Sea surface target detection device and method under unmanned ship load high sea condition

Abstract

The invention discloses a sea surface target detection device and method under an unmanned ship-loaded high sea condition, relates to the field of sea condition target detection, and aims to solve the problems that an existing sea surface target detection device is greatly interfered by the environment, the detection performance is reduced, and accurate detection and flexible real-time tracking of a target are difficult to realize under the high sea condition and severe weather. Comprising a telescopic device, a pose compensation device and a detection platform, the detection platform comprises a bottom support, and a monitoring radar and an electronic level meter which are mounted on the bottom support; the pose compensation device is positioned below the detection platform and is used for controlling the detection platform to turn over and move up and down; the bottom end of the telescopic device is fixed on a deck of the unmanned ship, and the other end of the telescopic device is connected with the center of the bottom of the pose compensation device and used for controlling the height of the pose compensation device and enabling the pose compensation device to rotate so as to further control continuous tracking of the monitoring radar on a target; the sea surface target detection device under the high sea condition further comprises an early warning device, a monitoring camera and an electronic gyroscope.

Description

Technical Field

[0001] The present invention relates to the technical field of sea state target detection, and in particular to a device and method for detecting sea surface targets in high sea conditions carried by an unmanned boat. Background Art

[0002] Sea surface target detection is an important part of ocean exploration. Target detection under high sea conditions is currently the focus and difficulty in the field of sea surface target detection. For example, improving the detection accuracy of sea surface targets, operating unmanned boats to quickly approach rescue targets and issue alarm information, providing important support for rescue personnel, improving the search and rescue efficiency of sea surface personnel and ships, and minimizing casualties and property losses; the sea surface target detection technology of unmanned boats can determine whether there are other ships or vessels on the navigation path in front of the unmanned boat. If found, an alarm will be issued through the sensor, and the driving path will be adjusted in time to ensure the normal driving of the unmanned boat. It plays an important role in the autonomous navigation and obstacle avoidance of the unmanned boat, and improves the navigation safety and efficiency of the unmanned boat; radar images can be used to detect oil pollution, floating objects, etc. on the sea surface, thereby protecting the marine ecological environment and providing important support for marine economic construction. The existing patent for an obstacle avoidance device for an unmanned boat (patent number 202321805916.5) uses a traditional acoustic detection device, which is greatly affected by temperature and wind direction. The existing patent for an unmanned boat surface target tracking method, device, electronic device and storage medium (patent number 202311696640.6) uses a visual image detection device, which is easily affected by severe weather such as rain, sea fog, and heavy snow, resulting in a decrease in detection performance. Summary of the invention

[0003] The technical problems to be solved by the present invention are:

[0004] In high sea conditions and bad weather, existing sea surface target detection devices are greatly disturbed by the environment, and their detection performance is reduced, making it difficult to achieve accurate detection and flexible real-time tracking of targets.

[0005] The present invention adopts the following technical solutions to solve the above technical problems:

[0006] The present invention provides a sea surface target detection device in high sea conditions carried by an unmanned boat, the sea surface target detection device in high sea conditions comprising: a telescopic device, a posture compensation device and a detection platform;

[0007] The detection platform includes a base and a monitoring radar and an electronic level meter installed on the base;

[0008] The posture compensation device is located below the detection platform and is used to control the flipping and up and down movement of the detection platform;

[0009] The bottom end of the telescopic device is fixed on the deck of the unmanned boat, and the other end of the telescopic device is connected to the bottom center of the posture compensation device, which is used to control the height of the posture compensation device and make it rotate, so as to further control the monitoring radar to continuously track the target;

[0010] The sea surface target detection device under high sea conditions also includes an early warning device, a monitoring camera and an electronic gyroscope.

[0011] Furthermore, the posture compensation device is provided with a base, and the base is connected to the base support through a connecting rod control structure arranged along the x-axis, y-axis and z-axis. The number of connecting rod control structures arranged along the x-axis and y-axis is an even number respectively. The connecting rod control structures arranged along the x-axis and y-axis are evenly arranged at the outer edges of the bottom of the base and the detection platform, and are used to control the flipping of the detection platform along the x-axis and y-axis directions. The number of connecting rod control structures arranged along the z-axis is 1, which is located at the center of the bottom of the base and the detection platform, and is used to control the up and down movement of the detection platform along the z-axis direction; the connecting rod control structure includes Hooke's hinges at both ends and a telescopic connecting rod in the middle, and the Hooke's hinges at both ends of the connecting rod control structure are respectively connected to the bottom of the detection platform and the base support.

[0012] Furthermore, the number of connecting rod control structures arranged along the x-axis and the y-axis is two respectively, and both are located at the outer edges of the base and the bottom bracket.

