A method and device for detecting dynamic targets on the sea surface in a polarization change scenario

Abstract

The application provides a sea surface dynamic target fast detection method and device under a polarization change scene, belongs to the fields of photoelectric detection, polarization imaging and target detection, and comprises the following steps: acquiring the longitude, latitude, attitude angle and world time of a platform through a navigation positioning system, combining a sun position calculation model to judge whether a current imaging area falls into a sun glare zone in real time; when it is judged that the imaging area falls into the glare zone, recording the brightness of each angle image and locking the optimal polarization angle corresponding to the minimum brightness, and then tracking and adjusting the optimal polarization angle in real time according to the platform attitude; when it is judged that the imaging area does not fall into the glare zone, driving a displacement table to move the polarization plate as a whole out of the light path; performing target detection processing based on a deep network on the low-glare image obtained when the polarization plate is in the light path, and outputting the ship target position and category. The application realizes the continuous, stable and efficient identification of ship targets in the strong glare interference area.

Description

Technical Field

[0001] This invention belongs to the fields of photoelectric detection, polarization imaging, and target detection, and specifically relates to a method and device for rapid detection of dynamic targets on the sea surface under polarization change scenarios. Background Technology

[0002] In the fields of marine remote sensing and sea surface target monitoring, solar flare, as a strong specular reflection signal, can cause problems such as local image saturation, high background noise, and obscuring of target information, severely affecting image quality and target detection capabilities. To address this issue, researchers have extensively explored solar flare suppression methods based on polarization information in recent years, with related research mainly focusing on the following aspects:

[0003] 1. Polarization Adaptive Filtering Imaging System Design: This system utilizes background polarization direction information to control the rotation angle of the polarizer, making it perpendicular to the background polarization direction, thereby enhancing the contrast between the target and the background. However, this system only controls the polarization direction adjustment based on brightness changes in the target's observed image, lacking a collaborative judgment mechanism between external observation geometry and the relative position of the sun. It cannot predict glare areas based on actual positions, nor can it actively turn off the polarizer in non-glare areas to improve the signal-to-noise ratio.

[0004] 2. Polarization suppression method based on Stokes vector and Mueller matrix modeling: This method analyzes the influence of solar zenith angle and observed zenith angle on polarization suppression through theoretical modeling, proposing that the optimal observation conditions lie in the region where the sum of the zenith angle and the solar angle is approximately 106°. While providing guidance for imaging angle selection, it still relies on static observation parameters and experimental scenario settings, lacks practical equipment and dynamic adaptive control capabilities, and does not possess real-time adaptability on airborne or flight platforms.

[0005] 3. Temporal Polarization Fusion Suppression Method: This method acquires video frame images at three polarization angles (0°, 45°, and 90°), calculates the Stokes vector, and performs temporal fusion to generate a suppressed image. While suitable for video sequence analysis, this method suffers from poor real-time performance and time-consuming fusion process, making it difficult to deploy on resource-constrained UAVs or shipborne platforms.

[0006] 4. A physical model-based method for polarization reflection estimation and intensity decomposition: This method constructs a complete polarization transmission model from the light source to the target to the detector, derives a decoupling model for solar flare and underwater target polarized light, and corrects it by incorporating water attenuation. Although it boasts high accuracy and reasonable modeling, the algorithm is complex and relies heavily on computational resources and prior environmental parameters, making it difficult to deploy quickly for dynamic imaging tasks.

[0007] 5. Adaptive Polarization + Digital Domain Fusion Imaging System: This system combines polarization measurement with TDI-CMOS digital delay integration technology to suppress glare and improve image quality for moving targets. Although it integrates hardware and software systems, its control mechanism is still mainly based on adjusting the image brightness distribution and lacks a mechanism to assist in judging the glare state using external navigation data. This results in response lag issues on high-dynamic platforms (such as UAVs or ships).

