An underwater unmanned vehicle

By combining acoustic detection and image analysis, the system achieves precise control of ballast balance and image clarity perception for underwater unmanned vehicles, solving the problem of poor image recognition accuracy in existing technologies and improving the image acquisition quality of underwater vehicles in complex environments.

CN123061561APending Publication Date: 2026-11-10大连星翼金属科技制造有限公司
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202611136488.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-29
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing underwater unmanned vehicles lack the fine perception for ballast balance control and the analysis of image acquisition clarity in complex underwater environments, resulting in poor image recognition accuracy and a lack of effective motion blur compensation mechanisms.

Method used

The system acquires real-time distance sequences using an acoustic detection device, determines image clarity by combining image entropy and exposure rate, uses a balance analysis module to predict ballast balance status and adjust valve opening, a response analysis module to distinguish the causes of image blurring, a fusion processing module to acquire images at stable moments, and a hysteresis processing module to compensate for attitude deviations, thereby achieving multi-frame fusion and hysteresis compensation of images.

Benefits of technology

It improves the accuracy of image recognition for underwater unmanned vehicles in complex underwater environments, and steadily acquires high-definition images through prediction and compensation measures, thereby enhancing the ability to cope with complex disturbances.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123061561A_ABST
    Figure CN123061561A_ABST
Patent Text Reader

Abstract

This invention relates to the field of vehicle technology, and more particularly to an underwater unmanned vehicle (UAV), comprising a detection and control mechanism including several sensors, an image acquisition device, an acoustic detection device, and a valve body; a balance analysis module for determining the ballast balance state of the UAV; a target operation module for determining whether the initial operation image is clear; a response analysis module for determining the cause of unclear initial operation images and processing strategies; a fusion processing module for determining the relatively stable moment and acquiring operation images; and a hysteresis processing module for determining the imbalance initiation moment based on drive correlation, determining the hysteresis rate and attitude deviation by combining the valve opening command issuance moment and the UAV's roll angle, and determining the valve opening adjustment amount based on the hysteresis rate and attitude deviation to reacquire operation images. This invention improves the accuracy of underwater UAV operation image recognition.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of vehicle technology, and more particularly to an underwater unmanned vehicle. Background Technology

[0002] In underwater target search and identification missions, unmanned underwater vehicles (UUVs) are prone to attitude changes such as roll and pitch due to the complex and variable underwater environment. This can lead to motion blur, geometric distortion, or exposure abnormalities in the images captured, severely reducing the usability of the operational images. If adjustment commands are not issued in advance at the appropriate time, the UUV is likely to enter a weak ballast balance state. Continuing to carry out shooting operations in this state will exacerbate image degradation. Existing motion blur compensation methods lack mechanisms for dynamic steady-state detection based on roll angular velocity and for adjusting the acquisition window length according to fusion quality. At the same time, there is a lack of compensation methods that link valve response lag with image acquisition timing, making it impossible to correct imaging failures caused by execution lag in a timely manner. Therefore, there is an urgent need for an unmanned underwater vehicle that can accurately sense the ballast balance state, distinguish the causes of image blurring, and implement multi-frame fusion or lag compensation to stably acquire high-definition operational images of seabed targets under complex underwater disturbances.

[0003] Chinese Patent Application Publication No. CN109591982A discloses an unmanned underwater aquaculture robot, including a vehicle body, a data acquisition module, and a data processing module. The data acquisition module includes a camera, a temperature and salinity sensor, and a five-constant water quality meter. The data processing module includes an image recognition module and a water product growth status extraction module. This invention uses a fusion algorithm of image and water quality parameters to extract the growth status of water products, improving the accuracy of recognition. It avoids the need to catch fish from the water for measurement, reducing harm to the fish, and allows for timely detection, control, and prevention of fish diseases. To a certain extent, it can shorten the fish fry breeding cycle and control fish fry diseases in a timely manner, thereby increasing income.

[0004] However, the existing technology still has the following problems: the existing technology does not take into account the impact of the underwater vehicle's own ballast balance control on the image acquisition clarity, and does not analyze the reasons affecting the image clarity and take corresponding processing strategies accordingly, resulting in poor recognition accuracy of the operation images. Summary of the Invention

[0005] To address this, the present invention provides an underwater unmanned vehicle to overcome the problems in the prior art that fail to consider the impact of the underwater vehicle's own ballast balance control on image acquisition clarity, and fail to analyze the causes affecting image clarity and thus fail to take corresponding processing strategies, resulting in poor recognition accuracy of operational images.

[0006] To achieve the above objectives, the present invention provides an underwater unmanned vehicle, comprising: The detection and control mechanism includes several sensors, an image acquisition device for capturing images of seabed targets, an acoustic detection device for detecting the real-time distance between the vehicle and the seabed, and a valve body for controlling the valve opening. The balance analysis module is used to determine the ballast balance state of the vehicle based on several real-time distances. If the ballast balance state of the vehicle is determined to be a weak ballast balance state, the advance adjustment time of the valve opening is determined according to the distance deviation value. The target operation module is used to determine whether the initial operation image is clear based on the image entropy and exposure rate of the initial operation image of the seabed target; The response analysis module is used to determine the cause and processing strategy of the unclear initial operation image based on the image offset value, and to determine the drive correlation based on the roll moment and roll angular velocity of the vehicle. The fusion processing module is used to determine a relatively stable moment based on the roll angular velocity when the processing strategy is determined to be image fusion, to acquire a number of operational images of the seabed target, to determine whether the image fusion is qualified or to determine the fusion offset value based on the image variance of the number of operational images, and to adjust the step size of the time acquisition window based on the fusion offset value to redetermine the relatively stable moment and reacquire operational images. The hysteresis processing module is used to determine the imbalance start time based on the drive correlation when the processing strategy is determined to be response hysteresis detection, and to determine the hysteresis rate and attitude offset by combining the valve opening command issuance time and the roll angle of the vehicle, respectively. Based on the hysteresis rate and the attitude offset, the adjustment amount of the valve opening is determined to reacquire the operation image.

