A method and system for detecting small targets at sea

Abstract

The application discloses a kind of offshore small target radar detection method and system, it is related to radio detection technical field;Method includes: real-time acquisition obtains radar echo data, visible light image, infrared image, AIS signal and navigation attitude data;Radar echo data is subjected to sea clutter suppression and pre-filtering, and target track is corrected in combination with historical radar plot set, obtains radar single-source target track set;Visible light image and infrared image are subjected to target detection, and obtain optical target detection set and infrared target detection set;Radar single-source target track set, optical target detection set, infrared target detection set, AIS signal and navigation attitude data are associated and fused, finally carry out deception feature identification and consistency check, obtain updated fusion target track set;The application can accurately identify offshore human confrontation and deception behavior, provide reliable support for offshore safety monitoring, maritime supervision and emergency response.

Description

Technical Field

[0001] This invention relates to the field of radio detection technology, and more specifically, to a radar detection method and system for small targets at sea. Background Technology

[0002] In practical applications where high reliability of detection is crucial, such as coastal defense, maritime surveillance, and security in key sea areas, it is essential to utilize small maritime target detection technology and multi-source information fusion systems to achieve accurate identification, continuous tracking, and risk warning of various targets, thereby ensuring maritime traffic order and preventing security threats. Existing technologies typically rely on multi-source sensing devices to acquire target appearance features, motion trajectories, and auxiliary correlation information. Through model training, they learn the inherent attributes and motion patterns of targets under normal conditions, as well as the inherent correlation between auxiliary information and targets, thereby establishing judgment criteria for target detection and identification. Specifically, these technologies utilize target morphological features captured by optical sensors and target reflection signal features acquired by radar equipment, combined with identity and navigation information provided by automatic identification systems (AIS), to construct a multi-dimensional feature fusion model. Target detection and verification are then achieved through pattern matching and feature comparison.

[0003] However, the core design logic of existing technologies relies on the feature distribution and correlation patterns under normal application scenarios, without considering situations where deliberate human intervention and deception are employed. In key monitored sea areas, some targets deliberately evade detection or mislead judgment through specific means. For example, they may use low-reflectivity coating materials to alter their reflection characteristics of radar signals, weaken the target morphological features under optical sensors through structural design or camouflage, deploy decoys with similar target signals or appearance features to interfere with detection, or even forge identification marks and navigation data in auxiliary information to confuse target attributes. Because existing technologies do not specifically model the feature distortions, false signals, and information falsification caused by such human intervention, the detection criteria trained by them will be undermined by these deception methods, making it impossible to distinguish between real targets and artificially created false targets, and also difficult to identify real targets that have been disguised.

[0004] The aforementioned problems can lead to systemic detection blind spots and judgment biases in existing technologies when faced with human adversarial and deception tactics. Specifically, this manifests as missed detection of disguised key targets, misjudging decoys as real targets, or misidentifying target attributes due to false auxiliary information. Such situations directly reduce the reliability of maritime surveillance and defense systems, preventing timely detection of security threats and potentially triggering unnecessary resource allocation and emergency response, resulting in a waste of human and material resources.

[0005] In view of this, the present invention proposes a radar detection method and system for small targets at sea to solve the above problems. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art and achieve the above objectives, the present invention provides the following technical solution: a method for detecting small targets at sea using radar, comprising:

[0007] Real-time acquisition of radar echo data, visible light images, infrared images, AIS signals, and navigation attitude data;

[0008] Sea clutter suppression and pre-filtering are performed on the radar echo data to obtain the initial radar target trace list;

[0009] By combining historical radar track sets with the target tracks in the initial radar target track list, a set of radar single-source target tracks is obtained.

[0010] Target detection is performed on visible light images and infrared images to obtain optical target detection sets and infrared target detection sets;

[0011] Spatially align and temporally synchronize radar single-source target track sets, optical target detection sets, infrared target detection sets, AIS signals, and navigation attitude data to obtain multi-source observation sets;

[0012] By correlating and fusing multi-source observation sets, a fused target track set is obtained;

[0013] Deception feature identification and consistency verification are performed on the fused target track set to obtain an updated fused target track set;

[0014] Decision output and closed-loop control are performed based on the updated fused target trajectory set.

[0015] Furthermore, a constant false alarm rate (CFAR) detection algorithm combined with sea clutter statistical characteristics is used to perform target screening on radar echo data to obtain suspected targets; feature parameters of each suspected target are extracted to obtain an initial radar target point list.

[0016] Furthermore, each detectable unit in the radar echo data is divided into adjacent reference units; the power statistics of sea clutter within the reference units are calculated, and a dynamic detection threshold is determined in combination with a preset false alarm rate; detectable units with echo power higher than the preset dynamic detection threshold are identified as suspected targets.

[0017] Furthermore, the reference cells for each unit to be detected are set individually based on the position of each unit to be detected in the two-dimensional plane, and they do not overlap.

[0018] Furthermore, by analyzing the changes in the number of suspected targets, spatial distribution characteristics, correlation probability with historical tracks, and whether the signal characteristics of the track itself meet the preset reasonable range in the radar initial target track list of the current frame, it is determined whether there is multiple false target interference among the suspected targets.

[0019] Furthermore, when it is determined that there is interference from multiple false targets, a threshold for the effective duration of the track is set. Tracks formed by points that have appeared consecutively for a number of frames greater than or equal to the threshold for the effective duration of the track and have established associations with other frame points are retained, while points that have appeared consecutively for a number of frames less than the threshold for the effective duration of the track and have not been able to establish associations with other frame points are removed.

[0020] Based on the position, timestamp, and azimuth of the dot, the motion speed, acceleration, and direction change of the dot are calculated. If any of the motion speed, acceleration, or direction change of a dot exceeds the preset reasonable range, the corresponding dot is filtered out.

[0021] Furthermore, methods for deception feature identification of fused target track sets include:

[0022] A radar signal camouflage determination mechanism is established based on the average reflection intensity of radar single-source target tracks, optical target detection sets, and infrared target detection sets.

[0023] Based on optical target detection features and infrared target detection features, an optical and infrared camouflage determination mechanism is established;

[0024] Based on the spatial distribution and movement correlation of the target group, a decoy target group determination mechanism is established;

[0025] Based on the protocol compliance and multi-source consistency of AIS data, a multi-level AIS spoofing verification mechanism is established.

[0026] Furthermore, the establishment of a radar signal camouflage determination mechanism specifically includes: if the geographic coordinates of the fused target track of the current target do not match any optical or infrared target detection, then the radar decoy determination process is initiated to determine whether the current target is a suspected radar decoy; otherwise, the radar stealth determination process is initiated to determine whether the current target is a suspected radar stealth target.

[0027] The establishment of an optical and infrared camouflage determination mechanism specifically includes: determining whether the current target is a suspected visual camouflage target based on the optical target detection features and infrared target detection features corresponding to the current fused target track; if the current target is not a suspected visual camouflage target, determining whether the current target is a suspected optical decoy based on the matching degree between the current target's motion speed and the sea surface wave motion speed at the same time.