[0008] In summary, existing methods for rapid detection of dynamic targets on the sea surface have the following problems: 1. Lack of proactive judgment capability for glare areas combined with POS (Positioning System) navigation information; 2. The working state and rotation angle adjustment of the polarizer rely heavily on image posterior analysis, resulting in slow response and poor stability; 3. Inability to automatically remove the polarization rotation module in non-glare areas, affecting imaging energy utilization efficiency; 4. Lack of integration and optimization with dedicated target detection algorithms, affecting target extraction capability. Summary of the Invention

[0009] To address the problems of poor real-time performance, low energy utilization, insufficient adaptability, and weak target detection performance in the detection of dynamic targets on the sea surface under solar glare interference, this invention provides a method and device for rapid detection of dynamic targets on the sea surface in polarization-changing scenarios. The system dynamically determines the state of the glare region based on the real-time position and attitude information of the imaging platform and automatically controls the switching and angle adjustment of the polarization rotation module. It also features a polarization imaging system with an efficient target detection algorithm. This invention can be widely applied to scenarios such as ship surveillance, maritime law enforcement patrols, and UAV remote sensing, achieving continuous, stable, and efficient identification of targets such as ships in areas with strong glare interference.

[0010] To achieve the above objectives, the present invention adopts the following technical solution:

[0011] A method for rapid detection of dynamic targets on the sea surface under polarization variation scenarios includes the following steps:

[0012] Step 1: Obtain latitude and longitude, platform attitude angle and world time through the navigation and positioning system, and determine in real time whether the current imaging area falls into the glare zone by combining the solar position calculation model;

[0013] Step 2: When it is determined that the light source has fallen into the glare zone, immediately drive the stepper motor's displacement stage to cut the polarizer into the optical path; when it is determined that the light source has fallen into the non-glare zone, drive the displacement stage to move the polarizer out of the optical path.

[0014] Step 3: After entering the glare zone, control the polarization rotation module to complete a 360° polarizer rotation scan at a constant angular velocity, record the imaging brightness at each angle, and obtain the optimal polarization suppression angle by filtering the minimum value; adjust and update the polarization angle in real time according to the platform attitude angle to obtain a low glare image after adding polarization glare suppression; the polarization suppression angle is defined as the rotation angle of the polarizer when suppressing sea surface glare interference to the greatest extent.

[0015] Step 4: Perform depth model-based target detection processing on the low-glare image obtained when the polarizer is in the optical path, and output the ship target position and category.

[0016] The present invention also provides a rapid detection device for dynamic targets on the sea surface under polarization variation scenarios, for implementing the above method, comprising the following modules:

[0017] The judgment module obtains latitude and longitude, platform attitude angle and world time through the navigation and positioning system, and combines the solar position calculation model to determine in real time whether the current imaging area falls into the glare zone.

[0018] The drive module immediately drives the displacement stage to insert the polarizer into the optical path when it determines that the light has fallen into the glare zone; when it determines that the light has fallen into the non-glare zone, it drives the displacement stage to move the polarizer out of the optical path.

[0019] The adjustment module, after entering the glare zone, controls the polarization rotation module to complete a 360° polarizer rotation at a constant angular velocity, records the imaging brightness at each angle, and obtains the optimal polarization suppression angle by filtering the minimum value; the polarization angle is adjusted and updated in real time according to the platform attitude to obtain a low glare image after adding polarization glare suppression; the polarization suppression angle is defined as the rotation angle of the polarizer when suppressing sea surface glare interference to the greatest extent.

[0020] The output classification module performs depth model-based target detection processing on the low-glare image obtained when the polarizer is in the optical path, and outputs the ship target position and category.

[0021] Beneficial effects:

[0022] 1. Real-time glare area determination capability, enhancing system initiative and response speed: This invention is the first to incorporate navigation and positioning system (POS) information into the joint calculation of imaging geometry and solar vectors, constructing a spatiotemporal prediction model for glare areas. Compared to existing methods that rely solely on image feedback to determine whether an area has entered a glare region, this invention achieves early judgment and early intervention, significantly reducing suppression failure and image saturation problems caused by delayed judgment.

