077] In this embodiment of the application, the confidence threshold is set to 60%. The confidence level of the signal in the suspected target area in the first detection result is evaluated to determine whether it is greater than the threshold. If the condition is met, a sensor data synchronization window of no more than 50ms is set to synchronously collect the first detection result (target distance, speed, suspected area location), the second detection result (image coordinates, target category) and the third detection result (sound wave presence determination, feature frequency matching degree, time and location).

[0078] Based on the above steps, highly reliable initial target information can be selected, ensuring the time consistency and effectiveness of subsequent data fusion.

[0079] S402. The target state is jointly estimated by the synchronized detection results using the Kalman filter algorithm to obtain the target's optimal position and optimal velocity information.

[0080] Among them, the Kalman filter algorithm is an algorithm that achieves the optimal estimation of the target state through a prediction-update iterative process, while the joint estimation of the target state is a process of integrating multi-source sensor data to calculate the unified state parameters of the target.

[0081] In this embodiment, the distance and velocity information in the first detection result after synchronization, the coordinate position information in the second detection result, and the position information in the third detection result are used as input data for Kalman filtering. The current state is predicted based on the target's historical state, and then the predicted value is corrected using multi-source sensor measurement data. The optimal three-dimensional position (x, y, z) and radial velocity information of the target are obtained through iterative calculation, thus eliminating the measurement error of a single sensor.

[0082] It should be noted that the Kalman filter algorithm has strong anti-interference capabilities and can adapt to the nonlinear motion state of the target.

[0083] As an example, in low-altitude airport detection, the distance of 1.2km and speed of 15m / s measured by radar, the coordinates (320,240) measured by photoelectric sensors, and the position information measured by acoustic sensors are fused together, and the optimal position (1200m,350m,100m) and optimal speed of 14.8m / s are estimated by Kalman filtering.

[0084] Based on the above steps, multi-dimensional measurement data can be integrated, noise interference can be reduced, and accurate target core state information can be output.

[0085] S403. Employs a joint probability data correlation filtering algorithm and a multi-model adaptive tracking algorithm to continuously track the target based on optimal position and optimal velocity information, and outputs the tracking results.

[0086] Among them, the Joint Probabilistic Data Association Filter (JPDAF) algorithm is used to calculate the association probability between the target and the measurement data of each sensor, and the multi-model adaptive tracking algorithm establishes multiple dynamic models for different motion modes of the target and adaptively switches between them.

[0087] In this embodiment, the correlation probability between the target and synchronized radar, photoelectric, and acoustic measurement data is first calculated using the JPDAF algorithm to select the most reliable measurement data correlation relationship. At the same time, multiple dynamic models such as uniform motion and uniform acceleration motion are established. The target motion mode is judged in real time based on the optimal position and velocity information, and the matching model is adaptively switched. The target state is continuously iterated and updated, and the tracking result containing the target's real-time three-dimensional position and continuous motion trajectory is output.

[0088] As an example, in an airport scenario, when a target changes from uniform flight to uniform acceleration during a turn, JPDAF associates data from multiple sensors, and the multi-model adaptive tracking algorithm switches to the uniform acceleration model to continuously track the target and record its trajectory.

[0089] Based on the above steps, stable and continuous tracking of low-altitude, slow-moving, and small targets can be achieved, maintaining tracking continuity even in complex motion or interference scenarios.

[0090] Based on the above technical solution, by screening effective initial data, accurately jointly estimating the target state, and continuously tracking with collaborative algorithms, deep fusion of multi-sensor data is achieved, which significantly improves the accuracy, stability, and anti-interference capability of tracking low-altitude, slow, and small targets.

[0091] In one possible implementation of the embodiments of this application, combined with Figure 2 ,like Figure 5 As shown, the above S205 can be implemented through the following S501, S502 and S503, which are explained in detail below: S501. Extract the high-resolution image of the target acquired by the photoelectric sensor from the tracking results.