[0007] Furthermore, if the standard deviation of the distance sequence is greater than the preset standard deviation, the balance analysis module determines that the ballast balance state of the aircraft is a weak ballast balance state. The distance sequence consists of several real-time distance data points continuously acquired within a preset distance acquisition window. The real-time distance is calculated based on the acoustic wave detection device transmitting acoustic pulses from the vessel to the seabed at a preset frequency, receiving the echo signals, and solving the data.

[0008] Furthermore, the balance analysis module determines the advance adjustment time of the valve opening based on the distance deviation value and the reference adjustment time; The distance deviation value is the difference between the standard deviation and the preset standard deviation.

[0009] Furthermore, the target task module determines that the initial task image is unclear when the overall image value is greater than the overall image threshold. The image composite value is the sum of the image entropy factor and the exposure rate factor. The image entropy factor is the ratio of the difference between the image entropy and the reference image entropy; The exposure rate factor is the ratio of the difference between the exposure rate and the reference exposure rate; The image entropy is the information entropy of the initial task image, and the exposure rate is the proportion of saturated pixels to total pixels in the initial task image.

[0010] Furthermore, the response analysis module generates an image offset value based on the image composite value and the image composite threshold to determine the reason why the initial working image is unclear; When the response analysis module determines that the unclear initial image is due to fluctuations in the internal image acquisition device, it determines that the processing strategy is image fusion. When the response analysis module determines that the unclear initial operation image is due to external valve body ballast imbalance, it determines that the processing strategy is response hysteresis detection.

[0011] Furthermore, the fusion processing module acquires a continuous time series of the roll angular velocity based on a preset time acquisition window, and determines the relatively stable moment based on the root mean square of the continuous time series.

[0012] Furthermore, when the image fusion fails, the fusion processing module increases the step size of the time acquisition window based on the fusion offset value; The fusion processing module determines whether the image fusion is qualified based on the image variance and the image variance threshold; The fusion offset value is the difference between the image variance and the image variance threshold.

[0013] Furthermore, when the hysteresis processing module determines that an imbalance event has occurred, it determines the end time of the previous jump window in the adjacent jump window as the start time of the imbalance. The lag processing module determines whether an imbalance event has occurred based on the lag time difference and the correlation difference; The correlation difference is the difference between the driving correlations calculated based on the imbalance time series in adjacent jump windows, and the lag time difference is the difference between the monitoring times corresponding to each driving correlation.

[0014] Furthermore, the hysteresis processing module determines the hysteresis rate based on the imbalance start time, the instruction issuance time, the imbalance roll angle, and the instruction roll angle; The hysteresis processing module determines the attitude offset based on the maximum roll angle within the imbalance fluctuation range and the imbalance roll angle. The imbalance fluctuation range has an upper limit at the time of the imbalance onset and a lower limit at the time of the instruction issuance.

[0015] Furthermore, the hysteresis processing module determines the adjustment amount based on the attitude offset and the proportional coefficient; The proportionality coefficient is positively correlated with the hysteresis rate.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention continuously acquires the real-time distance sequence between the vehicle and the seabed through the acoustic detection device, and uses whether the standard deviation of the distance sequence exceeds the preset standard deviation as the basis for judging the weak ballast balance state. It can sensitively capture early signs of increased depth fluctuations. When the weak ballast balance state is determined, the valve opening adjustment time is determined based on the magnitude of the distance deviation value, so that the ballast control is preventive and provides a more stable carrier attitude basis for subsequent shooting, thereby further improving the accuracy of underwater unmanned vehicle operation image recognition.

[0017] Furthermore, this invention uses a comprehensive image value, which is the sum of an image entropy factor and an exposure rate factor, to determine whether the initial operational image is clear. The image entropy factor reflects the richness of information, while the exposure rate factor reflects the tendency of overexposure or underexposure. The comprehensive image value characterizes the unclear image acquisition caused by uneven underwater lighting and scattering by suspended objects, providing a reliable basis for whether to initiate correction processing, thereby further improving the accuracy of underwater unmanned vehicle operational image recognition.

[0018] Furthermore, this invention, through the image offset value generated by the image synthesis value and the image synthesis threshold, combined with the analysis of the vehicle's roll moment and angular velocity, can accurately distinguish the reasons for unclear initial operational images. These reasons include fluctuations in the internal image acquisition device, i.e., motion blur caused by slight vibrations or fluctuations of the internal imaging device itself at the moment of image acquisition; and external valve body ballast imbalance, i.e., overall attitude deviation caused by ballast imbalance due to valve response lag and failure to achieve expected attitude recovery. Corresponding processing strategies are adopted according to different reasons. An image fusion processing strategy is used for internal fluctuations, and a response lag detection control compensation strategy is used for external imbalances. This improves the pertinence and processing efficiency in dealing with complex disturbances, thereby further improving the accuracy of underwater unmanned vehicle operational image recognition.

[0019] Furthermore, under the image fusion processing strategy, this invention determines a relatively stable moment within the time acquisition window by using the root mean square of the roll angular velocity and acquires multiple frames of operational images. Then, the difference between the image variance and the image variance threshold is used as the fusion offset value to determine whether the image fusion is qualified. If it is unqualified, the step size of the time acquisition window is increased according to the fusion offset value and the stable moment is re-determined, forming an optimization oriented towards the stable window, which is beneficial for the stable extraction of clear fused images in attitude fluctuations. Under the response lag detection processing strategy, the imbalance event and the imbalance start time are identified by driving the correlation change. The lag rate and attitude offset are calculated by combining the command issuance time and the roll angle. The valve opening adjustment amount is determined by a proportional coefficient positively correlated with the lag rate. Based on this compensation, the operational images are re-acquired, which effectively reduces the impact of valve response delay on attitude recovery and imaging quality, thereby further improving the accuracy of underwater unmanned vehicle operational image recognition. Attached Figure Description