[0023] 2. Dynamic switching of the polarization rotation module improves system imaging energy efficiency and image quality: This invention features a controllable entry / exit mechanism for the polarization rotation module, which automatically removes the polarizer from the imaging optical path in non-glare areas. This avoids unnecessary filtering loss by the polarizer, improves the overall light transmittance and signal-to-noise ratio of the image, and effectively ensures the detection capability of dim targets in non-glare areas. Most existing polarization rotation modules are permanently stationary, leading to unnecessary energy loss.

[0024] 3. Polarization angle self-calibration and dynamic tracking ensure optimal suppression effect: This invention obtains the optimal suppression angle through a single rotation scan of the polarization rotation module, and combines it with POS attitude information for real-time angle tracking and adjustment, ensuring that the suppression direction always remains orthogonal to the specular reflection direction, greatly improving the stability and accuracy of the suppression effect. Compared with existing suppression methods that only use fixed angle or interval scanning, this solution has a more precise suppression angle and stronger dynamic response capability.

[0025] 4. Integrating target detection algorithms to form a "suppression + recognition" closed-loop system: This invention not only completes the design of an imaging system for glare suppression, but also integrates a ship detection algorithm adapted to the characteristics of polarization-suppressed images. On the basis of reducing background clutter interference, it achieves high-precision and rapid recognition of sea surface targets, and constructs a complete "imaging-suppression-recognition" closed-loop system, which is significantly better than the traditional system that only focuses on the imaging level.

[0026] 5. The system is lightweight and intelligently controlled, making it easy to implement and deploy on platforms such as UAVs and ships: The entire system can use electronically controlled rotating polarizers, lightweight imaging modules and embedded navigation and control units in terms of hardware, and adopts a modular algorithm architecture in terms of software, which has high portability and engineering feasibility, and is conducive to its promotion to airborne, shore-based and shipborne remote sensing systems. Attached Figure Description

[0027] Figure 1a , Figure 1b , Figure 1c Here are the three views of the polarization imaging system; where, Figure 1a Main view, Figure 1b This is a side view. Figure 1c This is a top view;

[0028] Figure 2 This is a schematic diagram of the internal structure of a polarization imaging system;

[0029] Figure 3 This is a flowchart of a method for rapid detection of dynamic targets on the sea surface under polarization variation scenarios according to the present invention;

[0030] Figure 4a , Figure 4b , Figure 4c , Figure 4d , Figure 4e, Figure 4f , Figure 4g , Figure 4h This is an image showing the effect of continuous polarization suppression imaging; among them, Figure 4a , Figure 4c , Figure 4e , Figure 4g This is the original image of sea surface flare from a camera without the polarization rotation module engaged. Figure 4b , Figure 4d , Figure 4f , Figure 4h Image of the sea surface after glare suppression;

[0031] Figure 5a , Figure 5b , Figure 5c , Figure 5d , Figure 5e , Figure 5f This is a schematic diagram of the polarization target detection results; where, Figure 5a Scene 1 Figure 5b This is the second scene. Figure 5c This is scene 3. Figure 5d This is scene 4. Figure 5e This is scene 5. Figure 5f In scene 6, the red box indicates the target area marked by the object detection algorithm. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0033] like Figure 1a , Figure 1b , Figure 1c As shown, the polarization imaging system of this invention includes a front-facing lens, a polarization modulation assembly, an electronic compass, and a detector. Opening the device panel of the polarization imaging system reveals the main components, including a stepper motor, a polarization rotation module, a polarization rotation module support device, and a base, etc. Figure 2 As shown in the diagram, the stepper motor is connected to the base, and the polarization rotation module support device is fixed to the stepper motor's displacement platform using six Allen screws. The polarization rotation module (including the polarization rotation motor) is fixed to the top of the polarization rotation module support device using three Phillips head screws. The top plate of the polarization imaging system has multiple mounting holes for installing the required components.