[0092] The tracking results include status information such as the target's real-time position and motion trajectory, while the high-resolution target image is an image captured by the photoelectric sensor that clearly reflects the target's appearance features.

[0093] In this embodiment of the application, the tracking results record the temporal location information of the target. Based on the imaging data of the photoelectric sensor associated with the location, the frames in which the target is located in the center of the imaging field of view are selected, and the corresponding infrared thermal imaging high-resolution images and visible light high-resolution images are extracted to ensure that the images contain the complete appearance features of the target.

[0094] Based on the above steps, a clear image of the target's appearance can be obtained.

[0095] S502. Input the target high-resolution image into the pre-trained deep learning image classification algorithm.

[0096] Among them, the deep learning image classification algorithm is an algorithm based on deep neural networks learning the deep features of the target and making classification determination. In this application, it is specifically the ResNet algorithm or the DenseNet algorithm.

[0097] In this embodiment, a ResNet or DenseNet algorithm is trained in advance using a labeled image dataset containing various low-altitude, slow-moving, and small targets such as consumer drones, industrial drones, pigeons, and eagles. After training, the model has the ability to extract deep features such as target texture, shape, and color distribution. Before input, the high-resolution target image is normalized to fit the model's input requirements, and then the processed image is input into the trained model.

[0098] It should be noted that the model was trained on data from various scenarios, including rain, fog, and nighttime, and has strong environmental adaptability.

[0099] Based on the above steps, mature deep learning models can be used to efficiently extract key features of the target, providing technical support for classification and judgment.

[0100] S503. Obtain the target category probability output by the deep learning image classification algorithm, determine the specific type of the target based on the target category probability, and obtain the classification and recognition result.

[0101] Among them, the target category probability is the numerical value of the possibility that the target belongs to each preset category, and the classification and recognition result is a judgment conclusion that clearly defines the specific category of the target.

[0102] In this embodiment, the algorithm outputs the probability values ​​of the target belonging to each preset category (the sum is 100%). The preset categories include consumer drones, industrial drones, pigeons, eagles, etc. A probability threshold is set (e.g., 80%). If the probability of a certain category exceeds the threshold, the category is determined as the specific type of the target. If the probability of all categories does not exceed the threshold, it is marked as an "unknown target". Finally, a classification and recognition result containing the specific type of the target is formed.

[0103] Based on the above steps, it is possible to accurately distinguish different types of low-altitude, slow-moving, small targets.

[0104] Based on the above technical solution, by accurately extracting high-resolution images of targets, using pre-trained deep learning classification algorithms, and determining target types based on probability thresholds, efficient and accurate classification and identification of low-altitude, slow, and small targets is achieved, thereby improving the intelligence level of the detection system.

[0105] In one possible implementation of the embodiments of this application, combined with Figure 2 ,like Figure 6 As shown, the above S206 can be implemented through the following S601, S602 and S603, which are explained in detail below: S601. Calculate the target threat index and set the target threat index threshold.

[0106] Among them, the threat index is a risk assessment value that integrates multiple dimensions of information such as target distance, speed, and type, and the threat index threshold is the critical judgment standard for triggering an early warning.

[0107] In this embodiment of the application, the target threat index is calculated; the distance weight is set according to the distance between the target and the protected area, the closer the target is, the higher the weight (value 0-1); the speed weight is set according to the target's flight speed (the faster the speed, the higher the weight, value 0-1); the type weight is set according to the target category (e.g., industrial drone weight 0.9, bird weight 0.3); and the target threat index threshold is set to 0.8.

[0108] Optionally, the target threat index satisfies: Threat Index = 0.4 × Distance Weight + 0.3 × Speed ​​Weight + 0.3 × Type Weight.

[0109] As an example, in an airport scenario, a drone is 1km away from the protection boundary (distance weight 0.8), has a speed of 18m / s (speed weight 0.7), and a type weight of 0.8. The calculated threat index is 0.4×0.8+0.3×0.7+0.3×0.8=0.77.