[0013] Furthermore, a wave-transparent shell is provided on the upper surface of the base, and a detachable outer cover is provided on the front side thereof. The detachable outer cover is made of a ceramic-based composite material. The monitoring radar is arranged inside the wave-transparent shell, and the monitoring camera is installed on the top of the wave-transparent shell. A wedge-shaped block is arranged under the wave-transparent shell to make the illumination direction of the monitoring radar biased toward the sea surface.

[0014] Furthermore, it also includes a signal processing control unit and a data storage module, wherein the input end of the processing control unit is connected to the monitoring radar, the electronic level and the electronic gyroscope, and the output end of the processing control unit is connected to the early warning device, the telescopic device and the connecting rod control structure arranged along the x-axis and the y-axis, and the mechanical auxiliary compensation is performed by controlling the connecting rod control structure, and the telescopic device is controlled to enable the monitoring radar to continuously track the target, and the data storage module is used to temporarily store the electronic level data, the motion data collected by the electronic gyroscope and the radar detection data;

[0015] The signal processing control unit includes a signal processing module and a controller. The signal processing module contains an algorithm for fitting the three-dimensional motion data of the unmanned boat collected by the electronic gyroscope, predicting the posture of the unmanned boat at the current moment based on the motion data at the previous moment, and detecting sea surface targets based on the sea surface radar wave data of the monitoring radar and calculating the motion state of the target to be measured.

[0016] A method for detecting a sea surface target by using the sea surface target detection device in high sea conditions carried by an unmanned boat comprises the following steps:

[0017] Step 1: Fix the buffer system vertically on the deck, adjust the height of the posture compensation device and the detection platform through the telescopic device, and make the detection platform level;

[0018] Step 2: The electronic gyroscope collects the three-dimensional motion data of the unmanned boat within one cycle, pre-processes the three-dimensional motion data, fits the pre-processed three-dimensional motion data of the unmanned boat, predicts and updates the motion data of the fitted three-degree-of-freedom curve, predicts the posture of the unmanned boat at the current moment according to the motion data of the x-axis, y-axis and z-axis at the previous moment, and uses the measured value at the current moment to correct the estimated value in the prediction stage, adjusts the result, and uses mechanical auxiliary compensation to maintain the balance and stability of the detection platform;

[0019] Step 3: The monitoring radar collects sea surface information to obtain sea surface radar wave data, extracts features from radar wave signals, constructs radar signal feature data, detects sea surface targets based on a deep learning network, and when a target is detected, calculates the motion state of the target, controls the telescopic device to extend and rotate according to the motion state of the target, and realizes continuous tracking of the target by the monitoring radar;

[0020] Step 4: The early warning device sends out an alarm message when the target to be measured is detected, or when the inclination angle of the detection platform collected by the electronic level is too large or the detection platform cannot maintain balance.

[0021] Furthermore, the three-dimensional motion data is preprocessed as described in step 2. The specific process is: use the Leyte criterion to remove gross errors, and calculate the standard deviation σ and mean μ of the data at the current moment, and determine the value interval (μ-3σ, μ+3σ) according to a certain probability. It is considered that errors exceeding this interval are gross errors and this part of the data should be eliminated.

[0022] Furthermore, in step 2, the posture of the unmanned boat at the current moment is predicted based on the motion data of the x-axis, y-axis and z-axis at the previous moment. The specific process is as follows:

[0023] Assume that the fitted polynomial is an n-order polynomial, and the fitting parameters δ0, δ1, …, δn are the optimal solutions for each order parameter. For an n-degree polynomial:

[0024]

[0025] The right side of the equation is δ i Take the partial derivative and get the matrix form:

[0026]

[0027] Simplified to XA = Y, we get the coefficient matrix A;

[0028] Kalman filtering is used to predict the posture of the unmanned boat at the current moment based on the posterior estimated value in a certain direction of XYZ at the previous moment.

[0029] Furthermore, the radar wave signal characteristics in step 3 include: SPWVD marginal spectrum characteristics, phase characteristics, Doppler vector entropy, normalized Hurst exponent and relative average amplitude characteristics;

[0030] The extraction method of SPWVD marginal spectrum features is:

[0031] The SPWVD transform is expressed as follows:

[0032]

[0033] Among them, h(τ) and g(u) are window functions in the frequency domain and time domain respectively;

[0034] The corresponding discrete SPWVD is expressed as follows:

[0035]

[0036] Where L1 is the window function length of g(n), L2 is the window function length of h(k), 0≤n≤N, 0≤k≤N, N is the sampling length of the discrete sequence, Δf d represents the normalized Doppler frequency sampling interval;

[0037] The SPWVD marginal spectrum is characterized as follows:

[0038]

[0039] The phase feature extraction method is:

[0040] DEG(n)=deg[x(n)] (6)

[0041] Where deg[i] is the function for calculating the phase of the complex signal;

[0042] The method for extracting the Doppler vector entropy is:

[0043]