[0034] A polarizer is installed on the polarization-rotating motor, and a non-polarized window is installed on the right side of the polarization-rotating module support device, which is used for visible light. Thus, the polarization and non-polarization light paths can be switched by moving the stepper motor left and right.

[0035] like Figure 3 As shown, a rapid detection method for dynamic targets on the sea surface under polarization variation scenarios according to the present invention includes the following steps:

[0036] Step 1: Construct a real-time method for determining glare regions based on spatial location information:

[0037] Equipped with a POS (Positioning and Positioning System), it acquires its spatial position and attitude information in real time. Using a solar position calculation model, combined with the imaging perspective and solar vector, it determines whether the current imaging area falls within a potential solar flare zone.

[0038] Step 2: Construct the polarization modulation component:

[0039] The polarization modulation component includes a stepper motor, a polarization rotation module, and a polarization rotation module support device. The polarization rotation module support device is fixed on the displacement stage of the stepper motor. Based on the result of the real-time judgment method of the glare area, the cutting-in and cutting-out operations of the polarization rotation module are executed by controlling the movement of the displacement stage. The polarization rotation module includes a polarization rotation motor.

[0040] When it is determined that the area is in a glare region, the polarization rotation module is controlled to rotate the polarizer to enter the optical path.

[0041] When it is determined that the area is in a non-glare region, the polarization rotation module is automatically moved out of the optical path.

[0042] Step 2 ensures suppression capability while maximizing the preservation of incident light intensity, thereby improving image quality and signal-to-noise ratio in non-glare areas.

[0043] Step 3: Construct an optimal polarization suppression direction self-calibration mechanism:

[0044] After entering the glare area, the polarization rotation module is controlled to complete a 360° rotation of the polarizer at a preset speed, recording the imaging brightness at each angle. The optimal suppression angle is obtained by filtering for the minimum value. Then, the rotation angle of the polarization rotation module is dynamically adjusted according to the POS posture change to always maintain alignment with the direction of minimum glare. Step 3 can accurately suppress the reflection of the sunglass surface and avoid the problem of suppression failure caused by relative polarization angle shift due to platform shaking and movement after fixing the polarization angle.

[0045] Step 4: Construct the object detection algorithm:

[0046] By combining the image features after glare suppression, multi-scale candidate region extraction and screening are performed based on image brightness distribution, edge features and polarization contrast; at the same time, time series information is integrated to improve stability and denoising ability.

[0047] Specifically, step 1 includes:

[0048] Based on the navigation and positioning system, this invention can obtain the current latitude and longitude coordinates and world time. Using these as a basis, this invention can calculate the solar flare angle at the current moment. The calculation steps are as follows:

[0049] Step 1.1: Calculate Julian Day. Let the time be Y (year), M (month), D (day), h (hour), m (minute), and s (second) in UTC. The formula for calculating Julian Day (JD) is as follows:

[0050] ;

[0051] Where Y, M, D, h, m, and s represent year, month, day, hour, minute, and second, respectively.

[0052] Step 1.2: Calculate the century number T: ;

[0053] Step 1.3: Calculate the geometric mean ecliptic longitude of the Sun. (Unit: degrees):

[0054] ;

[0055] Step 1.4: Calculate the mean anomaly angle K of the sun (unit: degrees):

[0056] ;

[0057] Step 1.5: Calculate the solar central difference angle C (unit: degrees):

[0058] ;

[0059] Step 1.6: Calculate the true ecliptic longitude of the sun. (Unit: degrees):

[0060] ;

[0061] Step 1.7: Calculate the apparent ecliptic longitude of the sun. (Unit: degrees):

[0062] ;

[0063] Step 1.8: Calculate the solar declination angle (Unit: degrees):

[0064] ;

[0065] Among them, intermediate parameters , It represents the obliquity of the ecliptic.