[0020] Figure 1 This is a block diagram of the underwater unmanned vehicle according to an embodiment of the present invention; Figure 2 This is a logic diagram for determining the advance adjustment time for stopping operations or determining the valve opening in an embodiment of the present invention. Figure 3 This is a logic diagram for determining whether the initial working image is clear in an embodiment of the present invention; Figure 4 This is a logic diagram for determining whether image fusion is qualified according to an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0022] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0023] It should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0024] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation" and "connection" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0025] Please see Figure 1 The diagram shown is a block diagram of an underwater unmanned vehicle according to an embodiment of the present invention. The underwater unmanned vehicle according to an embodiment of the present invention includes: The detection and control mechanism includes several sensors, an image acquisition device for capturing images of seabed targets, an acoustic detection device for detecting the real-time distance between the vehicle and the seabed, and a valve body for controlling the valve opening. The balance analysis module, which is connected to the detection and control mechanism, is used to determine the ballast balance state of the vehicle based on several real-time distances. If the ballast balance state of the vehicle is determined to be a weak ballast balance state, the advance adjustment time of the valve opening is determined according to the distance deviation value. The target operation module, which is connected to the detection and control mechanism and the balance analysis module, is used to determine whether the initial operation image is clear based on the image entropy and exposure rate of the initial operation image of the seabed target. The response analysis module, which is connected to the detection control mechanism and the target operation module, is used to determine the cause of unclear initial operation images and processing strategies based on image offset values, and to determine the drive correlation based on the roll moment and roll angular velocity of the aircraft. The fusion processing module, which is connected to the detection control mechanism and the response analysis module, is used to determine the relative stable moment based on the roll angular velocity when the processing strategy is determined to be image fusion, to acquire several operational images of the seabed target, to determine the image fusion qualification or to determine the fusion offset value based on the image variance of several operational images, and to adjust the step size of the time acquisition window based on the fusion offset value to redetermine the relative stable moment and reacquire operational images. The hysteresis processing module, which is connected to the detection control mechanism and the response analysis module, is used to determine the imbalance start time based on the drive correlation when the processing strategy is determined to be response hysteresis detection. It determines the hysteresis rate and attitude offset by combining the valve opening command issuance time and the roll angle of the aircraft. Based on the hysteresis rate and attitude offset, it determines the valve opening adjustment amount to reacquire the operation image.

[0026] It is understood that the underwater unmanned vehicle provided in this embodiment of the invention can be applied to tasks such as nearshore subsea pipeline inspection, seabed topography mapping, or underwater target search. The vehicle needs to conduct fixed-altitude patrols within a height range of 3 to 5 meters above the seabed and hover to capture high-definition operational images when suspicious pipeline displacement, abnormal topography, or man-made targets are detected. The operating area has a water depth of 50 meters, and the seabed is mainly composed of sand and mud with natural and slow undulations. The vehicle is easily affected by the coupling of shallow water swells, bottom currents, and its own propulsion pulsations, which can easily generate continuous depth fluctuations and rolling motions. At the same time, the opening and closing response of its ballast tank valves has a physical lag, which makes it impossible to complete attitude recovery instantaneously. The above-mentioned disturbance factors make the vehicle prone to entering a weak ballast equilibrium state with depth divergence but not yet exceeding the limit, resulting in motion blur or attitude shift blur in the captured images.

[0027] Specifically, in order to stably acquire high-definition operational images of seabed targets under the above conditions, the vehicle in this embodiment of the invention is equipped with: an acoustic altimeter as an acoustic wave detection device, a high-resolution underwater camera as an image acquisition device; an inertial measurement unit for measuring attitude information such as the roll rate, roll moment and roll angle of the vehicle; and a valve body for adjusting the inflow and outflow of ballast water tanks.

[0028] Specifically, the balance analysis module determines that the ballast balance state of the aircraft is a weak ballast balance state when the standard deviation of the distance sequence is greater than the preset standard deviation. If the standard deviation of the distance sequence is less than or equal to the preset standard deviation, the balance analysis module determines that the ballast balance state of the aircraft is a strong ballast balance state. The distance sequence consists of several real-time distance data points continuously acquired within a preset distance acquisition window. The real-time distance is calculated based on the acoustic pulse emitted from the vehicle to the seabed at a preset frequency by the acoustic wave detection device, the reception of the echo signal, and the calculation.

[0029] In this embodiment, the balance analysis module includes a preset standard deviation determination unit, which is used to distinguish between strong ballast balance state and weak ballast balance state. The smaller the preset standard deviation, the more stable the vehicle is.

[0030] Specifically, the preset standard deviation is the product of the inherent fluctuation standard deviation and the preset redundancy coefficient. After confirming that the ocean current is stable, the vehicle's ballast is balanced, and it is in a stable hovering state, the real-time distance between the vehicle and the seabed is continuously collected at a preset frequency using an acoustic detection device to form a long-period distance reference sequence. The distance reference sequence is statistically analyzed, and the standard deviation of the distance reference sequence is calculated as the inherent fluctuation standard deviation. The inherent fluctuation standard deviation is the normal depth fluctuation level of the vehicle under the current operating conditions caused by factors such as thruster pulsation and microcurrent disturbance. The redundancy coefficient is a fixed value, and its function is to provide a reasonable tolerance bandwidth for the inherent fluctuation, so that the preset standard deviation can fully cover the normal hovering fluctuation and avoid false alarms, and can also ensure that when the depth fluctuation increases abnormally, it can be identified in time. The value of the redundancy coefficient follows the following calibration principle: In the controlled test, a weak external disturbance is artificially applied so that the depth fluctuation of the vehicle just exceeds the normal range. At this time, the calculated distance sequence standard deviation should just exceed the preset standard deviation.

[0031] In this embodiment, the balance analysis module includes a distance acquisition window determination unit, which is used to determine that the length of the distance acquisition window is an integer multiple of the main period of depth fluctuation, to acquire depth fluctuation data of the vehicle under strong ballast balance under typical operating sea conditions, and to extract the main period of its depth fluctuation through spectrum analysis. The main period can determine the dominant frequency characteristics of the depth fluctuation of the vehicle during normal hovering. When setting an integer multiple, the length of the distance acquisition window must not exceed the maximum allowable response time determined by the current operating altitude of the vehicle and the bottoming safety boundary. The maximum allowable response time is defined as the longest time that the system can use to complete the judgment and perform preventive control from the time the weakening trend of ballast condition is detected until the depth deviation of the vehicle develops to the point of touching the safety boundary.