[0044] Among them, f d represents the Doppler frequency, X(f d ) represents the Doppler amplitude spectrum;

[0045]

[0046] Where P is the number of reference sets;

[0047] The method for extracting the normalized Hurst index is:

[0048] The echo sequence received by the radar at a certain distance unit is regarded as a random walk process:

[0049]

[0050] Where m is the scale range, and the Hurst of the radar satisfies:

[0051] F(m)~m H (10) Where H is the Hurst index. The normalized Hurst index can be expressed as:

[0052]

[0053] Among them, μ is the mean value of Hurst index, σ is the standard deviation of Hurst index;

[0054] The method for extracting the relative average amplitude feature is:

[0055] The average amplitude of the echo signal of length N is:

[0056]

[0057] The relative average amplitude is:

[0058]

[0059] Furthermore, in step 3, the sea surface target is detected, specifically: a DBO-BiLSTM neural network with controllable false alarm rate is adopted; the DBO dung beetle optimization algorithm is adopted to calculate the relevant parameters suitable for the BiLSTM model, and the optimal number of hidden layers, learning rate and regularization coefficient are obtained;

[0060] Determine the constant false alarm detection threshold for classification detection, and control different false alarm rates by setting the threshold T:

[0061] T=O2(i) (15)

[0062] i=Pfa·N clutter (16)

[0063] Among them, O2 indicates that the clutter data is arranged in ascending order according to the second bit value, Pfa is the required false alarm rate, N clutter is the number of clutter samples;

[0064] The obtained detection threshold T is compared with the second bit value in the classification probability value obtained after the input data set. If the probability P is greater than the detection threshold T, it is judged as sea clutter; if it is less than the detection threshold T, it is judged as a target.

[0065] Compared with the prior art, the present invention has the following beneficial effects:

[0066] The unmanned boat-mounted sea surface target detection device under high sea conditions of the present invention adopts a telescopic device combined with a posture compensation device to freely adjust the detection height and rotation angle according to the distance of the target from the radar to continuously track the target, and utilizes the posture compensation device to compensate for the radar shaking caused by the hull deviation under high sea conditions, thereby maintaining the relative stability of the radar and improving the detection performance and anti-interference capability of sea surface targets.

[0067] The sea surface target detection method of the present invention predicts the posture of the unmanned boat based on the motion data transmitted by the electronic gyroscope, and dynamically adjusts the radar posture; uses the radar echo data to extract relevant features, adopts a deep learning network to detect sea surface targets, and reasonably controls the monitoring radar to enter the tracking mode, so as to realize the continuous tracking of the target by the monitoring radar, greatly improve the detection accuracy and sensitivity of the target, and realize efficient and comprehensive sea surface target detection and early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 Schematic diagram of the structure of a sea surface target detection device in high sea conditions carried by an unmanned boat in an embodiment of the present invention;

[0069] Figure 2 is a flow chart of posture compensation in an embodiment of the present invention;

[0070] Figure 3 This is a flow chart of target detection and tracking in an embodiment of the present invention;

[0071] Figure 4 is a flowchart of deep learning in an embodiment of the present invention;

[0072] Figure 5 A comparison diagram of target detection results before and after neural network processing in an embodiment of the present invention;

[0073] Figure 6 4 is a comparison chart of detection probabilities of four neural networks in an embodiment of the present invention.

[0074] Figure 7 This is a scene diagram of the experimental site in an embodiment of the present invention.

[0075] Description of reference numerals:

[0076] 11-wireless communication component, 12-power supply module, 13-early warning module, 14-electronic gyroscope, 21-telescopic device, 22-posture compensation device, 24-monitoring camera, 25-electronic level, 231-circular base, 232-wave-transparent shell. DETAILED DESCRIPTION

[0077] In order to enable those skilled in the art to better understand the scheme of the present invention, exemplary implementations or embodiments of the present invention will be described below in conjunction with the accompanying drawings. Obviously, the described implementations or embodiments are only implementations or embodiments of a part of the present invention, not all of them. Based on the implementations or embodiments of the present invention, all other implementations or embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present invention.

[0078] In the description of the present invention, it should be noted that the terminology nouns in the various embodiments, such as "up", "down", "front", "back", "left", "right", etc., which indicate directions, are only for simplifying the description of the positional relationship based on the drawings in the specification, and do not mean that the referred elements and devices must be operated according to the specific directions and defined operations and methods and structures in the specification. Such directional nouns do not constitute a limitation on the present invention.

[0079] In the description of the present invention, it should be noted that the terms "first", "second", and "third" mentioned in the embodiments of the present invention are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first", "second", and "third" may explicitly or implicitly include one or more of the features.