[0066] Step 1.9: Calculate the final solar hour angle H (unit: degrees) based on the local solar hour angle (Hour Angle), Greenwich Mean Sidereal Time (GST) (which can be looked up in a table or calculated), and local mean sidereal time (LST) = GST + λ:

[0067] ;

[0068] Where λ represents the longitude of the observation point, and α is the right ascension of the sun, calculated by the following formula:

[0069] ;

[0070] Step 1.10: Calculate the flare angles, as follows:

[0071] Solar altitude angle: ;

[0072] Sun azimuth: ;

[0073] in, The latitude of the observation point.

[0074] Based on the calculated solar altitude angle and solar azimuth angle, and referring to empirical values, the glare area can be selected within ±15° of the solar azimuth angle and solar altitude angle.

[0075] Specifically, step 2 includes:

[0076] A polarizer is fixed to a polarization motor, which is then mounted on a horizontal linear displacement stage. When a glare zone is identified, a displacement signal is transmitted to the stage, and the component carrying the polarization motor is inserted into the imaging optical path for polarization imaging. When the zone is identified as non-glare, the stage is controlled to remove the component carrying the polarization motor from the imaging optical path for non-polarization imaging. Figure 4a , Figure 4b , Figure 4c , Figure 4d , Figure 4e , Figure 4f , Figure 4g , Figure 4h This is an image showing the effect of continuous polarization suppression imaging; among them, Figure 4a , Figure 4c , Figure 4e , Figure 4g This is the original image of sea surface flare from a camera without the polarization rotation module engaged. Figure 4b , Figure 4d , Figure 4f , Figure 4h Image of the sea surface after glare suppression.

[0077] Specifically, step 3 includes:

[0078] The optimal polarization suppression direction self-calibration mechanism is one of the core innovations of this invention. It aims to precisely suppress solar flare over the sea surface, enabling the polarization rotation module to automatically align to the optimal polarization angle (i.e., the angle that minimizes flare) under different observation geometries and flare conditions. The specific implementation steps are as follows:

[0079] Step 3.1: After initially determining that the current observation area has entered the glare zone, control the polarizer to rotate at a constant angular velocity within the range of 0° to 360°.

[0080] The image sequence acquired at the corresponding angle is as follows: ;

[0081] in, Where i is the polarization angle, i is the angle count, and N is the number of angle samples.

[0082] Acquire images from each angle Calculate the average brightness of the image That is, selecting the average brightness within the ROI (target area):

[0083] ;

[0084] in, This represents the number of pixels in the ROI region. For at the polarization angle Below, the brightness of pixel (x,y) in the image, where (x,y) are the pixel coordinates.

[0085] Step 3.2: Find the minimum brightness angle, i.e., the optimal polarization suppression angle. The polarization suppression angle is defined as the rotation angle of the polarizer when sea surface glare interference is suppressed to the greatest extent.

[0086] ;

[0087] in, This represents the independent variable that yields the minimum value of the function.

[0088] At this point, the transmitted component of the glare after passing through the polarizer is the smallest, which means that the angle direction is orthogonal to the polarization direction of the glare, and is the best polarization suppression angle under the current environment.

[0089] Step 3.3: Record the polarization suppression angle - platform attitude:

[0090] Record the optimal polarization suppression angle at this time. Platform attitude (pitch) opt Roll angle opt Yaw opt ).

[0091] Step 3.4: Dynamically update the polarization angle when the platform attitude changes:

[0092] In subsequent observations, instead of re-rotating and scanning, the corresponding polarization angle is calculated in real time based on changes in the platform's attitude. ;

[0093] Since the polarizer can only be controlled by rotating within the roll angle plane, the primary focus is on recording the optimal polarization suppression angle. And the roll angle of the positioning and navigation system at that time opt Using the reference angle, the polarization angle is used as the roll angle changes with the body attitude. Roll needs to be adjusted based on the current roll angle. real-time Dynamic compensation is performed, that is:

[0094] ;

[0095] By controlling the polarization rotation motor to rotate to this angle, real-time dynamic tracking of the polarization suppression angle can be achieved.