[0028] Furthermore, the establishment of a decoy target group determination mechanism specifically includes: performing spatial clustering analysis on the fused target track set; if the number of tracks in the target group exceeds a preset group size threshold, then initiating the decoy target group analysis process to determine whether the target group is a spatially regularly distributed target group; if the target group is a spatially regularly distributed target group, then based on the multi-source observation coverage and motion similarity of the spatially regularly distributed target group, determining whether the target group is a suspected parent target and a suspected decoy group member;

[0029] The establishment of a multi-level AIS deception verification mechanism specifically includes: performing consistency verification on static attributes in AIS signals to determine whether static information is suspected to be falsified; performing rationality verification on dynamic data in AIS signals to determine whether dynamic data is suspected to be falsified; performing compliance verification on the arrival time interval of AIS signals to determine whether protocol violations are suspected; and determining whether the current target is a suspected ghost target or a suspected AIS cloaked target based on the multi-source observation coverage of the fused target track.

[0030] A radar detection system for small targets at sea, comprising:

[0031] The data acquisition module is used to acquire radar echo data, visible light images, infrared images, AIS signals, and navigation attitude data in real time.

[0032] The track acquisition module is used to suppress sea clutter and pre-filter radar echo data to obtain an initial list of radar target points.

[0033] The track correction module is used to correct the target tracks in the initial radar target track list by combining the historical radar track set, so as to obtain the radar single-source target track set.

[0034] The target detection module is used to perform target detection on visible light images and infrared images to obtain optical target detection sets and infrared target detection sets;

[0035] The data alignment module is used to spatially align and temporally synchronize the radar single-source target track set, optical target detection set, infrared target detection set, AIS signal, and navigation attitude data to obtain a multi-source observation set.

[0036] The data fusion module is used to correlate and fuse multi-source observation sets to obtain a fused target track set;

[0037] The data update module is used to perform deception feature identification and consistency verification on the fused target track set to obtain the updated fused target track set.

[0038] The decision output module performs decision output and closed-loop control based on the updated fused target trajectory set.

[0039] Compared with the prior art, the technical effects and advantages of the radar detection method and system for small targets at sea according to the present invention are as follows:

[0040] This invention acquires radar echo data, visible light images, infrared images, AIS signals, and navigation attitude data in real time. After sea clutter suppression and pre-filtering, it obtains an initial radar target track list. Combined with historical radar track sets, it corrects and obtains a radar single-source target track set. At the same time, it completes target detection of visible light and infrared images. After spatial alignment and temporal synchronization of multi-source data, it correlates and fuses the data. Then, it uses a multi-dimensional anti-spoofing criterion system to identify deception features and verify consistency. Finally, based on the verification results, it realizes hierarchical situation output and closed-loop control.

[0041] This invention effectively solves the problems of existing technologies that fail to consider human countermeasures and deception, leading to missed detection of camouflaged targets, misjudging decoys as real targets, and incorrect target attribute identification due to false information. These problems reduce the reliability of maritime surveillance and defense systems, waste resources, and even cause safety accidents. The invention can accurately identify various deception behaviors such as radar stealth, visual camouflage, optical decoys, and AIS spoofing, filter out false target interference, improve the accuracy of target detection and identification, provide reliable support for maritime safety monitoring, maritime supervision, and emergency response, and ensure maritime traffic order and the safety and stability of the sea area. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of a small target radar detection system at sea according to an embodiment of the present invention;

[0043] Figure 2 This is a flowchart of a radar detection method for small targets at sea according to an embodiment of the present invention;

[0044] Figure 3 This is a schematic diagram of a method for deception feature recognition of a fused target track set according to an embodiment of the present invention. Detailed Implementation

[0045] The technical solutions of the embodiments of the present invention will be described in detail, clearly, and completely below with reference to the accompanying drawings. It should be particularly noted that the specific embodiments described below are only for better illustrating and explaining the technical solutions of the present invention, and are intended to enable those skilled in the art to better understand and implement the present invention, and should not be construed as limiting the scope of protection of the present invention. Without departing from the spirit and substance of the present invention, those skilled in the art can modify, adjust, or make equivalent substitutions based on the content disclosed in the present invention, and these should all be considered within the scope of protection of the present invention.

[0046] Example 1:

[0047] Please see Figure 1As shown in the figure, this embodiment discloses a small target radar detection system at sea, including a data acquisition module, a track acquisition module, a track correction module, a target detection module, a data alignment module, a data fusion module, a data update module, and a decision output module. Each module is connected by wires and / or wirelessly to realize data transmission.

[0048] The data acquisition module is used to acquire radar echo data, visible light images, infrared images, AIS signals, and navigation attitude data in real time.

[0049] Radar echo data is obtained by the active transmission and reception of echoes from the surface surveillance radar carried by the unmanned platform, recording raw signals such as target azimuth, range, and echo power. Employing an X-band pulse compression system, it can detect targets in all weather conditions and complex sea states. Visible light images are obtained from visible light video frames or images acquired by the electro-optical imaging equipment carried by the unmanned platform. Typically, a high-definition camera captures sea surface scenes under daytime conditions, providing information on target shape and color. The camera is mounted at a high position and equipped with a stabilized gimbal to ensure a clear horizontal field of view on the moving platform. Infrared images are acquired by a mid-wave or long-wave infrared thermal imager carried by the unmanned platform, containing images of the thermal radiation intensity of the sea surface scene. Infrared images are used to detect the temperature characteristics of targets, and can still detect heat sources such as engines at night or under camouflage. The infrared camera is calibrated and its field of view is aligned with that of the visible light camera for subsequent fusion. AIS signals are broadcast information from the Automatic Identification System for Vessels, acquired by the AIS receiver on the unmanned platform. This includes static identification information such as the vessel's name, call sign, length, and type, as well as dynamic track data such as position, speed, and heading. This data is received via the VHF band and parsed to obtain the fields for each AIS message. Navigation attitude data consists of the unmanned platform's own GPS positioning and IMU inertial navigation data, used for time synchronization of sensor data and unifying the detection results to a geographic coordinate system. Navigation attitude data includes the unmanned platform's latitude, longitude, altitude, heading, attitude angles, and timing information, ensuring that the data from various sources are fused in a time- and spatially aligned manner.

[0050] The aforementioned data is collected in real time by their respective sensors and transmitted to the system, providing the original basis for subsequent multi-source information fusion and anti-spoofing analysis.

[0051] The track acquisition module is used to suppress sea clutter and pre-filter radar echo data to obtain an initial list of radar target points.

[0052] Methods for suppressing and pre-filtering sea clutter in radar echo data include:

[0053] A constant false alarm rate (CFAR) detection algorithm combined with sea clutter statistical characteristics is used for target screening. Specifically, in the range-azimuth two-dimensional data plane of the radar echo data, each target unit is independently divided into adjacent reference units. Each target unit's reference unit is individually set based on its position in the two-dimensional data plane, ensuring they do not overlap and only cover the surrounding sea clutter distribution area. By calculating the power statistics of sea clutter within the corresponding reference units for each target unit, and combining this with a preset false alarm rate, a dynamic detection threshold specific to each target unit is determined. Since the power distribution of sea clutter within the reference units corresponding to different target units varies, the dynamic detection threshold for each target unit is different to accurately adapt to the clutter intensity in the local area. The echo power of each target unit is compared one by one with its specific dynamic detection threshold. Target units with echo power higher than the dynamic detection threshold are identified as suspected targets. Feature parameters of each suspected target are extracted. These feature parameters are calculated based on the original measurement information and processed signal characteristics of the radar echo data, thus forming the initial radar target trace list for the current radar scanning cycle.