[0080] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0081] Example 1

[0082] like Figure 1As shown, the present invention provides a sea surface target detection device under high sea conditions carried by an unmanned boat, the sea surface target detection device under high sea conditions comprises: a telescopic device 21, a posture compensation device 22 and a detection platform; the detection platform comprises a base 231 and a monitoring radar and an electronic level 25 installed on the base 231; the electronic level 25 is used to collect horizontal data of the detection platform and monitor whether a tipping phenomenon occurs; the posture compensation device 22 is located below the detection platform and is used to control the flipping and up and down movement of the detection platform; the bottom end of the telescopic device 21 is fixed on the deck of the unmanned boat, and the other end of the telescopic device 21 is connected to the bottom center of the posture compensation device 22, which is used to control the height of the posture compensation device 22 and make it rotate, so as to further control the monitoring radar to continuously track the target; the sea surface target detection device under high sea conditions also includes an early warning device 13, a monitoring camera 24 and an electronic gyroscope 14.

[0083] When the monitoring radar detects higher altitudes and the observation distance is longer, more clutter is collected accordingly. When the monitoring radar detects lower altitudes and the observation distance is shorter, the less clutter interference there is, and the more accurate the result is. The telescopic device 21 includes a telescopic rod and a rotating motor. The detection device is provided with a detection mode and a tracking mode. When in the detection mode, the monitoring radar focuses on small targets in the direction of the unmanned boat's advance, while avoiding dangerous situations such as collisions during navigation. When in the tracking mode, the telescopic device 21 receives the angle information of the target to be measured, controls the height and angle of the monitoring radar, and continuously tracks the target. The two modes cooperate with each other to complete the tracking task of the target to be measured.

[0084] Preferably, the posture compensation device 22 is provided with a base, and the base is connected to the base support 231 through a connecting rod control structure arranged along the x-axis, y-axis and z-axis. The number of connecting rod control structures arranged along the x-axis and y-axis is an even number respectively, and the connecting rod control structures arranged along the x-axis and y-axis are evenly arranged at the outer edges of the bottom of the base and the detection platform, which are used to control the flipping of the detection platform along the x-axis and y-axis directions. The number of connecting rod control structures arranged along the z-axis is 1, which is located at the center of the bottom of the base and the detection platform, and is used to control the up and down movement of the detection platform along the z-axis direction; the connecting rod control structure includes Hooke's hinges at both ends and a telescopic connecting rod in the middle, and the Hooke's hinges at both ends of the connecting rod control structure are respectively connected to the bottom of the detection platform and the base support 231.

[0085] The working mechanism of the posture compensation device 22 is: constructing an XYZ three-dimensional support mechanism, extracting the three-dimensional data of the electronic gyroscope 14, realizing the acquisition of platform data under shaking conditions, further analyzing the three-dimensional motion data under different wave conditions, drawing and fitting three-degree-of-freedom curves, and performing collaborative prediction to realize the prediction and correction of the posture of the unmanned boat.

[0086] Preferably, the number of connecting rod control structures arranged along the x-axis and the y-axis is two respectively, and both are located at the outer edge of the base and the bottom bracket 231 .

[0087] Preferably, a wave-transparent housing 232 is provided on the upper surface of the base 231, and a detachable outer cover is provided on the front side thereof, and the detachable outer cover is made of a ceramic-based composite material. The monitoring radar is arranged inside the wave-transparent housing 232, and the monitoring camera 24 is installed on the top of the wave-transparent housing 232. The monitoring camera 24 is used to shoot the real-time scene of the sea surface, and is connected to the wireless communication component 11 through a cable. When the monitoring radar is in working state, the sea surface video is transmitted back to the remote receiving end in real time. A wedge-shaped block with an inclination angle θ of 2° is arranged under the wave-transparent housing 232 to make the irradiation direction of the monitoring radar deviate toward the sea surface.

[0088] Preferably, it also includes a signal processing control unit and a data storage module, wherein the input end of the processing control unit is connected to the monitoring radar, the electronic level 25 and the electronic gyroscope 14, and the output end of the processing control unit is connected to the early warning device 13, the telescopic device 21 and the connecting rod control structure arranged along the x-axis and the y-axis, and the mechanical auxiliary compensation is performed by controlling the connecting rod control structure, and the telescopic device is controlled to enable the monitoring radar to continuously track the target, and the data storage module is used to temporarily store the data of the electronic level 25, the motion data collected by the electronic gyroscope 14 and the radar detection data;

[0089] The signal processing control unit includes a signal processing module and a controller. The signal processing module contains an algorithm for fitting the three-dimensional motion data of the unmanned boat collected by the electronic gyroscope 14, predicting the posture of the unmanned boat at the current moment based on the motion data at the previous moment, and detecting sea surface targets based on the sea surface radar wave data of the monitoring radar and calculating the motion state of the target to be measured.