[0032] Specifically, real-time distance refers to the instantaneous vertical straight-line distance from the transducer at the bottom of the vehicle to the seabed reflector, measured and calculated by the acoustic wave detection device at the current moment. In this embodiment, an acoustic altimeter is used as the acoustic wave detection device, which emits a short acoustic pulse towards the seabed at a preset frequency. The receiver receives the echo signal reflected back from the seabed and accurately records the round-trip propagation time of the acoustic pulse from transmission to reception. The sound velocity of the water in the current sea area is measured in real time by a sound velocity meter. The real-time distance from the vehicle to the seabed is determined to be equal to the ratio of the product of the water sound velocity and the round-trip propagation time to 2. To improve data quality, the original calculation results can be filtered by median filtering to remove obvious abnormal jump points. The filtered real-time distance is used as the final data point for calculation.

[0033] In this embodiment, the balance analysis module includes a preset frequency determination unit, which is used to determine the number of real-time distance data points acquired per unit time based on the preset frequency.

[0034] Specifically, the depth fluctuation data sequence of the vehicle under strong ballast balance in typical operating sea conditions is acquired. Spectral analysis is performed on the depth fluctuation data sequence to extract the highest significant frequency component of the depth fluctuation. The highest significant frequency component refers to the frequency value in the power spectral density, excluding the zero-frequency component, where the power proportion exceeds a preset threshold. It characterizes the fastest changing component of depth fluctuation caused by factors such as propeller pulsation and microfluidic disturbances when the vehicle is hovering normally. According to the Nyquist sampling theorem, to fully capture the waveform characteristics of the highest significant frequency component, the preset frequency needs to be set to at least twice the highest significant frequency component. To improve the reconstruction accuracy of the sampled signal and the fidelity of the distance sequence in subsequent statistical analysis, the multiple can be rounded up to a higher integer multiple that is easily achievable in engineering. At the same time, the upper limit of the preset frequency is constrained by the hardware performance of the acoustic detection device. It must be ensured that the maximum detection distance corresponding to a single transmission and reception cycle at the corresponding frequency is greater than the maximum operating altitude of the vehicle, so as to ensure that the seabed echo signal can be fully received after each acoustic pulse transmission without distance ambiguity.

[0035] Specifically, this embodiment continuously acquires the real-time distance sequence between the vehicle and the seabed through an acoustic detection device, and uses whether the standard deviation of the distance sequence exceeds a preset standard deviation as the basis for determining a weak ballast balance state. This can sensitively capture early signs of increased depth fluctuations. When a weak ballast balance state is determined, the valve opening adjustment time is determined based on the magnitude of the distance deviation value, making ballast control preventive and providing a more stable carrier attitude basis for subsequent shooting, thereby further improving the accuracy of underwater unmanned vehicle operation image recognition.

[0036] Please see Figure 2 As shown, it is a logic decision diagram for determining the advance adjustment time of stopping operation or determining the valve opening in an embodiment of the present invention. The balance analysis module determines the advance adjustment time of the valve opening based on the distance deviation value and the reference adjustment time. The distance deviation value is the difference between the standard deviation and the preset standard deviation.

[0037] In this embodiment, the balance analysis module includes a reference adjustment time determination unit, which is used to determine the reference adjustment time as the basis for calculating the valve opening advance adjustment time, and to determine the time difference required for the system to intervene in advance under the standard deviation.

[0038] In this embodiment, the value of the reference adjustment time mainly depends on the response characteristics of the valve actuator and the flow characteristics of the ballast water tank. In order to balance the timeliness and stability of the adjustment, the preferred value in this embodiment is 4 seconds.

[0039] Specifically, the balance analysis module also includes an advance adjustment time determination unit, which is used to determine the advance adjustment time as the product of the distance deviation value and the reference adjustment time. The adjustment direction is determined by the sign of the distance deviation and the current average depth of the vehicle.

[0040] If the mean of the distance sequence is consistently greater than the target altitude, it indicates that the vehicle is too high and the ballast water volume needs to be increased to make the vehicle sink and restore the target altitude. If the mean of the distance sequence is consistently less than the target altitude, it indicates that the vehicle is too low and the ballast water volume needs to be reduced to make the vehicle float and restore the target altitude.

[0041] Please see Figure 3 As shown, it is a logic determination diagram for determining whether the initial task image is clear in an embodiment of the present invention. When the image comprehensive value is greater than the image comprehensive threshold, the target task module determines that the initial task image is not clear. The target task module determines that the initial task image is clear when the image synthesis value is less than or equal to the image synthesis threshold; The overall image value is the sum of the image entropy factor and the exposure rate factor. The image entropy factor is the ratio of the difference between the image entropy and the reference image entropy; The exposure rate factor is the ratio of the difference between the exposure rate and the baseline exposure rate; Image entropy is the information entropy of the initial job image, and exposure rate is the proportion of saturated pixels in the initial job image to the total number of pixels.

[0042] In this embodiment, the target operation module includes an image comprehensive threshold determination unit, which is used to determine the image comprehensive threshold to distinguish between usable images and images with degradation defects that need to be corrected. The value of the image comprehensive threshold is determined based on the readability requirements of target recognition for medium-scale detail features such as weld seams and corrosion points of submarine pipelines, as well as the degree of characterization of image degradation by composite indicators. The setting of the image comprehensive threshold should ensure a balance between retaining usable images and filtering defective images.