[0054] Doppler filtering is applied to remove slow clutter caused by sea waves. By designing a targeted Doppler filter bank, the difference in distribution between the slow clutter and the target signal in the Doppler frequency domain is utilized. This preserves the signal components within the Doppler frequency shift range corresponding to the target, while filtering out slow sea clutter outside this range. Furthermore, pulse accumulation gain is used to improve the signal-to-noise ratio (SNR) of low radar cross-section (RCS) targets. By superimposing the energy of the echo signals of the same target across multiple consecutive pulse periods, the target energy dispersed in multiple pulses is concentrated and aggregated. Simultaneously, the sea clutter energy, lacking a fixed phase correlation, does not synchronously increase, thus relatively increasing the amplitude level of the target signal and improving the SNR of low RCS targets. This improved SNR allows the signals of low RCS targets, previously masked by sea clutter, to become more prominent, reducing missed detections due to weak signals and providing more reliable signal support for subsequent target identification and trace extraction.

[0055] After the above processing, the radar outputs a list of initial target points within the current radar scanning cycle, which includes information such as the polar coordinates, reflection intensity, Doppler frequency shift, signal amplitude, dynamic detection threshold, and point confidence score for each point.

[0056] The track correction module is used to correct the target tracks in the initial radar target track list by combining the historical radar track set, so as to obtain a set of radar single-source target tracks.

[0057] The radar system continuously collects data according to a fixed scanning cycle. Each scanning cycle generates a corresponding initial radar target track list. Therefore, each frame in the multi-target tracking process has an independent current frame radar initial target track list. A multi-target tracking algorithm is used to temporally correlate the radar initial target tracks of consecutive frames. The specific correlation process is as follows: First, an existing target trajectory model is constructed based on the historical radar track set. The state of the next frame for each historical trajectory is predicted using a Kalman filter to obtain the predicted position, velocity, and other state information of the target. Then, each track in the current frame radar initial target track list is matched with all predicted states. If the nearest neighbor method is used, the spatial distance and feature similarity between the current track and each predicted state are calculated, and the predicted trajectory that is closest and meets the similarity threshold is selected as the matching object. If the joint probabilistic data correlation algorithm is used, the correlation probability between the current track and each predicted trajectory is calculated, and the optimal correlation combination is determined according to the probability distribution. The temporal correlation of the tracks of consecutive frames is completed in the above manner, thereby generating a stable target trajectory.

[0058] False candidate targets that are not actual targets in the radar's initial target track list are considered multiple false target interference. Multiple false target interference consists of false components in the suspected targets generated by factors such as sea clutter remnants and electronic interference. The method for determining the existence of multiple false target interference is to analyze the changes in the number of suspected targets in the radar's initial target track list in the current frame, their spatial distribution characteristics, their correlation probability with historical tracks, and whether the signal characteristics of the tracks themselves conform to a preset reasonable range. Specific methods include:

[0059] The system pre-generates the average number of suspected targets under normal sea conditions and sets a threshold for the ratio of the number of suspected targets to the average number. When the ratio of the number of suspected targets in the current frame to this average number is greater than the threshold, it is determined that the number of targets has increased abnormally. The system calculates the correlation probability between newly added suspected targets and historical tracks in the current frame, and presets a correlation threshold. This correlation threshold is determined based on the statistical results of the correlation probability between historical real targets and historical tracks. If the correlation probability between all newly added suspected targets and historical tracks is lower than the preset correlation threshold, and an effective temporal correlation cannot be established, then the correlation probability judgment condition is met. The system also pre-generates the reflection intensity range and Doppler frequency shift range of typical real targets and constructs the reflection intensity range and Doppler frequency shift range corresponding to sea clutter and electronic interference. The proportion of newly added suspected targets whose reflection intensity falls within the range of sea clutter or electronic interference reflection intensity, and the proportion of targets whose Doppler frequency shift falls within the range of sea clutter or electronic interference Doppler frequency shift, are calculated. A feature matching threshold is set. If both proportions are higher than the feature matching threshold, the signal features are determined to be concentrated in the interference feature range, which does not match the feature distribution of typical real targets. Spatial distribution analysis is performed on newly added suspected targets in the current frame. The geographical coordinate distance between any two newly added suspected targets is calculated to obtain the mean distance and standard deviation of the distance. A dispersion distance threshold and a distribution discreteness threshold are set. If the mean distance is greater than the dispersion distance threshold and the standard deviation of the distance is greater than the distribution discreteness threshold, the target is determined to be spatially irregularly dispersed.

[0060] An abnormal increase in the number of false targets is a direct manifestation of multi-false target interference. A high degree of matching between signal characteristics and interference intervals is the core basis for distinguishing false targets from real targets. A correlation probability below the correlation threshold indicates the lack of a sustained trajectory. These three factors together constitute the core judgment conditions for multi-false target interference; satisfying these three core conditions is sufficient to preliminarily determine the presence of multi-false target interference. Irregular spatial distribution is a typical spatial characteristic of multi-false target interference, serving as an auxiliary judgment condition. If all four conditions are met, the credibility of the judgment result is further enhanced, avoiding misjudgments caused by anomalies in a single dimension. This setting is suitable for scenarios with clear core characteristics, such as strong electronic interference, reducing missed detections, while also improving judgment accuracy through auxiliary conditions, thus balancing application needs under different sea conditions and interference intensities.

[0061] When multiple false target interference is detected, isolated random points are filtered out through trajectory consistency checks. The specific operations are as follows: a threshold for the effective duration of the track is generated in advance. The threshold for the effective duration of the track is determined based on the radar scan cycle and the statistical results of the average number of continuous occurrences of historical real targets. Cross-frame association is performed on the radar initial target point list of the current frame and multiple consecutive frames. Only tracks formed by points whose consecutive occurrences are greater than or equal to the threshold for the effective duration of the track and which have established associations with points in other frames are retained. Points whose consecutive occurrences are less than the threshold for the effective duration of the track and which cannot establish associations with points in other frames are removed.

[0062] The consistency of the motion state of the points is verified by pre-generating a reasonable range of motion speeds for typical small maritime targets. This reasonable range includes a lower speed threshold and an upper speed threshold. The motion speed is calculated by the ratio of the positional change between adjacent frames to the time interval. If the motion speed is less than the lower speed threshold or greater than the upper speed threshold, it is considered an abnormal motion speed. An azimuth angle abrupt change threshold is pre-generated, calculated based on the maximum turning angular velocity of a typical small maritime target and the radar scan period. The absolute value of the azimuth angle difference between adjacent frames is calculated. If the absolute value of the azimuth angle difference is greater than the azimuth angle abrupt change threshold, it is considered a sudden change in motion direction. A maximum acceleration threshold for typical small maritime targets is pre-generated. The motion acceleration is calculated by the ratio of the velocity change between adjacent frames to the time interval. If the motion acceleration is greater than the maximum acceleration threshold, the motion state is considered to have no reasonable physical explanation. When a point meets any one of the three conditions—abnormal motion speed, sudden change in motion direction, and motion acceleration greater than the maximum acceleration threshold—it is considered an isolated random point and filtered out.

[0063] The final output is a set of radar single-source target tracks. Each radar single-source target track contains information such as track number, position, azimuth, velocity, acceleration, mean Doppler frequency shift, mean reflection intensity, track confidence, number of continuous tracking frames, and state estimation covariance matrix. This provides a structured radar observation sequence for subsequent multi-source data spatial alignment and time synchronization, as well as multi-source target association and fusion.