[0090] The data collected by the monitoring radar is processed by the signal processing control unit and then transmitted to the wireless communication component 11 for long-distance control and observation; the electronic gyroscope 14 is fixed on the surface of the unmanned boat deck, and transmits the detected three-dimensional motion data to the signal processing control unit, and the posture compensation device 22 dynamically adjusts the radar posture according to the processed three-dimensional motion data information; the power supply module 12 is used to power the wireless communication component 11, the early warning device 13, the electronic gyroscope 14, the telescopic device 21, the posture compensation device 22, the detection platform, the monitoring camera 24 and the electronic level 25; the early warning device 13 is used to remind the relevant operators and send the radar information of the target to be measured through the wireless communication component 11 after detecting the target to be measured, and is used to send an alarm message when the inclination angle of the detection platform is too large or the detection platform is difficult to maintain self-balance, reminding the relevant operators to make timely adjustments, and transmit the alarm message to the wireless communication component 11 for remote alarm.

[0091] The signal processing module is used to process the data of the monitoring radar, analyze the speed, distance, angle, and acceleration information of the target, and transmit the data to the controller and the wireless communication component 11; the controller is used to receive the radar information processed by the signal processing module, control the telescopic movement of the telescopic device 21 according to the position information of the target to be measured, adjust the detection height of the monitoring radar to match the position of the target to be measured, and adjust the rotation angle of the telescopic device according to the angle information of the target to be measured, so that the target to be measured always remains within the radar line of sight for continuous tracking. The monitoring radar is controlled to work periodically. When there is no target, the working time interval is 2 minutes, and each time it works for 1 minute. When the target to be measured is detected, the working state is switched, the working time interval is set to 0, and the telescopic device 21 drives the detection platform to rotate to ensure that the target to be measured is always within the radar line of sight for stable detection; the mechanism of adjusting the height of the telescopic device 21 is that as the distance between the measured target and the monitoring radar increases, the detection height of the monitoring radar increases, and the maximum detection distance of the monitoring radar is 1km. As the measured target is closer to the monitoring radar, the detection height of the monitoring radar decreases. At this time, less clutter information enters the monitoring radar, and the detection accuracy of the monitoring radar is higher; the posture compensation device 22 is used to receive the electronic gyroscope data processed by the signal processing unit, predict the wave motion state, and try to keep the monitoring radar in a horizontal state, so as to reduce the influence of the unmanned boat on the working performance of the monitoring radar due to the ups and downs of the waves.

[0092] Example 2

[0093] The present invention also provides a sea surface target detection method based on the sea surface target detection device under high sea conditions carried by an unmanned boat, such as Figure 2 and Figure 3 As shown, the following steps are included:

[0094] Step 1: Fix the buffer system vertically on the deck, adjust the height of the posture compensation device through the telescopic device, and make the detection platform horizontal;

[0095] Step 2: The electronic gyroscope collects the three-dimensional motion data of the unmanned boat within one cycle, pre-processes the three-dimensional motion data, fits the pre-processed three-dimensional motion data of the unmanned boat, predicts and updates the motion data of the fitted three-degree-of-freedom curve, predicts the posture of the unmanned boat at the current moment according to the motion data of the x-axis, y-axis and z-axis at the previous moment, and uses the measured value at the current moment to correct the estimated value in the prediction stage, adjusts the result, and uses mechanical auxiliary compensation to maintain the balance and stability of the detection platform;

[0096] Step 3: The monitoring radar collects sea surface information to obtain sea surface radar wave data, extracts features from radar wave signals, constructs radar signal feature data, detects sea surface targets based on a deep learning network, and when a target is detected, calculates the motion state of the target, controls the telescopic device to extend and rotate according to the motion state of the target, and realizes continuous tracking of the target by the monitoring radar;

[0097] Step 4: the warning device 13 issues an alarm message when the target to be measured is detected, or when the inclination angle of the detection platform collected by the electronic level 25 is too large or the detection platform cannot maintain balance.

[0098] Preferably, the three-dimensional motion data is preprocessed as described in step 2, and the specific process is: use the Leyte criterion to remove gross errors, and calculate the standard deviation σ and mean μ of the data at the current moment, determine the value interval (μ-3σ, μ+3σ) according to a certain probability, and consider that errors exceeding this interval are gross errors, and this part of the data should be eliminated.

[0099] Preferably, in step 2, the posture of the unmanned boat at the current moment is predicted based on the motion data of the x-axis, y-axis and z-axis at the previous moment, and the specific process is:

[0100] Assume that the fitted polynomial is an n-order polynomial, and the fitting parameters are δ0, δ1, …, δ n As the optimal solution for each order parameter, For an n-degree polynomial:

[0101]

[0102] The right side of the equation is δ i Take the partial derivative and get the matrix form:

[0103]

[0104] Simplified to XA = Y, we get the coefficient matrix A;

[0105] Kalman filtering is used to predict the posture of the unmanned boat at the current moment based on the posterior estimated value in a certain direction of XYZ at the previous moment.