[0043] In this embodiment, the target task module includes an image entropy determination unit, which is used to characterize the richness of information contained in the initial task image. Clear images usually have sharp edges and rich textures, and have high information entropy; blurry images lose details, the grayscale distribution tends to be concentrated, and the information entropy decreases. During calculation, the initial task image is first converted into a grayscale image, and the brightness value of each pixel is between 0 and 255. The grayscale image is traversed, and the number of pixels corresponding to each brightness value is counted one by one. The number of pixels for each brightness value is divided by the total number of pixels in the image to obtain the probability of the corresponding brightness value. The probability of each brightness value is multiplied by the base-2 logarithm of the probability, and the sum of the products of all brightness values ​​is taken as the negative number to obtain the information entropy of the image.

[0044] In this embodiment, the target operation module includes an exposure rate determination unit, which is used to determine the exposure rate as the proportion of saturated pixels to the total number of pixels in the initial operation image. The higher the exposure rate, the more areas of information loss in the image and the worse the image quality. In the calculation, pixels with a brightness value of 255 are first identified as overexposed pixels, and pixels with a brightness value of 0 are identified as underexposed pixels. In this embodiment, saturated pixels include overexposed pixels and underexposed pixels. All pixels in the initial operation image are traversed, and the number of saturated pixels is counted, which is the sum of the number of overexposed pixels and the number of underexposed pixels. The counted number of saturated pixels is divided by the total number of pixels in the image to obtain the exposure rate of the current image.

[0045] Specifically, the baseline image entropy and baseline exposure rate are both reference benchmarks for ideal, clear images and need to be calibrated in advance. The specific process is to calibrate in a sandy or muddy environment with clear water, good lighting conditions, and a seabed type similar to the operating area. The vehicle is hovered at a standard operating height of 4 meters above the seabed. Under the condition that the balance analysis module determines that the current state is one of strong ballast balance and the root mean square value of the roll angular velocity is lower than the preset stability judgment threshold, several frames of images are taken of the standard target board with a checkerboard pattern. According to the above calculation method for image entropy and exposure rate, the image entropy and exposure rate of multiple frames are calculated respectively, and the average value is taken as the baseline image entropy and baseline exposure rate and stored in the memory of the target operation module.

[0046] Specifically, the image entropy factor is the ratio of the absolute value of the difference between the image entropy and the reference image entropy to the reference image entropy, and the exposure rate factor is the ratio of the absolute value of the difference between the exposure rate and the reference exposure rate to the reference exposure rate.

[0047] Specifically, this embodiment uses the combined image value, which is composed of the image entropy factor and the exposure rate factor, to determine whether the initial operation image is clear. The image entropy factor reflects the richness of information, while the exposure rate factor reflects the tendency of overexposure or underexposure. The combined image value characterizes the unclear image acquisition caused by uneven underwater lighting and scattering of suspended objects, providing a reliable basis for whether to initiate correction processing, thereby further improving the accuracy of underwater unmanned vehicle operation image recognition.

[0048] Specifically, the response analysis module generates an image offset value based on the image composite value and the image composite threshold to determine the reason for the unclear initial working image; When the response analysis module determines that the unclear initial image is due to fluctuations in the internal image acquisition device, it determines that the processing strategy is image fusion. When the response analysis module determines that the unclear initial operation image is due to external valve body ballast imbalance, the processing strategy is response hysteresis detection.

[0049] In this embodiment, the image offset value is the difference between the image composite value and the image composite threshold.

[0050] Specifically, this embodiment uses the image offset value generated by the image synthesis value and the image synthesis threshold, combined with the analysis of the vehicle's roll moment and angular velocity, to accurately distinguish the reasons for unclear initial operation images. These reasons include fluctuations in the internal image acquisition device, i.e., motion blur caused by slight vibrations or fluctuations of the internal imaging device itself at the moment of image acquisition; and external valve body ballast imbalance, i.e., overall attitude deviation caused by ballast imbalance due to valve response lag and failure to achieve expected attitude recovery. Corresponding processing strategies are adopted according to different reasons. An image fusion processing strategy is used for internal fluctuations, and a response lag detection control compensation strategy is used for external imbalances. This improves the pertinence and processing efficiency in dealing with complex disturbances, thereby further improving the accuracy of underwater unmanned vehicle operation image recognition.

[0051] Specifically, the fusion processing module acquires a continuous time series of roll angular velocity based on a preset time acquisition window, and determines the relatively stable moment based on the root mean square of the continuous time series.

[0052] In this embodiment, the fusion processing module includes a time acquisition window determination unit, which is used to determine the length of the time acquisition window based on the main period of the roll motion of the vehicle. The main period of the roll motion is obtained by spectral analysis of the roll angular velocity time series of the vehicle under typical operating sea conditions. The length of the time acquisition window is set to an integer multiple of the main period of the roll motion, and the integer multiple ensures that the length of the time acquisition window covers at least two complete roll cycles. This ensures that the roll motion data contained in the time acquisition window can fully reflect the fluctuation characteristics of the vehicle in multiple oscillation cycles, so that the root mean square of the roll angular velocity can effectively measure the comprehensive fluctuation level in the corresponding time period and avoid the distortion of stability assessment caused by period truncation.

[0053] In this embodiment, the fusion processing module includes an inertial measurement unit, which outputs roll angular velocity data in real time at a fixed sampling frequency of 100Hz. Starting from the current decision moment, the fusion processing module extracts a continuous time series of roll angular velocity with a preset time acquisition window length from the data buffer of the inertial measurement unit, with a sampling interval of 0.01 seconds.

[0054] In this embodiment, the fusion processing module includes a root mean square (RMS) determination unit for continuous time series, which is used to characterize the overall intensity of the roll motion of the aircraft within the time acquisition window. The smaller the RMS value, the lower the overall amplitude of the roll angular velocity within the time acquisition window and the more stable the attitude. The RMS value of the 1000 roll angular velocity data points intercepted by the inertial measurement unit is calculated according to the RMS calculation formula of discrete sampling signals, which is the RMS value of the continuous time series.

[0055] In this embodiment, the fusion processing module includes a relatively stable moment determination unit, which slides the time acquisition window backward step by step in increments of 0.1 seconds, traversing all historical data of roll angular velocity. Each time the window is slid, the root mean square value of the current time acquisition window is recalculated. The minimum value is found from all the calculated root mean square values, and the midpoint of the time acquisition window corresponding to the minimum value is determined as the relatively stable moment.