[0064] The target detection module is used to perform target detection on visible light images and infrared images to obtain optical target detection sets and infrared target detection sets.

[0065] First, horizon detection is performed to segment the visible light image into sky and sea areas to narrow the search range. In the sea area near the horizon, a trained small target detection model is applied to scan the visible light image. The training method for the small target detection model is as follows: based on a sample dataset containing small targets at different sea conditions, camouflage types, and scales, each sample in the dataset is labeled with its bounding box, segmentation contour, and corresponding camouflage type label. An adapted network architecture for small target detection is adopted, using a joint loss function consisting of target classification loss and bounding box regression loss as the optimization objective. Through multiple rounds of iterative training, the network parameters are adjusted to enable the small target detection model to accurately identify small targets from complex sea backgrounds.

[0066] To counter maritime camouflage, a combination of background subtraction and motion detection is employed. The specific methods include: constructing a Gaussian mixture model to dynamically model the pixel value distribution of the sea surface background; updating the background model in real time to adapt to dynamic changes in wave undulation; performing a difference operation between the current frame's pixel value and the background model to obtain foreground candidate regions, filtering out background regions where pixel value changes are less than a set threshold; calculating the grayscale difference between the current frame and several adjacent frames, retaining regions with significant differences, and taking the intersection with the foreground candidate regions obtained from background subtraction to initially screen moving objects; further, combining optical flow field analysis to calculate the motion vectors of each pixel between adjacent frames, statistically analyzing the directional consistency and velocity stability of all pixel motion vectors within the target candidate region, eliminating regions with chaotic motion vectors and velocity fluctuations exceeding reasonable ranges, and retaining target candidates with unified motion direction, stable velocity, and continuous shape contours, thereby reducing interference from blurred target-background boundaries caused by maritime camouflage.

[0067] For targets disguised by color or shape, a disguised target detection technique is introduced to conduct refined analysis in the dimensions of texture and chromaticity. Specific methods include: texture feature extraction using local binary mode and gray-level co-occurrence matrix to calculate texture feature vectors for the target candidate region and the surrounding background region, and quantifying the texture similarity between the two using cosine similarity; chromaticity analysis converting the image to the Lab color space and calculating the mean, variance, and peak value of the target candidate region and the surrounding background region in the L, a, and b channels respectively, constructing a chromaticity feature difference matrix to quantify the difference between the two in each color channel; combining texture similarity and chromaticity difference to obtain a fusion difference value, setting a fusion difference threshold, and identifying regions with fusion difference values ​​higher than the threshold as abnormal regions. Contrast enhancement, morphological dilation, and erosion processing are used to highlight the contour details of abnormal regions, thereby identifying disguised targets whose color is similar to the background and whose shape is obscured by the background texture.

[0068] After the above processing, an optical target detection set is output. This set includes targets identified through background subtraction, motion detection, and camouflage target detection techniques, as well as uncamouflaged small maritime targets. Specifically, the optical target detection set includes pixel coordinates or line-of-sight orientation, bounding boxes, segmentation contours, texture feature vectors, chromaticity feature parameters, and target confidence scores for each type of target. These are used for subsequent multi-source data spatial alignment and temporal synchronization, as well as multi-source target association and fusion.

[0069] Infrared images are preprocessed to eliminate non-uniformity, and horizon detection is used for region filtering to exclude non-target areas such as the sky and bright clouds, searching only for targets within the sea surface background area. Since small infrared targets typically appear as tiny bright spots lacking shape details, this embodiment employs multi-scale filtering combined with local contrast enhancement to improve the distinction between the target and the background. Specifically, multi-scale filtering involves constructing multiple filter templates of different sizes to perform layered filtering on the infrared image. Small-sized templates enhance the signal strength of tiny bright spot targets, while medium- and large-sized templates suppress sea surface background clutter at different scales. Through complementary filtering of multi-scale templates, preliminary separation of the target signal from background clutter is achieved. Local contrast enhancement involves defining a fixed-size local region centered on each pixel, calculating the average grayscale difference between the central pixel and all other pixels within the local region, and using this average grayscale difference as the enhanced grayscale value of the central pixel. By amplifying the grayscale difference between the target pixel and the surrounding background pixels, the prominence of the target in the complex sea surface background is further enhanced. After the above operations, adaptive threshold detection is performed on the enhanced image. The detection threshold is dynamically adjusted according to the gray-scale statistical characteristics of the local area of ​​the image, and several bright pixel clusters are initially selected as target candidates.

[0070] To reduce false alarms caused by sunlight reflection from the sea surface or cloud noise, features of each target candidate are extracted, including hotspot size, brightness distribution, and shape moments. These features are then input into a trained classifier to filter out false alarms that do not conform to the characteristics of small targets. Since infrared decoy interference can be used to deceive thermal imaging, this process also analyzes the target's movement and persistence characteristics: real targets are usually persistent and move smoothly over time, while infrared decoys may be briefly bright and drift irregularly. Infrared hotspots that do not conform to the actual flight path pattern are eliminated through inter-frame correlation and trajectory smoothing. The specific operations are as follows: Inter-frame correlation calculates the spatial distance and grayscale feature similarity between each target candidate in the current frame and all target candidates in the previous and two previous frames to establish the correspondence between target candidates in different frames; Trajectory smoothing performs trajectory fitting on the target candidate points with established correspondence, calculates the motion speed and direction change rate of the temporary trajectory, and if the number of continuous frames of the temporary trajectory is less than a set threshold, the motion speed fluctuation exceeds the reasonable motion range of small targets at sea, or the motion direction changes abruptly without a reasonable physical explanation, it is determined to be an infrared hotspot corresponding to the infrared decoy and is eliminated; For infrared hotspots corresponding to temporary trajectories with a number of continuous frames that meet the set requirements and whose motion speed and direction change are stable, they are retained as valid target candidates.

[0071] The final infrared target detection set includes information such as the field of view direction, thermal intensity, pixel coordinates, bounding box, average brightness, number of frames that persist, motion speed, and target confidence score of effective target candidates. This information is used to reference the infrared characteristics of the target during subsequent spatial alignment and temporal synchronization of multi-source data and multi-source target association and fusion.

[0072] The data alignment module is used to spatially align and temporally synchronize radar single-source target track sets, optical target detection sets, infrared target detection sets, AIS signals, and navigation attitude data to obtain multi-source observation sets.

[0073] The spatial alignment process uses a unified geographic coordinate system as a reference, and performs coordinate transformation on the observation data of each sensor one by one. Specific methods include:

[0074] Based on the latitude, longitude, altitude, and attitude angle of the unmanned platform in the navigation attitude data, and combined with the polar coordinates of each track in the radar single-source target track set, a coordinate mapping is completed through a polar coordinate-geographic coordinate transformation model. This converts the relative polar coordinate position measured by the radar into absolute coordinates in the geographic coordinate system. At the same time, an attitude angle compensation factor is introduced to correct the impact of platform attitude changes on radar measurement accuracy, ensuring the accuracy of the target position after conversion.