[0096] Specifically, step 4 includes:

[0097] Using the SINet-V2 detection algorithm, we trained a model on surface ship samples with polarized glare suppression added to the COD10K target detection dataset, enabling the deep model to learn the ability to detect low-contrast ship targets.

[0098] The following is the specific process of using the SINet-V2 detection algorithm for low-contrast target recognition:

[0099] Step 1: Collect ship samples under typical glare backgrounds. Approximately 240 ship samples were selected from multiple sets of images collected during airborne flight tests under glare conditions on the water surface. The ship targets were manually interpreted; they can be mainly categorized as: fishing boats, surface cages, and large vessels.

[0100] Step 2: Perform polygon semantic segmentation-level annotation on the targets in the samples, and then use the subsequent processing to create training samples that can be used for the SINet-V2 deep model.

[0101] Step 3: Randomly divide all target samples into training and test sets in a 7:3 ratio for model training. After 100 training rounds, select the best-performing model as the prediction model for testing. Some target detection results are shown below. Figures 5a-5f In each image, the top left corner is the original image, the top right corner is the binary image with the identified target, the bottom left corner is the target enhancement and highlighting image, and the bottom right corner is the target labeling image.

[0102] Step 4: Verify the transfer learning capability of the SINet-V2 deep model after introducing the ship target dataset. First, no ship target samples were included in the training dataset for deep model training. The best-performing model was then selected for target detection on the target dataset.

[0103] Step 5: Introduce ship target samples to retrain the deep model, and select the best-performing model to perform target detection on the ship target dataset.

[0104] A comparison of the target detection performance before and after the introduction reveals the following:

[0105] (1) Ship targets are related to ship targets in the COD10K dataset used for SINet-V2 model training. Some ship targets with poor glare interference effects can also be detected by the native SINet-V2 model.

[0106] (2) After introducing ship target samples, SINet-V2 learns the features of ship targets in a targeted manner, and can better detect low-contrast ship targets.

[0107] The present invention also provides a rapid detection device for dynamic targets on the sea surface under polarization variation scenarios, for implementing the above method, comprising the following modules:

[0108] The judgment module obtains latitude and longitude, platform attitude angle and world time through the navigation and positioning system, and combines the solar position calculation model to determine in real time whether the current imaging area falls into the glare zone.

[0109] The drive module immediately drives the displacement stage to insert the polarizer into the optical path when it determines that the light has fallen into the glare zone; when it determines that the light has fallen into the non-glare zone, it drives the displacement stage to move the polarizer out of the optical path.

[0110] The adjustment module, after entering the glare zone, controls the polarization rotation module to complete a 360° polarizer rotation scan at a constant angular velocity, records the imaging brightness at each angle, and obtains the optimal polarization suppression angle by filtering the minimum value; the polarization angle is adjusted and updated in real time according to the platform attitude to obtain a low glare image after adding polarization glare suppression.

[0111] The output classification module performs depth model-based target detection processing on the low-glare image obtained when the polarizer is in the optical path, and outputs the ship target position and category.

[0112] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for rapid detection of dynamic targets on the sea surface under polarization variation scenarios, characterized in that, Includes the following steps: Step 1: Obtain latitude and longitude, platform attitude angle and world time through the navigation and positioning system, and determine in real time whether the current imaging area falls into the glare zone by combining the solar position calculation model; Step 2: When it is determined that the light source has fallen into the glare zone, immediately drive the stepper motor's displacement stage to cut the polarizer into the optical path; when it is determined that the light source has fallen into the non-glare zone, drive the displacement stage to move the polarizer out of the optical path. Step 3: After entering the glare zone, control the polarization rotation module to complete a 360° polarizer rotation scan at a constant angular velocity, record the imaging brightness at each angle, and obtain the optimal polarization suppression angle by filtering the minimum value; adjust and update the polarization angle in real time according to the platform attitude angle to obtain a low glare image after adding polarization glare suppression; the polarization suppression angle is defined as the rotation angle of the polarizer when suppressing sea surface glare interference to the greatest extent. Step 4: Perform depth model-based target detection processing on the low-glare image obtained when the polarizer is in the optical path, and output the ship target position and category.