[0106] Preferably, for the radar echo data x(n), relevant features are extracted from the signal distance unit and the clutter distance unit in the echo signal, and a certain amount of data set X(n) and reference set X are prepared. P (n), split the data set X(n) into short vectors with a sequence length of m and a segmentation interval of n, and divide them into training set X(n) and training set X(n) in a ratio of 7:3. P (n) and the test set X T (n);

[0107] The radar wave signal characteristics in step 3 include: SPWVD marginal spectrum characteristics, phase characteristics, Doppler vector entropy, normalized Hurst exponent and relative average amplitude characteristics;

[0108] The extraction method of SPWVD marginal spectrum features is:

[0109] The SPWVD transform is expressed as follows:

[0110]

[0111] Among them, h(τ) and g(u) are window functions in the frequency domain and time domain respectively;

[0112] The corresponding discrete SPWVD is expressed as follows:

[0113]

[0114] Where L1 is the window function length of g(n), L2 is the window function length of h(k), 0≤n≤N, 0≤k≤N, N is the sampling length of the discrete sequence, Δf d represents the normalized Doppler frequency sampling interval;

[0115] The SPWVD marginal spectrum is characterized as follows:

[0116]

[0117] The phase feature extraction method is:

[0118] DEG(n)=deg[x(n)] (6)

[0119] Where deg[i] is the function for calculating the phase of the complex signal;

[0120] The method for extracting the Doppler vector entropy is:

[0121]

[0122] Among them, f d represents the Doppler frequency, X(f d ) represents the Doppler amplitude spectrum;

[0123]

[0124] Where P is the number of reference sets;

[0125] The method for extracting the normalized Hurst index is:

[0126] The echo sequence received by the radar at a certain distance unit is regarded as a random walk process:

[0127]

[0128] Where m is the scale range, and the Hurst of the radar satisfies:

[0129] F(m)~m H (10) Where H is the Hurst index. The normalized Hurst index can be expressed as:

[0130]

[0131] Among them, μ is the mean value of Hurst index, σ is the standard deviation of Hurst index;

[0132] The method for extracting the relative average amplitude feature is:

[0133] The average amplitude of the echo signal of length N is:

[0134]

[0135] The relative average amplitude is:

[0136]

[0137] Preprocess the echo signal:

[0138] The obtained SPWVD marginal spectrum features, phase features, Doppler vector entropy features, normalized Hurst exponents, and relative average amplitude features are fused into five-dimensional features, and the training data set X(n) and reference set X are constructed through the five-dimensional features. P (n), are all matrices with m rows and s columns, where m is the total number of training samples and s is the feature dimension. The training labels are added, and the training labels of sea clutter are set to 0 and the training labels of targets are set to 1. The final constructed data set X(n) is a matrix with m rows and (s+1) columns, and the last column is the training label. Then the data set X(n) and the reference set X P (n) Perform standardization preprocessing, namely Z-score standardization. The mean of the processed data is 0 and the standard deviation is 1. The formula is:

[0139]

[0140] in, is the mean of the original data, and σ is the standard deviation of the original data.

[0141] Preferably, in step three, the sea surface target is detected, such as Figure 4As shown in the figure, specifically: a DBO-BiLSTM neural network with controllable false alarm rate is used; the DBO dung beetle optimization algorithm is used to calculate the relevant parameters suitable for the BiLSTM model, and its optimal number of hidden layers, learning rate and regularization coefficient are obtained; then the divided training set is input into the neural network for training, and verified in the test set; different training sets and test sets are randomly selected each time to accurately test its detection performance;

[0142] Determine the constant false alarm detection threshold for classification detection. The values ​​of each node of the network output are between [0,1], which are the probability of being judged as a target and the probability of being judged as sea clutter. The sum must be 1, so the output vector of the output layer is 1×2 data, where the first value in the array represents the probability that the sample to be tested is a target, and the second value represents the probability that the sample to be tested is clutter. Different false alarm rates are controlled by setting the threshold T:

[0143] T=O2(i) (15)

[0144] i=Pfa·N clutter (16)

[0145] Among them, O2 indicates that the clutter data is arranged in ascending order according to the second bit value, Pfa is the required false alarm rate, N clutter is the number of clutter samples;

[0146] The obtained detection threshold T is compared with the second bit value in the classification probability value obtained after the input data set. If the probability P is greater than the detection threshold T, it is judged as sea clutter; if it is less than the detection threshold T, it is judged as a target.