[0075] The pixel coordinates of targets in the optical and infrared target detection sets are converted into ray direction vectors in the camera coordinate system using camera intrinsic parameters. Then, combined with the platform's GPS positioning information, platform attitude angles, and camera extrinsic parameters in the navigation attitude data, the ray direction vectors are converted into azimuth and elevation angles relative to the platform. A dual-mode strategy is adopted to obtain the target distance. Specifically, if the target is covered by the radar single-source target track set, the target distance measured by the radar is directly reused. If the target is detected purely visually or purely by infrared, two types of ranging methods are used to obtain the distance value. One is a geometric ranging method based on the horizon height, which uses the camera installation height, the pixel position of the horizon in the image, and the camera intrinsic parameters to derive the target distance through geometric calculations. The other is an auxiliary ranging method, which directly obtains the target distance through the parallax calculation of the binocular camera or the measurement data of the laser rangefinder. Finally, the target's azimuth, elevation angle, and distance value are combined to complete the conversion from relative platform coordinates to the geographic coordinate system.

[0076] The target latitude and longitude information carried in the AIS signal is directly mapped to the geographic coordinate system. In response to the discrepancy between the timestamp of the AIS signal and the unified time reference, a linear extrapolation model is used based on the speed and heading in the AIS signal to extrapolate the AIS target position to the unified target time, ensuring the time consistency of AIS observation with observations from other sensors.

[0077] The methods for time synchronization include: using GPS timing information in navigation attitude data as a unified time reference, assigning precise timestamps to each observation data in radar single-source target track sets, optical target detection sets, infrared target detection sets, and AIS signals; addressing the inconsistency in observation times caused by differences in sampling frequencies of different sensors, a combination of time interpolation and extrapolation is used for correction: for sensor observation data with higher sampling frequencies, linear interpolation is used to supplement the observation state at missing moments; for sensor data with lower sampling frequencies or observation gaps, extrapolation is performed based on the historical motion state of the target to obtain target observation data under a unified time reference; simultaneously, a fixed-duration time window is set, and multi-source observation data within the same time window are grouped together to ensure that the time deviation of each source observation data within the group is within a preset allowable range, thereby achieving approximate synchronization of multi-source data.

[0078] The final output multi-source observation set includes the geographic coordinates of each target in each group of synchronous observations, the original feature parameters of each source, timestamps, coordinate transformation accuracy indicators, and observation source identifiers, providing a unified, accurate, and spatiotemporally consistent data source support for subsequent multi-source target association and fusion.

[0079] The data fusion module is used to correlate and fuse multi-source observation sets to obtain a fused target track set.

[0080] To achieve complementary enhancement of multi-source information and improve the accuracy of target state estimation and the reliability of attribute identification, a multi-source target association and fusion operation is performed based on the spatiotemporally aligned multi-source observation set. Through cross-source association matching, weighted information fusion and conflict collaborative processing, a unified and robust fused target track set is constructed.

[0081] The association matching phase employs a cross-source association strategy based on radar single-source target tracks as the core benchmark. This strategy combines adaptive spatial neighborhood constraints and multi-dimensional similarity verification to achieve accurate matching. Specifically, the method includes: first, dynamically setting an adaptive spatial neighborhood threshold for each radar single-source target track based on the measurement error characteristics of each sensor, target motion patterns, and environmental interference intensity under current sea conditions. This threshold is dynamically adjusted according to target distance, speed, and sea clutter intensity to ensure the targeted and effective association search; then, for each radar single-source target track, searching within the adaptive spatial neighborhood threshold around its geographic coordinates for the same time... The system collects observation data corresponding to optical target detection, infrared target detection, and AIS signals within the window. For the searched candidate cross-source observations, it calculates multi-dimensional correlation similarity indices, including azimuth deviation, range deviation, speed, heading matching degree, and feature similarity. Feature similarity includes the degree of matching between radar reflection intensity and optical texture and color features, as well as infrared thermal intensity features. The comprehensive correlation similarity score is obtained by weighted summation. A correlation similarity threshold is set. When the comprehensive correlation similarity score between the candidate cross-source observation and the radar single-source target track is higher than the correlation similarity threshold, it is determined that the two belong to the same real target, and a cross-source correlation relationship is established.

[0082] For scenarios without radar coverage, optical target detection and infrared target detection are used as joint correlation benchmarks. Their spatial deviation, feature similarity, and motion state consistency are calculated to establish an optical-infrared cross-source correlation. The observation data corresponding to AIS signals are then compared with optical and infrared target detection data to calculate positional deviations and analyze motion state matching, thus determining the correlation between AIS and optical and infrared observations. Targets observed by only a single sensor are retained as candidate tracks, labeled with single-source observation identifiers and sensor types, and included in the fused target track set for subsequent tracking and verification, ensuring that no potential real targets are missed.

[0083] In the multi-source information fusion stage, dynamic fusion weights are assigned to different source observation data based on the observation reliability and measurement accuracy of each sensor. Specifically, radar single-source target tracks have significant advantages in distance measurement accuracy and are given a high position fusion weight; optical target detection has outstanding azimuth measurement accuracy and is given a high azimuth fusion weight; the thermal intensity characteristics of infrared target detection are environmentally independent and are given a high attribute fusion weight; and the identity information of AIS signals, after compliance verification, is given a high attribute fusion weight, and its dynamic track data is given corresponding auxiliary position fusion weights according to the degree of position deviation. The fusion process employs an extended Kalman filter algorithm to achieve accurate estimation of motion state. Specifically, the method includes: using the state estimation result of a single-source radar target track as the initial prediction value; using the successfully matched position and azimuth observation data of optical and infrared target detection, and the position and speed observation data of AIS signals as update inputs; and achieving optimal fusion estimation of the target's geographic coordinates, velocity, acceleration, azimuth, and other motion states through a prediction-correction iterative process. Weighted fusion of attribute information from each source is performed, including the mean radar reflection intensity, texture feature vectors and chromaticity feature parameters of optical target detection, the mean thermal intensity of infrared target detection, and the ship name, call sign, length, ship type, speed, and heading of the AIS signals, to generate a comprehensive set of target attribute features.

[0084] In the information conflict handling phase, a multi-scenario conflict determination and collaborative processing mechanism is established. Specific methods include: setting a conflict determination threshold; when the positional deviation, motion state deviation, or attribute characteristic deviation of different source observation data exceeds the threshold, an information conflict is determined to exist. For conflicts between radar single-source target tracks and AIS signals, two independent tracks are retained, labeled as the radar main navigation track and the AIS main navigation track respectively, and the conflict type is recorded as positional conflict, motion state conflict, or attribute conflict. Simultaneously, the deviation value or inconsistency degree is recorded as a conflict quantification indicator. For conflicts between optical target detection, infrared target detection, and radar single-source target tracks, the reliability assessment results of each sensor under the current environment are considered. For example, the reliability of optical sensors decreases in foggy conditions, while the stability of radar observations improves in rainy conditions. The fusion weights of each source are readjusted. If the conflict is still not alleviated after adjustment, the radar single-source target track is retained as the main navigation track, and the conflicting optical or infrared observations are marked as suspicious observations and included in the observation data of subsequent time windows for consistency verification. For situations where multiple source observations conflict with each other, each single-source candidate track is retained, marked with a high-conflict indicator, and awaits cross-verification of subsequent multi-frame observation data to clarify the target's authenticity.

[0085] The final fused target track set contains a unique target identifier, geographic coordinates, velocity, acceleration, azimuth, mean reflection intensity, texture feature vector, chromaticity feature parameters, mean thermal intensity, AIS identity information, static attributes and dynamic track data, fusion weight allocation records for each source observation, cross-source association matching logs, conflict identifiers and conflict details, track confidence scores and continuous fusion frame counts. This provides comprehensive, accurate and confidence-labeled fused data support for subsequent deception feature identification and consistency verification.