[0110] Based on the above steps, the threat level of a target can be quantified, providing an objective basis for early warning.

[0111] S602. When the target threat index is greater than the target threat index threshold and the target enters the GIS electronic fence area, an early warning is triggered.

[0112] The GIS electronic fence is a virtual boundary of a protected area preset by a geographic information system, used to define the spatial range for target warning. In this embodiment, the target's location coordinates are obtained in real time through the GIS system to determine whether it falls within the preset electronic fence area. If the target threat index is greater than the target threat index threshold and the location is within the electronic fence, an early warning is immediately triggered. The early warning methods include audible and visual alarms at the monitoring center, pop-up notifications on the screen, sending text messages and push notifications to relevant personnel, and synchronously transmitting the target's real-time location, flight speed, specific type, and movement trajectory information.

[0113] It should be noted that the shape (such as circle or polygon) and range of the GIS electronic fence can be customized according to protection needs, and the early warning methods can be combined and activated as needed.

[0114] As an example, in an airport scenario, an industrial-grade drone with a threat index of 0.86 enters a circular electronic fence with a radius of 3km, triggering an audible and visual alarm and a push notification to the security personnel's mobile application, simultaneously transmitting the target trajectory information.

[0115] Based on the above steps, it is possible to achieve accurate early warning of both risk compliance and spatial boundary violations, thus avoiding interference from ineffective early warnings.

[0116] S603. After triggering the warning, the interception device is activated and a dispatch message is sent.

[0117] Interception devices are used to jam or obstruct low-altitude, slow-moving, and small targets, while dispatch information is an instruction to relevant personnel to handle the situation, containing core target information.

[0118] In this embodiment, the corresponding interception device is activated according to the target type: for drone targets, directional acoustic jamming equipment or laser jamming equipment is activated to block their control signals; at the same time, dispatch information containing the target's real-time location, movement trajectory, specific type and threat level is generated and sent to the terminal of the security department or emergency response personnel.

[0119] Based on the above steps, threats can be responded to quickly, and the target risks can be effectively contained through a combination of technical interception and manual handling.

[0120] Based on the above technical solutions, a closed-loop mechanism of assessment, early warning, and response has been established through quantitative threat assessment, dual-condition early warning, and targeted response. This significantly improves the response efficiency and accuracy of handling threats from low-altitude, slow, and small targets, and effectively safeguards the security of key areas.

[0121] The foregoing mainly describes the solutions of the embodiments of this application from the perspective of device implementation. It is understood that each device, such as a low-altitude slow-moving small target detection device, includes at least one of the hardware structures and software modules corresponding to the execution of each function in order to achieve the above-mentioned functions. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0122] This application embodiment can divide the low-altitude slow-moving small target detection device into functional units according to the above method example. For example, each function can be divided into separate functional units, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software functional unit. It should be noted that the unit division in this application embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0123] When using integrated units, Figure 7 A possible structural schematic diagram of the low-altitude slow-moving small target detection device (referred to as low-altitude slow-moving small target detection device 70) involved in the above embodiments is shown. The low-altitude slow-moving small target detection device 70 includes a processing unit 701 and a communication unit 702, and may also include a storage unit 703. Figure 7 The schematic diagram shown can be used to illustrate the structure of the low-altitude slow-moving small target detection device involved in the above embodiments.

[0124] when Figure 7 The schematic diagram shown illustrates the structure of the low-altitude slow-moving small target detection device involved in the above embodiments. The processing unit 701 is used to control and manage the operation of the low-altitude slow-moving small target detection device, the communication unit 702 is used for the low-altitude slow-moving small target detection device to communicate with other devices, and the storage unit 703 is used to store the program code and data of the low-altitude slow-moving small target detection device.