[0056] Please see Figure 4 As shown, it is a logic judgment diagram for determining whether image fusion is qualified in an embodiment of the present invention. When the image fusion is unqualified, the fusion processing module increases the step size of the time acquisition window based on the fusion offset value. The fusion processing module determines that the image fusion is qualified when the image variance is less than or equal to the image variance threshold; The fusion processing module determines that the image fusion is unqualified when the image variance is greater than the image variance threshold; The fusion offset value is the difference between the image variance and the image variance threshold.

[0057] In this embodiment, the fusion processing module includes an image variance determination unit, which is used to determine the image variance by generating an inter-frame difference image after the two acquired operation images are generated. The gray-scale variance value of the inter-frame difference image is calculated to characterize the image fusion effect. When the image variance is less than or equal to the image variance threshold, it indicates that the attitude of the vehicle is more stable at the moment of acquisition, the consistency between the source images is higher, and the fusion quality is better.

[0058] In this embodiment, the fusion processing module includes an image variance threshold determination unit, which is used to determine the image variance threshold. Specifically, under controlled experimental conditions, the aircraft performs multiple sets of image acquisition and fusion under different levels of roll disturbance. The variance value of the inter-frame difference image is calculated for each set of results, and the image representation value of the corresponding fused image is obtained according to the image representation value calculation method in the target operation module. A corresponding curve between image variance and image representation value is established through regression analysis. The corresponding curve shows that the larger the image variance, the more significant the inter-frame difference, the more severe the loss of texture details in the fused image, and the higher the corresponding image representation value, i.e., the worse the image quality. The image representation threshold set in the target operation module is used as the quantification boundary of the sharpness requirement. The image variance value corresponding to the critical condition that the image representation value is exactly equal to the image representation threshold is calibrated as the image variance threshold.

[0059] In this embodiment, the fusion processing module includes a step size adjustment unit for the time acquisition window. When the image fusion is deemed unqualified, the unit determines the degree of unqualification based on the fusion offset value and increases the step size of the time acquisition window to re-determine the relatively stable moment. For example, when the fusion offset value is within the range of (0, 200), the step size of the time acquisition window is increased by 1 second; when the fusion offset value is within the range of (200, 400), the step size of the time acquisition window is increased by 2 seconds; and when the fusion offset value is greater than 400, the step size of the time acquisition window is increased by 3 seconds.

[0060] Specifically, when the lag processing module determines that an imbalance event has occurred, it determines the end time of the previous jump window in the adjacent jump window as the start time of the imbalance. When the lag time difference is greater than the lag time difference threshold or the correlation difference is greater than the correlation difference threshold, the lag processing module indicates that the response delay of the valve between adjacent jump windows has been significantly prolonged, or that the degree to which the aircraft's roll motion is driven by the valve action has changed beyond the normal range, thus determining that an imbalance event has occurred. When the lag time difference is less than or equal to the lag time difference threshold and the correlation difference is less than or equal to the correlation difference threshold, the lag processing module indicates that the valve response between adjacent jump windows is normal and the degree of the aircraft's roll motion driven by the valve action has not changed beyond the normal range, thus confirming that no imbalance event has occurred. Among them, the correlation difference is the difference between the driving correlations calculated based on the imbalance time series in adjacent jump windows, and the lag time difference is the difference between the monitoring times corresponding to each driving correlation.

[0061] Specifically, the correlation difference is the absolute value of the difference between the driving correlations of two adjacent transition windows, and the lag time difference is the absolute value of the difference between the monitoring times of two adjacent transition windows.

[0062] In this embodiment, the hysteresis processing module includes a jump window determination unit, which is used to determine the jump window as a sliding time window for monitoring changes in the vehicle's drive correlation. That is, by comparing the changes in drive correlation between two adjacent jump windows and the corresponding time interval, it can identify whether there is an imbalance event. In the near-shore scenario of this embodiment, the vehicle's roll period is 4 to 6 seconds, and the length of the jump window is set to 2 to 3 times the vehicle's roll period. Therefore, the length of a single jump window is set to 12 seconds. In order to ensure sufficient time resolution and provide sufficient data overlap for continuous monitoring, the sliding step size of adjacent jump windows is set to one-third of the length of a single jump window, that is, 4 seconds.

[0063] Specifically, when an imbalance event is determined to occur, the end time of the previous window in the adjacent transition window is taken as the start time of the imbalance.

[0064] In this embodiment, the hysteresis processing module includes an imbalance start time determination unit. The process of determining the imbalance start time is as follows: taking the moment when the response hysteresis detection is triggered as the endpoint, 60 seconds of historical data are extracted backward, and several jump windows are sequentially divided according to the jump window length. Each jump window is marked by its start time. Two adjacent jump windows form a group of adjacent jump windows. The end time of the previous jump window is the start time of the next jump window. For example, if window A corresponds to the time interval 0-12 seconds and window B corresponds to the time interval 4-16 seconds, then window A and window B constitute a group of adjacent jump windows. If an imbalance event is determined to occur in the adjacent jump window, then the end time of window A, 12 seconds, is the imbalance start time.

[0065] In this embodiment, the hysteresis processing module includes a drive correlation determination unit, which is used to characterize the correlation strength between the roll motion of the aircraft and the ballast adjustment command. The higher the drive correlation value, the greater the degree to which the roll is driven by the valve action. Within a single jump window, the roll angular velocity time series of the aircraft and the valve opening command time series are used as the basic data. The roll angular velocity time series and the valve opening command time series are time-aligned, and the sampling frequency is uniformly set to 100Hz. Normalized cross-correlation analysis is performed on the roll angular velocity series and the valve opening command series within the jump window. The valve opening command series is used as the input signal, and the roll angular velocity series is used as the output response signal. The maximum value of the cross-correlation coefficient between the two is calculated, which is the drive correlation of the current jump window. At the same time, the monitoring time corresponding to the drive correlation is recorded.