[0086] The data update module is used to perform deception feature identification and consistency verification on the fused target track set to obtain the updated fused target track set.

[0087] Please see Figure 3 As shown, to accurately identify human-induced countermeasures and deception behaviors against maritime targets and ensure the reliability of target detection and identification, a multi-dimensional anti-deception criterion system is constructed based on multi-source observation data, comprehensive motion state estimation results, and comprehensive attribute feature sets from the fused target track set. Through scenario-specific deception feature extraction, quantitative verification, and multi-source consistency fusion, a deception risk assessment and deception type labeling are completed for each fused target track. Specific methods include:

[0088] Based on the average reflection intensity of radar single-source target tracks, optical target detection sets, and infrared target detection sets, a radar signal camouflage determination mechanism is established. The specific method includes: for each fused target track, extracting the average reflection intensity corresponding to the radar single-source target track; simultaneously, based on the geographic coordinates of each fused target track, defining a preset spatial neighborhood in the optical and infrared target detection sets, and searching for matching optical or infrared target detections; if only radar observation data exists, and there are no matching observations within the preset spatial neighborhood of the optical and infrared target detection sets, then the radar decoy determination process is initiated. Specifically, this involves calling a pre-built typical maritime target radar cross-section (RCS) database, which contains the RCS values ​​of typical targets of different ship types and sizes under different sea conditions; setting an abnormal RCS threshold; if the RCS of the current target is less than or equal to the average RCS of a typical small boat of the same size... If the ratio exceeds the radar cross section anomaly threshold and there is no optical or infrared observation evidence for multiple consecutive frames, the current target's deception type is labeled as a suspected radar decoy, and key information such as the radar cross section anomaly multiple and the number of frames without multi-source evidence is recorded. If there is a clear matching observation in the preset spatial neighborhood of the optical target detection set and the infrared target detection set, but the average reflection intensity of the current target is lower than the preset ratio threshold of the average reflection intensity of typical targets of the same size, the radar stealth determination process is initiated. The specific method is as follows: after confirming that the target is a real entity by combining the target's optical morphology and infrared thermal intensity characteristics, the current target's deception type is labeled as a suspected radar stealth target, the threat rating of the current target is increased to the preset high threat level, and sensor control commands are generated to reduce the radar detection threshold of the corresponding area to the preset sensitivity threshold range in subsequent radar scanning cycles, or switch to the radar high-gain scanning mode to extend the radar dwell time in the area and enhance the continuous tracking of the current target.

[0089] Based on optical and infrared target detection features, an optical and infrared camouflage determination mechanism is established. The specific method includes: extracting optical and infrared target detection features corresponding to each fused target track; optical target detection features include texture feature vectors, chromaticity feature parameters, and target confidence scores; infrared target detection features include average thermal intensity, average brightness, and number of sustained frames. If the average thermal intensity of the infrared target detection is higher than a preset thermal intensity threshold, and the target confidence score of the optical target detection is lower than a preset optical confidence threshold, it is determined to be a suspected visual camouflage scene. The mechanism analyzes whether the shape of the infrared target's heat source conforms to the typical heat source distribution of a ship's engine, and whether the changes in speed and azimuth angle conform to the ship's navigation patterns. If both analysis results meet preset conditions, the current target's deception type is labeled as a suspected visual camouflage target, triggering a special alarm indicating the presence of a visually stealth target. If the target confidence score of the optical target detection is higher than a preset optical confidence threshold, and the average thermal intensity of the infrared target detection is lower than a preset optical confidence threshold, it is determined to be a suspected visual camouflage scene. If the average temperature is lower than the preset low heat threshold, it is judged as a suspected optical decoy scenario. The target's movement mode is analyzed, and the matching degree between the target's movement speed and the movement speed of the sea surface waves at the same time is calculated. If the matching degree is higher than the preset wave matching threshold, it is judged as a static decoy floating with the waves. If the target's movement trajectory is associated with other targets at a fixed distance and the synchronicity of its movement speed is higher than the preset synchronization threshold, it is judged as a decoy being towed. If the target's movement speed exceeds the reasonable speed range of typical ships or the movement direction changes irregularly, the deception type label of the current target is marked as a suspected optical decoy, and its track confidence score is reduced to the preset low confidence interval.

[0090] Based on the spatial distribution and motion correlation of target groups, a decoy target group determination mechanism is established. The specific steps are as follows: Spatial clustering analysis is performed on the fused target track set. The geographical coordinate distance between any two fused target tracks is calculated. Tracks with a distance less than a preset clustering distance threshold are grouped into the same target group. The number of tracks in each target group is counted. If the number of tracks in a target group exceeds a preset group size threshold, the decoy target group analysis process is initiated. The spatial layout characteristics of the target group are extracted. The distance and angle distributions of each track relative to the center coordinates of the target group are calculated. If the standard deviation of the distance distribution is less than a preset uniformity threshold and the angle distribution shows a uniform interval characteristic, the target group is determined to be a spatially regularly distributed target group. The target group is checked for multi-source observation coverage. If only radar observations completely cover the target group, and optical and infrared target detections only have a single low-confidence observation in the central area of ​​the target group, with no matching observations in other areas, the process of distinguishing between the parent target and the decoy is initiated. The specific method is as follows: calculate the similarity of the motion trajectory of each track in the target group with the central low-confidence observation, mark the central track with the highest trajectory similarity as the suspected parent target, increase its track confidence score, mark the deception type label of other tracks in the surrounding area as suspected decoy group members, reduce their display priority or temporarily store them in the candidate track list in subsequent processing, and only retain the core tracking information of the parent target.

[0091] Based on the protocol compliance and multi-source consistency of AIS data, a multi-level AIS spoofing verification mechanism is established. Specific methods include: extracting the ship length and hull type from the static attributes of the AIS signal and comparing them with the target size and morphological features of optical target detection calculated based on pixel coordinates and ranging data; calculating the ship length error rate and hull type matching degree; if the ship length error rate exceeds a preset size error threshold or the hull type matching degree is lower than a preset type matching threshold, the static attributes of the AIS signal are marked as suspected static information spoofing; using a Kalman trajectory filtering algorithm to smooth the position, speed, and heading in the dynamic data of the AIS signal, calculating the filtered speed change rate and heading change rate; if the change rate exceeds a preset physical limit threshold, it is determined that there is physically impossible motion; calculating the deviation value between the AIS dynamic position and the radar and optical fusion position; if the deviation value exceeds a preset threshold for multiple consecutive frames, it is considered as a spoofing. A position deviation threshold is set, and targets are marked as potentially spoofed dynamic data. The arrival time interval of AIS signals is monitored, and the broadcast frequency per unit time is counted. If the arrival time interval does not conform to the time slot allocation rules specified in the TDMA protocol, or the broadcast frequency exceeds the preset reasonable frequency range, the arrival time interval of AIS signals is marked as a suspected protocol violation. If the fused target track only contains AIS signal observations, and there are no radar, optical, or infrared observations within the preset spatial neighborhood, and the theoretical value of the radar cross-section corresponding to the ship type declared by AIS is higher than the preset detection threshold, then the current target's deception type label is marked as a suspected ghost target. If the fused target track contains radar, optical, and infrared observations, but there is no corresponding AIS signal reception record, and the target size conforms to the AIS mandatory installation standard, then the current target's deception type label is marked as a suspected AIS stealth target, and its suspicious level is increased.