[0125] For example, communication unit 702 is used to acquire a detection dataset; the detection dataset includes first data, second data, and third data; the first data is acquired by millimeter-wave radar; the second data is acquired by an optoelectronic sensor; the third data is acquired by an acoustic sensor; the optoelectronic sensor includes an infrared thermal imager and a visible light camera; The processing unit 701 is used to preprocess the detection dataset; perform target detection on the preprocessed detection dataset to obtain target detection results; the target detection results include a first detection result, a second detection result, and a third detection result; perform multi-sensor data fusion on the first detection result, the second detection result, and the third detection result, and continuously track the target using a target tracking algorithm to obtain tracking results; based on the tracking results, classify and identify the tracking results using an image classification algorithm to obtain classification and identification results; based on the target detection results, tracking results, and classification and identification results, set corresponding warning thresholds to issue warnings and take emergency response measures.

[0126] In one possible implementation, the processing unit 701 is further configured to preprocess the detection dataset, including: amplifying the first data, filtering out interference signals using a bandpass filter, and extracting effective radar signals including target range and velocity information; grayscale processing the second data, removing image noise using median filtering or Gaussian filtering algorithms, and enhancing the target edge contour using the Sobel operator or Canny operator; filtering the third data using an adaptive bandstop filter with an adjustable center frequency, separating mixed sound sources through independent component analysis, extracting the target acoustic wave characteristic frequency band, and performing normalization processing.

[0127] In one possible implementation, the processing unit 701 is further configured to perform target detection on the preprocessed detection dataset to obtain target detection results, including: using the CFAR algorithm to perform threshold decision on the preprocessed first data, and then obtaining the first detection result through Doppler frequency shift analysis; using a deep learning-based target detection algorithm to perform target detection and classification on the preprocessed second data to obtain the second detection result; and performing energy detection and feature frequency analysis on the preprocessed third data to determine the target acoustic wave features and obtain the third detection result.

[0128] In one possible implementation, the first detection result includes the target's distance information, velocity information, and location information of the suspected target area; the second detection result includes the target's coordinate position in the image and target category information; and the third detection result includes the determination result of the presence of the target's acoustic features, the feature frequency matching degree, and the signal occurrence time and location information.

[0129] In one possible implementation, the processing unit 701 is further configured to perform multi-sensor data fusion on the first detection result, the second detection result, and the third detection result, and continuously track the target using a target tracking algorithm to obtain tracking results. This includes: setting a confidence threshold; determining whether the confidence level of the first detection result is greater than the confidence threshold; if so, setting a sensor data synchronization window to synchronize the three types of detection results; performing joint target state estimation on the synchronized detection results using a Kalman filter algorithm to obtain the target's optimal position and optimal velocity information; and continuously tracking the target based on the optimal position and optimal velocity information using a joint probabilistic data association filtering algorithm and a multi-model adaptive tracking algorithm, and outputting the tracking results. The tracking results include the target's real-time position and motion trajectory.

[0130] In one possible implementation, a joint probability data association filtering algorithm is used to calculate the association probability between the target and the measurement data of each sensor; a multi-model adaptive tracking algorithm establishes multiple dynamic models for different motion modes of the target and adaptively switches models according to the actual motion of the target.

[0131] In one possible implementation, the processing unit 701 is further configured to classify and identify the tracking results based on the tracking results using an image classification algorithm, thereby obtaining classification and identification results. This includes: extracting a high-resolution image of the target acquired by a photoelectric sensor from the tracking results; inputting the high-resolution image of the target into a pre-trained deep learning image classification algorithm; the deep learning image classification algorithm being either a ResNet algorithm or a DenseNet algorithm; obtaining the target category probability output by the deep learning image classification algorithm; determining the specific type of the target based on the target category probability; and obtaining the classification and identification results.

[0132] In one possible implementation, the processing unit 701 is further configured to set corresponding warning thresholds based on target detection results, tracking results, and classification and identification results, and to take emergency response measures, including: calculating the target threat index and setting a target threat index threshold; triggering a warning when the target threat index is greater than the target threat index threshold and the target enters the GIS electronic fence area; the warning includes audible and visual alarms, screen pop-ups, SMS messages, and application push notifications, and synchronously transmitting target location, speed, type, and trajectory information; after triggering the warning, activating the interception device and issuing dispatch information; the dispatch information is used to dispatch personnel to the target location for handling.