2. The method for rapid detection of dynamic targets on the sea surface under polarization variation scenarios according to claim 1, characterized in that, In step 1, the solar position calculation model takes world time as input and continuously calculates the solar altitude angle and solar azimuth angle by taking Julian day, century number, solar geometric longitude, mean anomalous angle, central difference angle, solar true longitude, solar apparent longitude, solar declination angle, Greenwich mean sidereal time, local mean sidereal time, and solar hour angle, and delineates the boundary of the flare zone accordingly.

3. The method for rapid detection of dynamic targets on the sea surface under polarization variation scenarios according to claim 1, characterized in that, In step 1, the glare zone is determined based on the longitude, latitude, pitch angle, roll angle, heading angle and UTC time output by the navigation and positioning system. The determination result directly triggers the cut-in or move-out action.

4. The method for rapid detection of dynamic targets on the sea surface under polarization variation scenarios according to claim 1, characterized in that, Step 2 includes constructing a polarization modulation component, which includes a stepper motor, a polarization rotation module, and a polarization rotation module support device. The polarization rotation module support device is fixed on the displacement stage of the stepper motor. The polarization rotation module includes a polarization rotation motor, on which a polarizer is mounted.

5. The method for rapid detection of dynamic targets on the sea surface under polarization variation scenarios according to claim 1, characterized in that, In step 3, during a single 360° polarizer rotation scan, images are acquired synchronously and the average brightness of the entire image or the target area is calculated. The angle corresponding to the minimum brightness value is taken as the optimal polarization suppression angle.

6. The method for rapid detection of dynamic targets on the sea surface under polarization variation scenarios according to claim 1, characterized in that, In step 3, real-time tracking and adjustment includes continuously reading the attitude angle updated by the navigation and positioning system during subsequent imaging, calculating the compensation amount, and driving the polarization rotation motor to complete the angle correction.

7. The method for rapid detection of dynamic targets on the sea surface under polarization variation scenarios according to claim 1, characterized in that, In step 4, the depth model is SINet-V2.

8. A method for rapid detection of dynamic targets on the sea surface under polarization variation scenarios according to claim 7, characterized in that, The training data for the deep model was supplemented with 240 ship samples from the Yaoguang scene based on the COD10K public dataset.

9. A method for rapid detection of dynamic targets on the sea surface under polarization variation scenarios according to claim 8, characterized in that, The ship samples in the Yaoguang scene were manually annotated with polygons and divided into training and testing sets in a 7:3 ratio.

10. A rapid detection device for dynamic targets on the sea surface under polarization variation scenarios, characterized in that, Includes the following modules: The judgment module obtains latitude and longitude, platform attitude angle and world time through the navigation and positioning system, and combines the solar position calculation model to determine in real time whether the current imaging area falls into the glare zone. The drive module immediately drives the displacement stage to insert the polarizer into the optical path when it determines that the light has fallen into the glare zone; when it determines that the light has fallen into the non-glare zone, it drives the displacement stage to move the polarizer out of the optical path. The adjustment module, after entering the glare zone, controls the polarization rotation module to complete a 360° polarizer rotation scan at a constant angular velocity, records the imaging brightness at each angle, and obtains the optimal polarization suppression angle by filtering the minimum value; the polarization angle is adjusted and updated in real time according to the platform attitude to obtain a low glare image after adding polarization glare suppression; the polarization suppression angle is defined as the rotation angle of the polarizer when suppressing sea surface glare interference to the greatest extent. The output classification module performs depth model-based target detection processing on the low-glare image obtained when the polarizer is in the optical path, and outputs the ship target position and category.

Images