[0066] In this embodiment, the hysteresis processing module includes a hysteresis time difference threshold determination unit, which is used to determine the significance of the change in the valve's roll response delay time in adjacent jump windows based on the hysteresis time difference value. The value of the hysteresis time difference threshold should be able to effectively capture the hysteresis trend of the valve response, while ensuring that it will not be falsely triggered due to normal fluctuations. The hysteresis time difference threshold is set as the mean of the monitoring time sample set plus a preset integer multiple of the standard deviation of the monitoring time sample set.

[0067] Specifically, during the system integration testing phase, the aircraft is placed in a state of strong ballast balance. Standard valve opening adjustment commands are manually issued through the balance analysis module. The amplitude and rate of change of the commands should cover the typical range that may be used in the mission. After each command is issued, the time series of valve opening commands and the time series of the aircraft's roll rate response are continuously recorded. For the response process of each standard command, the same method used for imbalance event detection is used to divide the transition windows and calculate the drive correlation and corresponding monitoring time of each transition window, that is, the time delay between the valve command and the roll response. Since the aircraft is in a fault-free normal operating condition at this time, the drive correlation and its monitoring time should remain relatively stable. The monitoring times corresponding to all transition windows are extracted to form a monitoring time sample set for normal operating conditions. The selection principle of the preset integer multiple is that within the normal operating condition sample set, the difference in monitoring time between any adjacent transition windows is not greater than the lag time difference threshold.

[0068] In this embodiment, the hysteresis processing module includes a correlation difference threshold determination unit, which is used to determine the significance of the abrupt change in the degree of roll motion driven by valve action in adjacent jump windows based on the correlation difference value. The correlation difference threshold value should be able to promptly detect the correlation difference caused by the abrupt change in roll response due to valve hysteresis, while filtering out random small-amplitude correlation fluctuations caused by ocean current disturbances. Using the drive correlation data of each jump window under normal operating conditions obtained during the above-mentioned hysteresis time difference threshold calibration process, a drive correlation sample set under normal operating conditions is constructed. The drive correlation difference value of all adjacent jump window pairs under normal operating conditions is calculated successively to construct a correlation difference sample. This episode obtains the maximum value among them. The correlation difference threshold is set as a preset multiple of the maximum value of the correlation difference sample set. The preset multiple is slightly greater than one to leave a small tolerance margin for normal fluctuations. Any correlation difference exceeding the correlation difference threshold indicates that the degree of drive of the yaw motion between adjacent windows by the valve action has changed beyond the normal range. The change cannot be caused by continuous and stable ocean current disturbances, because ocean current disturbances are random and continuously changing and will not produce a step-like jump in drive correlation between adjacent windows. Therefore, the correlation difference threshold can effectively capture the transient changes caused by valve response lag, while filtering out random small fluctuations caused by ocean current disturbances.

[0069] Specifically, the lag processing module determines the lag rate based on the imbalance start time, the command issuance time, the imbalance roll angle, and the command roll angle; The hysteresis processing module determines the attitude offset based on the maximum roll angle and the unbalanced roll angle within the unbalanced fluctuation range; The imbalance fluctuation range is defined by the time the imbalance begins as the upper limit and the time the instruction is issued as the lower limit.

[0070] Specifically, the unbalanced roll angle is the aircraft roll angle corresponding to the moment the unbalance begins, and the command roll angle is the aircraft roll angle corresponding to the moment the command is issued.

[0071] Specifically, the attitude deviation is the absolute value of the difference between the maximum roll angle within the imbalance fluctuation range and the imbalance roll angle, used to characterize the maximum additional attitude deviation caused by the hysteresis effect.

[0072] In this embodiment, the hysteresis processing module includes a hysteresis rate determination unit, which is used to characterize the rate of change of the aircraft's roll attitude response delay after the valve command is issued. The hysteresis rate is equal to the ratio of the absolute value of the difference between the unbalanced roll angle and the commanded roll angle to the difference between the unbalanced start time and the command issuance time.

[0073] In this embodiment, the hysteresis processing module includes a maximum roll angle determination unit, which is used to determine the maximum roll angle as the maximum amplitude of the aircraft's roll motion within the imbalance fluctuation range. It extracts all roll angle data sequences within the imbalance fluctuation range from the inertial measurement unit, traverses the sequence, and records the maximum absolute roll angle, which is the maximum roll angle.

[0074] Understandably, the imbalance fluctuation range has a lower limit at the time the command is issued and an upper limit at the time the imbalance begins. The imbalance fluctuation range fully covers the entire development process from the issuance of the ballast adjustment command to the system confirming that the valve response lag has caused significant imbalance.

[0075] Specifically, the hysteresis processing module determines the adjustment amount based on the attitude offset and the proportional coefficient; Among them, the proportional coefficient is positively correlated with the lag rate.

[0076] In this embodiment, the hysteresis processing module includes a proportional coefficient determination unit, which is used to determine the proportional coefficient as a conversion factor that maps the attitude offset to the valve opening adjustment. The larger the hysteresis rate, the larger the value of the proportional coefficient, and the larger the adjustment amount. If the hysteresis rate is in the range of [0° / s, 0.5° / s], the proportional coefficient is determined to be 0.5% / °. If the hysteresis rate is in the range of (0.5° / s, 1.5° / s], the proportional coefficient is determined to be 1.5% / °. If the hysteresis rate is in the range of (1.5° / s, 3.0° / s], the proportional coefficient is determined to be 3.0% / °. If the hysteresis rate is greater than 3.0° / s, the proportional coefficient is determined to be 5.0% / °.

[0077] In this embodiment, the hysteresis processing module includes an adjustment amount determination unit, which determines the adjustment amount as the product of a proportional coefficient and an attitude offset, to characterize the amount of valve opening change required to eliminate the hysteresis effect. The unit of adjustment amount is %, and the adjustment direction is determined by the sign of the unbalanced roll angle and the roll direction of the aircraft. If the aircraft is in the positive direction within the unbalanced fluctuation range, such as a downward tilt to the starboard side causing an attitude offset, the valve opening should be adjusted in the direction to offset the offset, that is, the valve opening of the corresponding ballast water tank should be increased or decreased in the opposite direction.