[0092] Based on the evidence-based reasoning fusion theory, a multi-source information consistency assessment mechanism is established. The specific methods include: constructing a multi-source consistency assessment index system, including position consistency, attribute consistency, and motion state consistency, and assigning preset weights to each assessment index; employing an evidence-based reasoning fusion algorithm to assign initial confidence levels to radar, optical, infrared, and AIS source observation data; if a source observation data shows no anomalies after the aforementioned verification, the initial confidence level is retained; if suspected deception features are present, the confidence level of the corresponding source is reduced according to the degree of anomaly; calculating the multi-source consistency comprehensive score; if the comprehensive score is higher than a preset high consistency threshold, it is determined that the multi-source information is consistent, and the track confidence score is increased; if the comprehensive score is lower than a preset low consistency threshold, it is determined that the multi-source information is conflicting, and the track confidence score is reduced, recording information such as the conflict source type and conflict quantification indicators; for multi-source information conflict scenarios, if the conflict originates from a contradiction between single-source observation data and other multi-source data, the observation results of the majority of sources are taken as the standard, and the fusion weight of the conflict source is weakened; if there are multi-source cross-conflicts, the independent observation information of each source is retained, marked as a high-conflict target, and included in the subsequent multi-frame cross-validation queue, awaiting further verification with more observation data.

[0093] Through the coordinated execution of the aforementioned multi-dimensional anti-spoofing criteria, accurate identification and classification of various deception behaviors, such as radar decoys, visual camouflage, optical decoys, and AIS forgery, are achieved, effectively filtering out false target interference and preventing single deception information from dominating the fusion result. By dynamically adjusting the sensor operating mode and target threat rating, the tracking and monitoring of high-risk deception targets are strengthened, improving the system's target detection reliability and anti-interference capability in adversarial environments. The final output is an updated set of fused target tracks, with each track accompanied by a complete deception type label, track confidence score, key parameters of deception features, and multi-source consistency assessment results, providing a comprehensive deception risk assessment basis for subsequent decision output and closed-loop processing.

[0094] The decision output module performs decision output and closed-loop control based on the updated fused target trajectory set.

[0095] To provide users with clear, accurate, and risk-oriented target situation information, and to enhance continuous monitoring capabilities in adversarial environments through dynamic adjustments to system operation strategies, a hierarchical situation output and closed-loop control system is constructed based on a fused target track set annotated with deception identification. Specific methods include:

[0096] The fused target track set labeled with deception identification is classified into three categories based on track confidence score and deception type label: real targets, suspicious targets, and decoy targets.

[0097] The set of fused target tracks with a confidence score higher than a preset high confidence threshold and no deception type labels or only verified non-deception anomaly markers are classified as real targets. Complete situational information is output, including the target's unique identifier, geographic coordinates, velocity, acceleration, azimuth, mean reflectance intensity, texture feature vector, chromaticity feature parameters, mean thermal intensity, AIS identity information and static attributes, dynamic track data, continuous tracking frame count, and multi-source consistency assessment results. Fusion target tracks with a confidence score between a preset medium confidence threshold and a high confidence threshold, or those carrying deception type labels such as suspected radar stealth targets, suspected visual camouflage targets, or suspected AIS stealth targets, are classified as real targets. The dataset is categorized into suspicious targets, and core situational information is output, including target geographic coordinates, key motion parameters, main attribute characteristics, deception type labels, and suspicion level. These are also highlighted on the display interface using a preset special identifier to prompt operators to pay close attention. Fusion target tracks with confidence scores below a preset low confidence threshold and carrying deception type labels such as suspected radar decoys, suspected optical decoys, suspected decoy swarm members, or suspected ghost targets are categorized as decoy targets. Simplified situational information is output, retaining only the target geographic coordinates, deception type labels, and confidence scores. These are either faded out or hidden on the display interface to avoid interfering with the operator's judgment of genuine targets.

[0098] The system integrates situational information from three target categories to generate a structured target situational list. The list is sorted in descending order by target suspicion level or threat rating, ensuring high-risk targets are prioritized. For suspected radar stealth targets, suspected visual camouflage targets, and suspected AIS concealed targets, corresponding alerts are triggered. Alert information includes target location, deception type, threat level, and handling recommendations, and is simultaneously pushed to display terminals and the command system via audio-visual signals or pop-ups. Standardized logs are maintained for all detected deception-related events. Log content includes event timestamps, target geographic coordinates, deception type tags, radar cross-section anomaly multiples, AIS data deviation values, multi-source consistency conflict indicators, track confidence score changes, handling measures, and execution results. Log data is stored in a local database according to a preset format and simultaneously reported to the superior command system in real time via an encrypted communication link, ensuring the traceability of deception events and timely response to command decisions.

[0099] Based on target situation classification results and deception type labels, targeted sensor control commands are generated to dynamically optimize sensor operating parameters and detection strategies, achieving closed-loop operation of perception, identification, decision-making, and control. Specific methods include:

[0100] For suspected radar stealth targets, parameter adjustment commands are sent to the radar system to lower the radar detection threshold in the target area to a preset sensitivity threshold range, switch to high-gain radar scanning mode, extend the radar dwell time in the area, and improve the radar's detection sensitivity for targets with low reflectivity. Simultaneously, the optical and infrared imaging equipment is controlled to adjust its focal length and field of view, focusing on the target area for high-frequency sampling to enhance the capture of target morphology and thermal characteristics. For suspected visual camouflage targets or suspected optical decoys, control commands are sent to the optical imaging equipment to activate multispectral imaging mode or polarization imaging mode to enhance the texture and color differences between the target and the background; the infrared thermal imager is controlled to improve thermal sensitivity, expand the thermal intensity detection range, and accurately capture weak heat source signals that may exist within the target, providing... Subsequent secondary verification of deception features provides data support; for suspected AIS cloaked targets or suspected ghost targets, a monitoring enhancement command is issued to the AIS receiver to expand the receiving bandwidth of the target's frequency band, improve signal receiving sensitivity, and continuously monitor the AIS signals that the target may transmit; at the same time, the radar system is coordinated to shorten the scanning cycle of the area, and the authenticity and motion status of the target are cross-verified by combining optical and infrared observation data; for decoy target groups, a command is issued to the multi-sensor collaborative control module to adjust the radar scanning strategy, reduce the detection priority of the area where the decoy target group is located, and allocate more detection resources to the area of ​​real targets or suspected targets; the optical and infrared imaging equipment is controlled to skip high-frequency sampling in the dense decoy area, and only periodic overview detection is retained to avoid resource waste.

[0101] The entire processing flow is executed cyclically according to a preset cycle. Each cycle updates the detection strategy and processing parameters for the next cycle based on the target situation output and sensor control feedback data of the current cycle. By analyzing the detection data after sensor adjustment, the accuracy of deception feature recognition and the effectiveness of sensor control strategy are evaluated, and various thresholds and sensor operating parameters are dynamically optimized. For persistent high-conflict targets or suspicious targets, a multi-frame cross-verification mechanism is initiated to integrate multi-cycle observation data to gradually clarify the authenticity of the target, ensuring that the system maintains a high level of reliable situational awareness capability in complex adversarial environments.