[0133] In one possible implementation, the deep learning-based object detection algorithm is either the YOLO series or the Faster R-CNN algorithm.

[0134] The processing unit 701 can be a processor or a controller, and the communication unit 702 can be a communication interface, transceiver, transceiver circuit, transceiver device, etc. The term "communication interface" is a general term and may include one or more interfaces. The storage unit 703 can be a memory. When the low-altitude slow-moving small target detection device 70 is a chip, the processing unit 701 can be a processor or a controller, and the communication unit 702 can be an input interface and / or an output interface, pins, or circuits, etc. The storage unit 703 can be a storage unit within the chip (e.g., a register, cache, etc.) or a storage unit located outside the chip (e.g., read-only memory (ROM), random access memory (RAM, etc.).

[0135] The communication unit can also be called a transceiver unit. The antenna and control circuit with transceiver functions in the low-altitude slow-moving small target detection device 70 can be considered as the communication unit 702 of the low-altitude slow-moving small target detection device 70, and the processor with processing functions can be considered as the processing unit 701 of the low-altitude slow-moving small target detection device 70. Optionally, the device in the communication unit 702 used to implement the receiving function can be considered as the communication unit. The communication unit is used to execute the receiving steps in the embodiments of this application, and the communication unit can be a receiver, a receiver circuit, etc. The device in the communication unit 702 used to implement the transmitting function can be considered as the transmitting unit. The transmitting unit is used to execute the transmitting steps in the embodiments of this application, and the transmitting unit can be a transmitter, a transmitter, a transmitting circuit, etc.

[0136] Figure 7 If the integrated units in the process are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, in essence, or the parts that contribute to the prior art, or all or part of the technical solutions, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. Storage media for storing computer software products include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0137] Figure 7 The units in the process can also be called modules; for example, a processing unit can be called a processing module.

[0138] This application also provides a hardware structure diagram of a low-altitude slow-moving small target detection device (referred to as low-altitude slow-moving small target detection device 80), see [link to diagram]. Figure 8 The low-altitude slow-moving small target detection device 80 includes a processor 801, and optionally, a memory 802 connected to the processor 801.

[0139] In the first possible implementation, see Figure 8 The low-altitude slow-moving small target detection device 80 also includes a transceiver 803. The processor 801, memory 802, and transceiver 803 are connected via a bus. The transceiver 803 is used to communicate with other devices or communication networks. Optionally, the transceiver 803 may include a transmitter and a receiver. The device in the transceiver 803 that implements the receiving function can be considered as a receiver, which is used to perform the receiving steps in the embodiments of this application. The device in the transceiver 803 that implements the transmitting function can be considered as a transmitter, which is used to perform the transmitting steps in the embodiments of this application.

[0140] Based on the first possible implementation method Figure 8 The schematic diagram shown can be used to illustrate the structure of the low-altitude slow-moving small target detection device involved in the above embodiments.

[0141] in, Figure 8 This can also be illustrated by the system chip in the low-altitude slow-moving small target detection device. In this case, the actions performed by the aforementioned low-altitude slow-moving small target detection device can be implemented by this system chip. The specific actions performed can be found above and will not be repeated here.

[0142] In implementation, each step of the method provided in this embodiment can be completed by integrated logic circuits in the processor or by instructions in software form. The steps of the method disclosed in the embodiments of this application can be directly manifested as being executed by a hardware processor, or being executed by a combination of hardware and software modules in the processor.