[0147] Example 3

[0148] This embodiment uses a 77GHz millimeter-wave radar to obtain measured data of the remote dock area. The radar single beam width is 2.5°, the detection pitch angle is 14°, and the radar parameter configuration is as follows:

[0149]

[0150] The test site is selected from Figure 7As shown in the figure, inside the Yuanyao Fishing Port in Weihai, Shandong, the weak target to be detected on the sea surface is a steel oil drum with a height of 72 cm and a diameter of 44 cm. It is about 50 m away from the radar. The population size of the DBO optimization algorithm is set to 20, the maximum number of iterations is set to 10, the number of optimization parameters is set to 3, and the input dimension of the BiLSTM network is set to 3, the number of nodes in the hidden layer is set to 6, the maximum number of training times is 500, the initial learning rate is 0.01, and the Adam gradient descent algorithm is used. For the self-test data set, each frame of data in the data set collected by the radar contains 128 chirps. A single data cannot meet the actual false alarm rate requirements. Therefore, the continuous multi-frame data is spliced, and the overlapping cutting method is used to increase the number of non-repeated samples. Finally, the false alarm rate of 10 is obtained by training the measured data set. -2 The target detection probability is 0.8558. The comparison of target detection results before and after neural network processing is shown in the figure below. Figure 5 shown.

[0151] like Figure 6 As shown in FIG. 1 , the detection probabilities of the four existing detectors when Pfa=0.01 are compared. It can be seen that the detector based on the DBO-BiLSTM neural network of the present invention shows better detection performance when processing the self-test sea surface data set, and the detection probability is significantly better than that of the other three detectors.

[0152] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will fall within the protection scope of the present invention.

Claims

1. An unmanned boat-mounted sea surface target detection device under high sea conditions, characterized in that: The sea surface target detection device under high sea conditions comprises: a telescopic device (21), a posture compensation device (22) and a detection platform; The detection platform comprises a base (231) and a monitoring radar and an electronic level (25) installed on the base (231); The posture compensation device (22) is located below the detection platform and is used to control the flipping and up and down movement of the detection platform; The bottom end of the telescopic device (21) is fixed on the deck of the unmanned boat, and the other end of the telescopic device (21) is connected to the bottom center of the posture compensation device (22) to control the height of the posture compensation device (22) and make it rotate, so as to further control the monitoring radar to continuously track the target; The sea surface target detection device under high sea conditions also includes an early warning device (13), a monitoring camera (24) and an electronic gyroscope (14).

2. The unmanned boat-mounted sea surface target detection device under high sea conditions according to claim 1, characterized in that: The posture compensation device (22) is provided with a base, and the base is connected to the base support (231) via a connecting rod control structure arranged along the x-axis, y-axis and z-axis. The number of connecting rod control structures arranged along the x-axis and y-axis is an even number respectively. The connecting rod control structures arranged along the x-axis and y-axis are evenly arranged at the outer edges of the bottom of the base and the detection platform, and are used to control the detection platform to flip along the x-axis and y-axis directions. The number of connecting rod control structures arranged along the z-axis is one, which is located at the center of the bottom of the base and the detection platform, and is used to control the detection platform to move up and down along the z-axis direction. The connecting rod control structure includes Hooke's hinges at both ends and a telescopic connecting rod in the middle. The Hooke's hinges at both ends of the connecting rod control structure are respectively connected to the bottom of the detection platform and the base support (231).

3. The unmanned boat-mounted sea surface target detection device under high sea conditions according to claim 2, characterized in that: The number of connecting rod control structures arranged along the x-axis and the y-axis is two respectively, and both are located at the outer edges of the base and the bottom bracket (231).

4. The unmanned boat-mounted sea surface target detection device under high sea conditions according to claim 3, characterized in that: The upper surface of the base (231) is provided with a wave-transmitting housing (232), the front of which is a detachable outer cover, the detachable outer cover is made of a ceramic-based composite material, the monitoring radar is arranged inside the wave-transmitting housing (232), the monitoring camera (24) is installed on the top of the wave-transmitting housing (232), and a wedge block is arranged under the wave-transmitting housing (232) for making the irradiation direction of the monitoring radar deviate towards the sea surface.

5. The unmanned boat-mounted sea surface target detection device under high sea conditions according to claim 4, characterized in that: It also includes a signal processing control unit and a data storage module, wherein the input end of the processing control unit is connected to the monitoring radar, the electronic level (25) and the electronic gyroscope (14), and the output end of the processing control unit is connected to the early warning device (13), the telescopic device (21) and the connecting rod control structure arranged along the x-axis and the y-axis, and mechanical auxiliary compensation is performed by controlling the connecting rod control structure, and the telescopic device is controlled so that the monitoring radar can continuously track the target, and the data storage module is used to temporarily store the data of the electronic level (25), the motion data collected by the electronic gyroscope (14) and the radar detection data; The signal processing control unit comprises a signal processing module and a controller. The signal processing module contains an algorithm for fitting the three-dimensional motion data of the unmanned boat collected by the electronic gyroscope (14), predicting the posture of the unmanned boat at the current moment according to the motion data at the previous moment, and detecting sea surface targets according to sea surface radar wave data of the monitoring radar and calculating the motion state of the target to be detected.