[0102] This embodiment achieves accurate information transmission through hierarchical situational awareness output and strengthens targeted detection of deceptive targets through closed-loop control strategies. It effectively enhances the system's ability to resist various camouflage and deception methods, ensuring that the output target situational awareness list and alarms have high credibility and practical guidance value, and providing efficient and reliable decision support for maritime safety monitoring, maritime supervision and emergency response.

[0103] Example 2:

[0104] Please see Figure 2 As shown, this embodiment provides a method for detecting small targets at sea using radar, including:

[0105] Real-time acquisition of radar echo data, visible light images, infrared images, AIS signals, and navigation attitude data;

[0106] Sea clutter suppression and pre-filtering are performed on the radar echo data to obtain the initial radar target trace list;

[0107] By combining historical radar track sets with the target tracks in the initial radar target track list, a set of radar single-source target tracks is obtained.

[0108] Target detection is performed on visible light images and infrared images to obtain optical target detection sets and infrared target detection sets;

[0109] Spatially align and temporally synchronize radar single-source target track sets, optical target detection sets, infrared target detection sets, AIS signals, and navigation attitude data to obtain multi-source observation sets;

[0110] By correlating and fusing multi-source observation sets, a fused target track set is obtained;

[0111] Deception feature identification and consistency verification are performed on the fused target track set to obtain an updated fused target track set;

[0112] Decision output and closed-loop control are performed based on the updated fused target trajectory set.

[0113] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0114] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A radar detection method for small targets at sea, characterized in that, include: Real-time acquisition of radar echo data, visible light images, infrared images, AIS signals, and navigation attitude data; Sea clutter suppression and pre-filtering are performed on the radar echo data to obtain the initial radar target trace list; By combining historical radar track sets with the target tracks in the initial radar target track list, a set of radar single-source target tracks is obtained. Target detection is performed on visible light images and infrared images to obtain optical target detection sets and infrared target detection sets; Spatially align and temporally synchronize radar single-source target track sets, optical target detection sets, infrared target detection sets, AIS signals, and navigation attitude data to obtain multi-source observation sets; By correlating and fusing multi-source observation sets, a fused target track set is obtained; Deception feature identification and consistency verification are performed on the fused target track set to obtain an updated fused target track set; Methods for deception feature identification of fused target track sets include: A radar signal camouflage determination mechanism is established based on the average reflection intensity of radar single-source target tracks, optical target detection sets, and infrared target detection sets. Based on optical target detection features and infrared target detection features, an optical and infrared camouflage determination mechanism is established; Based on the spatial distribution and movement correlation of the target group, a decoy target group determination mechanism is established; Based on the protocol compliance and multi-source consistency of AIS data, a multi-level AIS deception verification mechanism is established. This mechanism specifically includes: consistency verification of static attributes in AIS signals to determine if static information is suspected of being falsified; rationality verification of dynamic data in AIS signals to determine if dynamic data is suspected of being falsified; compliance verification of the arrival time interval of AIS signals to determine if protocol violations are suspected; and determination of whether the current target is a suspected ghost target or a suspected AIS occult target based on the multi-source observation coverage of the fused target trajectory. Decision output and closed-loop control are performed based on the updated fused target trajectory set.

2. The method for detecting small targets at sea using radar according to claim 1, characterized in that, A constant false alarm rate (CFAR) detection algorithm combined with sea clutter statistical characteristics is used to perform target screening on radar echo data to obtain suspected targets; feature parameters of each suspected target are extracted to obtain an initial radar target trace list.

3. The method for detecting small targets at sea using radar according to claim 2, characterized in that, Each target unit in the radar echo data is divided into adjacent reference units; the power statistics of sea clutter within the reference units are calculated, and a dynamic detection threshold is determined in combination with a preset false alarm rate; target units with echo power higher than the preset dynamic detection threshold are identified as suspected targets.

4. The method for detecting small targets at sea using radar according to claim 3, characterized in that, The reference cells for each unit to be detected are set individually based on the position of each unit to be detected in the two-dimensional plane, and they do not overlap.

5. The method for detecting small targets at sea using radar according to claim 3, characterized in that, By analyzing the changes in the number of suspected targets, spatial distribution characteristics, correlation probability with historical tracks, and whether the signal characteristics of the tracks themselves meet the preset reasonable range in the radar's initial target track list for the current frame, it can be determined whether there is interference from multiple false targets among the suspected targets.

6. The method for detecting small targets at sea using radar according to claim 5, characterized in that, When multiple false target interference is determined, a threshold for the effective duration of the track is set. Tracks formed by points that have a consecutive number of frames greater than or equal to the threshold for the effective duration of the track and have established association with other frame points are retained. Tracks that have a consecutive number of frames less than the threshold for the effective duration of the track and have not established association with other frame points are removed. Based on the position, timestamp, and azimuth of the dot, the motion speed, acceleration, and direction change of the dot are calculated. If any of the motion speed, acceleration, or direction change of a dot exceeds the preset reasonable range, the corresponding dot is filtered out.

7. The method for detecting small targets at sea using radar according to claim 1, characterized in that, The establishment of a radar signal camouflage determination mechanism specifically includes: if the geographic coordinates of the fused target track of the current target do not match the optical target detection or infrared target detection, then the radar decoy determination process is initiated to determine whether the current target is a suspected radar decoy; otherwise, the radar stealth determination process is initiated to determine whether the current target is a suspected radar stealth target. The establishment of an optical and infrared camouflage determination mechanism specifically includes: determining whether the current target is a suspected visual camouflage target based on the optical target detection features and infrared target detection features corresponding to the current fused target track; if the current target is not a suspected visual camouflage target, determining whether the current target is a suspected optical decoy based on the matching degree between the current target's motion speed and the sea surface wave motion speed at the same time.

8. The method for detecting small targets at sea using radar according to claim 1, characterized in that, The specific steps for establishing a decoy target group determination mechanism include: performing spatial clustering analysis on the fused target track set; if the number of tracks in the target group exceeds a preset group size threshold, then initiating the decoy target group analysis process to determine whether the target group is a spatially regularly distributed target group; if the target group is a spatially regularly distributed target group, then based on the multi-source observation coverage and motion similarity of the spatially regularly distributed target group, determining whether the target group is a suspected parent target and a suspected decoy group member.

9. A radar detection system for small maritime targets, used to implement the radar detection method for small maritime targets as described in any one of claims 1-8, characterized in that, include: The data acquisition module is used to acquire radar echo data, visible light images, infrared images, AIS signals, and navigation attitude data in real time. The track acquisition module is used to suppress sea clutter and pre-filter radar echo data to obtain an initial list of radar target points. The track correction module is used to correct the target tracks in the initial radar target track list by combining the historical radar track set, so as to obtain the radar single-source target track set. The target detection module is used to perform target detection on visible light images and infrared images to obtain optical target detection sets and infrared target detection sets; The data alignment module is used to spatially align and temporally synchronize the radar single-source target track set, optical target detection set, infrared target detection set, AIS signal, and navigation attitude data to obtain a multi-source observation set. The data fusion module is used to correlate and fuse multi-source observation sets to obtain a fused target track set; The data update module is used to perform deception feature identification and consistency verification on the fused target track set to obtain the updated fused target track set. The decision output module performs decision output and closed-loop control based on the updated fused target trajectory set.

Images