[0143] The processor in this application may include, but is not limited to, at least one of the following: a central processing unit (CPU), a microprocessor, a digital signal processor (DSP), a microcontroller unit (MCU), or an artificial intelligence processor, etc., which are various computing devices that run software. Each computing device may include one or more cores for executing software instructions to perform calculations or processing. The processor may be a separate semiconductor chip or integrated with other circuits into a single semiconductor chip. For example, it may be integrated with other circuits (such as encoding / decoding circuits, hardware acceleration circuits, or various bus and interface circuits) to form a SoC (System-on-a-Chip), or it may be integrated as a built-in processor within an ASIC. The ASIC with the integrated processor may be packaged separately or together with other circuits. In addition to the cores for executing software instructions to perform calculations or processing, the processor may further include necessary hardware accelerators, such as field-programmable gate arrays (FPGAs), PLDs (programmable logic devices), or logic circuits that implement dedicated logic operations.

[0144] The memory in the embodiments of this application may include at least one of the following types: read-only memory (ROM) or other types of static storage devices capable of storing static information and instructions; random access memory (RAM) or other types of dynamic storage devices capable of storing information and instructions; or electrically erasable programmable-only memory (EEPROM). In some scenarios, the memory may also be a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures that can be accessed by a computer, but is not limited thereto.

[0145] This application also provides a computer-readable storage medium including instructions that, when run on a computer, cause the computer to perform any of the methods described above.

[0146] This application also provides a computer program product containing instructions that, when run on a computer, cause the computer to perform any of the methods described above.

[0147] This application also provides a chip including a processor and an interface circuit. The interface circuit is coupled to the processor. The processor is used to run computer programs or instructions to implement the above-described method. The interface circuit is used to communicate with other modules outside the chip.

[0148] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software programs, implementation can be, in whole or in part, in the form of a computer program product. This computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device containing one or more servers, data centers, etc., that can be integrated with the medium. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state disks (SSDs)).

[0149] Although this application has been described herein in conjunction with various embodiments, those skilled in the art, by reviewing the accompanying drawings, disclosure, and appended claims, will understand and implement other variations of the disclosed embodiments in carrying out the claimed application. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple instances. A single processor or other unit can implement several functions listed in the claims. While different dependent claims may recite certain measures, this does not mean that these measures cannot be combined to produce good results.

[0150] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of this application as defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from the spirit and scope of this application. Thus, if such modifications and modifications of this application fall within the scope of the claims of this application and their equivalents, this application is also intended to include such modifications and modifications.

Claims

1. A method for detecting low-altitude, slow-moving, small targets, characterized in that, include: A detection dataset is acquired; the detection dataset includes first data, second data, and third data; the first data is acquired by millimeter-wave radar; the second data is acquired by a photoelectric sensor; the third data is acquired by an acoustic sensor; the photoelectric sensor includes an infrared thermal imager and a visible light camera; The detection dataset is preprocessed; Target detection was performed on the preprocessed detection dataset to obtain the target detection results; The target detection results include a first detection result, a second detection result, and a third detection result; The first detection result, the second detection result, and the third detection result are fused together using multi-sensor data, and the target is continuously tracked using a target tracking algorithm to obtain the tracking result. Based on the tracking results, the tracking results are classified and identified using an image classification algorithm to obtain the classification and identification results. Based on the target detection results, the tracking results, and the classification and recognition results, a corresponding early warning threshold is set to issue an early warning, and emergency response measures are taken.

2. The method according to claim 1, characterized in that, The detection dataset is preprocessed, including: The first data is amplified, and a bandpass filter is used to filter out interference signals to extract effective radar signals including target distance and speed information. The second data is converted to grayscale, and image noise is removed by median filtering or Gaussian filtering algorithms. The Sobel operator or Canny operator is used to enhance the target edge contour. The third data is filtered using an adaptive band-stop filter with an adjustable center frequency. The mixed sound sources are separated by independent component analysis, and the characteristic frequency bands of the target sound wave are extracted and normalized.