[0078] Specifically, in this embodiment, under the image fusion processing strategy, the root mean square of the roll angular velocity is used to determine a relatively stable moment within the time acquisition window, and multiple frames of operational images are acquired. The difference between the image variance and the image variance threshold is used as the fusion offset value to determine whether the image fusion is qualified. If it is unqualified, the step size of the time acquisition window is increased according to the fusion offset value, and the stable moment is re-determined, forming an optimization oriented towards the stable window, which is beneficial for the stable extraction of clear fused images in attitude fluctuations. Under the response lag detection processing strategy, the imbalance event and the imbalance start time are identified by driving the correlation change. The lag rate and attitude offset are calculated by combining the command issuance time and the roll angle. The valve opening adjustment amount is determined by a proportional coefficient positively correlated with the lag rate. After compensation, the operational images are re-acquired, which effectively reduces the impact of valve response delay on attitude recovery and imaging quality, thereby further improving the accuracy of underwater unmanned vehicle operational image recognition.

[0079] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. An underwater unmanned vehicle, characterized in that, include: The detection and control mechanism includes several sensors, an image acquisition device for capturing images of seabed targets, an acoustic detection device for detecting the real-time distance between the vehicle and the seabed, and a valve body for controlling the valve opening. The balance analysis module is used to determine the ballast balance state of the vehicle based on several real-time distances. If the ballast balance state of the vehicle is determined to be a weak ballast balance state, the advance adjustment time of the valve opening is determined according to the distance deviation value. The target operation module is used to determine whether the initial operation image is clear based on the image entropy and exposure rate of the initial operation image of the seabed target; The response analysis module is used to determine the cause and processing strategy of the unclear initial operation image based on the image offset value, and to determine the drive correlation based on the roll moment and roll angular velocity of the vehicle. The fusion processing module is used to determine a relatively stable moment based on the roll angular velocity when the processing strategy is determined to be image fusion, to acquire a number of operational images of the seabed target, to determine whether the image fusion is qualified or to determine the fusion offset value based on the image variance of the number of operational images, and to adjust the step size of the time acquisition window based on the fusion offset value to redetermine the relatively stable moment and reacquire operational images. The hysteresis processing module is used to determine the imbalance start time based on the drive correlation when the processing strategy is determined to be response hysteresis detection, and to determine the hysteresis rate and attitude offset by combining the valve opening command issuance time and the roll angle of the vehicle, respectively. Based on the hysteresis rate and the attitude offset, the adjustment amount of the valve opening is determined to reacquire the operation image.

2. The underwater unmanned vehicle according to claim 1, characterized in that, The balance analysis module determines that the ballast balance state of the aircraft is a weak ballast balance state when the standard deviation of the distance sequence is greater than the preset standard deviation. The distance sequence consists of several real-time distance data points continuously acquired within a preset distance acquisition window. The real-time distance is calculated based on the acoustic wave detection device transmitting acoustic pulses from the vessel to the seabed at a preset frequency, receiving the echo signals, and solving the data.

3. The underwater unmanned vehicle according to claim 2, characterized in that, The balance analysis module determines the advance adjustment time of the valve opening based on the distance deviation value and the reference adjustment time; The distance deviation value is the difference between the standard deviation and the preset standard deviation.

4. The underwater unmanned vehicle according to claim 3, characterized in that, When the image composite value is greater than the image composite threshold, the target task module determines that the initial task image is unclear. The image composite value is the sum of the image entropy factor and the exposure rate factor. The image entropy factor is the ratio of the difference between the image entropy and the reference image entropy; The exposure rate factor is the ratio of the difference between the exposure rate and the reference exposure rate; The image entropy is the information entropy of the initial task image, and the exposure rate is the proportion of saturated pixels to total pixels in the initial task image.

5. The underwater unmanned vehicle according to claim 4, characterized in that, The response analysis module generates an image offset value based on the image composite value and the image composite threshold to determine the reason why the initial operation image is unclear. When the response analysis module determines that the unclear initial image is due to fluctuations in the internal image acquisition device, it determines that the processing strategy is image fusion. When the response analysis module determines that the unclear initial operation image is due to external valve body ballast imbalance, it determines that the processing strategy is response hysteresis detection.

6. The underwater unmanned vehicle according to claim 5, characterized in that, The fusion processing module acquires a continuous time series of the roll angular velocity based on the preset time acquisition window, and determines the relatively stable moment based on the root mean square of the continuous time series.

7. The underwater unmanned vehicle according to claim 6, characterized in that, When the image fusion fails, the fusion processing module increases the step size of the time acquisition window based on the fusion offset value. The fusion processing module determines whether the image fusion is qualified based on the image variance and the image variance threshold; The fusion offset value is the difference between the image variance and the image variance threshold.

8. The underwater unmanned vehicle according to claim 7, characterized in that, When the hysteresis processing module determines that an imbalance event has occurred, it determines the end time of the previous jump window in the adjacent jump window as the start time of the imbalance. The lag processing module determines whether an imbalance event has occurred based on the lag time difference and the correlation difference; The correlation difference is the difference between the driving correlations calculated based on the imbalance time series in adjacent jump windows, and the lag time difference is the difference between the monitoring times corresponding to each driving correlation.

9. The underwater unmanned vehicle according to claim 8, characterized in that, The hysteresis processing module determines the hysteresis rate based on the imbalance start time, the command issuance time, the imbalance roll angle, and the command roll angle. The hysteresis processing module determines the attitude offset based on the maximum roll angle within the imbalance fluctuation range and the imbalance roll angle. The imbalance fluctuation range has an upper limit at the time of the imbalance onset and a lower limit at the time of the instruction issuance.

10. The underwater unmanned vehicle according to claim 9, characterized in that, The hysteresis processing module determines the adjustment amount based on the attitude offset and the proportional coefficient; The proportionality coefficient is positively correlated with the hysteresis rate.

Citation Information

Patent Citations

  • Unmanned-monitoring underwater cultivation robot

    CN109591982A