6. A method for detecting sea surface targets based on the sea surface target detection device in high sea conditions carried by an unmanned boat according to any one of claims 1 to 5, characterized in that: The following steps are involved: Step 1: Fix the buffer system vertically on the deck, adjust the height of the posture compensation device and the detection platform through the telescopic device, and make the detection platform level; Step 2: The electronic gyroscope collects the three-dimensional motion data of the unmanned boat within one cycle, pre-processes the three-dimensional motion data, fits the pre-processed three-dimensional motion data of the unmanned boat, predicts and updates the motion data of the fitted three-degree-of-freedom curve, predicts the posture of the unmanned boat at the current moment according to the motion data of the x-axis, y-axis and z-axis at the previous moment, and uses the measured value at the current moment to correct the estimated value in the prediction stage, adjusts the result, and uses mechanical auxiliary compensation to maintain the balance and stability of the detection platform; Step 3: The monitoring radar collects sea surface information to obtain sea surface radar wave data, extracts features from radar wave signals, constructs radar signal feature data, detects sea surface targets based on a deep learning network, and when a target is detected, calculates the motion state of the target, controls the telescopic device to extend and rotate according to the motion state of the target, and realizes continuous tracking of the target by the monitoring radar; Step 4: When the target to be measured is detected, the early warning device (13) issues an alarm message, or when the inclination angle of the detection platform collected by the electronic level (25) is too large or the detection platform cannot maintain balance, the early warning device (13) issues an alarm message.

7. The sea surface target detection method according to claim 6, characterized in that: The three-dimensional motion data is preprocessed as described in step 2. The specific process is: use the Leyte criterion to remove gross errors, and calculate the standard deviation σ and mean μ of the data at the current moment, determine the value interval (μ-3σ, μ+3σ) according to a certain probability, and consider that errors exceeding this interval are gross errors, and this part of the data should be eliminated.

8. The sea surface target detection method according to claim 7, characterized in that: In step 2, the posture of the unmanned boat at the current moment is predicted based on the motion data of the x-axis, y-axis and z-axis at the previous moment. The specific process is as follows: Assume that the fitted polynomial is an n-order polynomial, and the fitting parameters δ0, δ1, …, δn are the optimal solutions for each order parameter. For an n-degree polynomial: The right side of the equation is δ i Take the partial derivative and get the matrix form: Simplified to XA=Y, we get the coefficient matrix A; Kalman filtering is used to predict the posture of the unmanned boat at the current moment based on the posterior estimated value in a certain direction of XYZ at the previous moment.

9. The method for detecting sea surface targets according to claim 6, characterized in that: The radar wave signal characteristics in step 3 include: SPWVD marginal spectrum characteristics, phase characteristics, Doppler vector entropy, normalized Hurst exponent and relative average amplitude characteristics; The extraction method of SPWVD marginal spectrum features is: The SPWVD transform is expressed as follows: Among them, h(τ) and g(u) are window functions in the frequency domain and time domain respectively; The corresponding discrete SPWVD is expressed as follows: Where L1 is the window function length of g(n), L2 is the window function length of h(k), 0≤n≤N, 0≤k≤N, N is the sampling length of the discrete sequence, Δf d represents the normalized Doppler frequency sampling interval; The SPWVD marginal spectrum is characterized as follows: The phase feature extraction method is: DEG(n)=deg[x(n)] (6) Where deg[i] is the function for calculating the phase of the complex signal; The method for extracting the Doppler vector entropy is: Among them, f d represents the Doppler frequency, X(f d ) represents the Doppler amplitude spectrum; Where P is the number of reference sets; The method for extracting the normalized Hurst index is: The echo sequence received by the radar at a certain distance unit is regarded as a random walk process: Where m is the scale range, and the Hurst of the radar satisfies: F(m)~m H (10) Where H is the Hurst index, and the normalized Hurst index can be expressed as: Among them, μ is the mean value of Hurst index, σ is the standard deviation of Hurst index; The method for extracting the relative average amplitude feature is: The average amplitude of the echo signal of length N is: The relative average amplitude is:

10. The sea surface target detection method according to claim 9, characterized in that: In step 3, the sea surface target is detected, specifically: a DBO-BiLSTM neural network with controllable false alarm rate is used; the DBO dung beetle optimization algorithm is used to calculate the relevant parameters suitable for the BiLSTM model, and its optimal number of hidden layers, learning rate and regularization coefficient are obtained; Determine the constant false alarm detection threshold for classification detection, and control different false alarm rates by setting the threshold T: T=O2(i) (15) i=Pfa·N clutter (16) Among them, O2 indicates that the clutter data is arranged in ascending order according to the second bit value, Pfa is the required false alarm rate, N clutter is the number of clutter samples; The obtained detection threshold T is compared with the second bit value in the classification probability value obtained after the input data set. If the probability P is greater than the detection threshold T, it is judged as sea clutter; if it is less than the detection threshold T, it is judged as a target.

Images