3. The method according to claim 1, characterized in that, The process of performing target detection on the preprocessed detection dataset to obtain target detection results includes: The CFAR algorithm is used to make a threshold decision on the preprocessed first data, and then the first detection result is obtained through Doppler frequency shift analysis. A deep learning-based object detection algorithm is used to perform object detection and classification on the preprocessed second data to obtain the second detection result. Energy detection and characteristic frequency analysis are performed on the preprocessed third data to determine the target acoustic wave characteristics and obtain the third detection result.

4. The method according to claim 3, characterized in that, The first detection result includes the target's distance information, speed information, and location information of the suspected target area; the second detection result includes the target's coordinate position in the image and target category information; the third detection result includes the determination result of the presence of the target's acoustic features, the feature frequency matching degree, and the signal occurrence time and location information.

5. The method according to claim 1, characterized in that, The first detection result, the second detection result, and the third detection result are fused using multi-sensor data, and the target is continuously tracked using a target tracking algorithm to obtain the tracking result, including: Set a confidence threshold, determine whether the confidence of the first detection result is greater than the confidence threshold, and if so, set a sensor data synchronization window to synchronize the three types of detection results; The target state is jointly estimated by the synchronized detection results using the Kalman filter algorithm to obtain the target's optimal position and optimal velocity information. A joint probabilistic data association filtering algorithm and a multi-model adaptive tracking algorithm are used to continuously track the target based on the optimal position and optimal velocity information, and output the tracking results; the tracking results include the target's real-time position and motion trajectory.

6. The method according to claim 5, characterized in that, The joint probability data association filtering algorithm is used to calculate the association probability between the target and the measurement data of each sensor; the multi-model adaptive tracking algorithm establishes multiple dynamic models for different motion modes of the target and adaptively switches models according to the actual motion of the target.

7. The method according to claim 1, characterized in that, Based on the tracking results, an image classification algorithm is used to classify and identify the tracking results, resulting in classification and identification results, including: Extract the high-resolution image of the target acquired by the photoelectric sensor from the tracking results; The target high-resolution image is input into a pre-trained deep learning image classification algorithm; the deep learning image classification algorithm is either the ResNet algorithm or the DenseNet algorithm. Obtain the target category probability output by the deep learning image classification algorithm, determine the specific type of the target based on the target category probability, and obtain the classification and recognition result.

8. The method according to claim 1, characterized in that, Based on the target detection results, the tracking results, and the classification and recognition results, a corresponding early warning threshold is set to issue an early warning, and emergency response measures are taken, including: Calculate the target threat index and set the target threat index threshold; When the target threat index exceeds the target threat index threshold and the target enters the GIS electronic fence area, an early warning is triggered; the early warning includes audible and visual alarms, screen pop-ups, SMS messages, and application push notifications, and simultaneously transmits the target's location, speed, type, and trajectory information; Upon triggering the warning, the interception device is activated and a dispatch message is sent; the dispatch message is used to dispatch personnel to the target location for handling.

9. The method according to claim 3, characterized in that, The deep learning-based object detection algorithm is either the YOLO series algorithm or the Faster R-CNN algorithm.

10. A low-altitude, slow-moving, small target detection system, characterized in that, The system includes: millimeter-wave radar, photoelectric sensors, acoustic sensors, and electronic devices; The millimeter-wave radar is used to collect the first data; The photoelectric sensor includes an infrared thermal imager and a visible light camera, used to collect second data; The acoustic sensor is used to collect third data; The electronic device is used to acquire a detection dataset; the detection dataset includes first data, second data, and third data; preprocess the detection dataset; perform target detection on the preprocessed detection dataset to obtain target detection results; the target detection results include a first detection result, a second detection result, and a third detection result; fuse the first detection result, the second detection result, and the third detection result into multi-sensor data, and continuously track the target using a target tracking algorithm to obtain tracking results; based on the tracking results, classify and identify the tracking results using an image classification algorithm to obtain classification and identification results; based on the target detection results, the tracking results, and the classification and identification results, set corresponding early warning thresholds to issue early warnings and take emergency response measures.

Images