Shipborne water cannon platform adjusting method and device based on jet deviation prediction and ship attitude prediction

Abstract

The invention provides a shipborne water cannon platform adjustment method and device based on jet deviation prediction and ship attitude prediction, and relates to the technical field of water cannon platform adjustments, the method comprises the following steps: obtaining a real-time video frame image, a real-time brightness value, a real-time wave height, a real-time wave frequency, real-time air humidity and real-time visibility; establishing an image segmentation model in which the brightness value, the wave height, the wave frequency, the air humidity and the visibility are introduced, and inputting the real-time data into the image segmentation model to segment a water column image, a striking drop point image and a target image to be struck; and determining a real object distance deviation and iteratively adjusting a yaw angle and a pitch angle of the water cannon platform until the target is completely hit. According to the method, during image segmentation, the image segmentation model in which the brightness value, the wave height, the wave frequency, the air humidity and the visibility are introduced is used for segmenting the image, the influence of external environment factors is fully considered, the segmented image is more accurate, and it is guaranteed that a follow-up water cannon platform is accurately adjusted to hit a target.

Description

Technical Field

[0001] The present invention relates to the technical field of water cannon platform adjustment, and in particular to a method and a device for adjusting a shipborne water cannon platform based on jet deviation prediction and ship attitude prediction. Background Art

[0002] With the development of global economic globalization, more ships are traveling on major shipping routes around the world. Random moorings on the waterway hinder traffic, foreign ships cross the border, pirates rob merchant ships, etc., which interfere with the normal development of the world's shipping industry. The common way to solve the above problems is to use water cannons to expel pirates, but these law enforcement water cannons require manual observation to see if there is a jet attack, adjust the caliber of the water cannon, and observe whether the pirate ship is hit. The work is cumbersome and labor-intensive.

[0003] In the prior art, a method for evaluating the effect of ship strike based on jet edge analysis and strike deviation feedback is disclosed with publication number "CN110245569B", which includes the following steps: S1, collecting video image information through an optoelectronic pan-tilt device; S2, processing the video image information through a jet segmentation model to extract the jet contour and the ship contour; S3, judging whether there is a jet strike based on the extraction result, if there is no jet, returning to S2; if there is a jet, entering S4; S4, judging whether the jet hits the ship based on contour analysis, if the jet does not hit the ship, entering S5; if the jet hits the ship, taking the ship position as the strike coordinate and entering S7; S5, calculating the deviation of water cannon shooting in combination with the coordinate position of the target ship and the position of the actual strike point of the jet; S6, based on the position of the actual strike point of the jet and the calculated deviation, converting it into the strike coordinate of the water cannon through a coordinate conversion method, and S7, feeding back the strike coordinate to the water cannon controller to correct the control parameters of the water cannon. This method can not only determine whether there is a jet attack, but also obtain the actual jet attack point, which is fed back to the water cannon system to adjust the water cannon attack direction and achieve precise attack on the pirate ship. It solves the problem of determining whether the target is hit and obtaining the attack effect, which is difficult to achieve with traditional solutions.

[0004] However, the existing technology still has major defects. For example, when processing video images to extract jet contours and ship contours, the existing technology does not take into account the impact of external environmental factors on image segmentation accuracy (for example, the brightness value of the area to be hit, air humidity, visibility and water surface disturbance will all affect the image segmentation accuracy), resulting in the problem that the jet contours and ship contours formed by the segmentation are not accurate enough, which in turn has an adverse effect on the subsequent calculation deviation and the process of adjusting the water cannon angle, making it difficult for the water cannon to accurately hit the target after adjustment.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The object of the present invention is to provide a method and device for adjusting a shipborne water cannon platform based on jet deviation prediction and ship attitude prediction, so as to solve the problems raised in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for adjusting a shipborne water cannon platform based on jet deviation prediction and ship attitude prediction, the specific steps comprising:

[0009] S1, real-time video recording of the water cannon strike area to obtain real-time video frame images of the water cannon strike area, and processing the real-time video frame images to obtain real-time brightness value, real-time wave height and real-time wave frequency of the water cannon strike area, and obtain real-time air humidity and real-time visibility at the water cannon platform;

[0010] S2, establishing an image segmentation model that introduces brightness value, wave height, wave frequency, air humidity and visibility, and inputting the data collected in step S1 into the image segmentation model to segment the water column image, the impact point image and the target image to be hit;

[0011] S3, calculating the intersection-and-union ratio between the impact point image and the target object image, and when the intersection-and-union ratio is greater than the intersection-and-union ratio threshold, continue shooting at the current muzzle angle, and when the intersection-and-union ratio is not greater than the intersection-and-union ratio threshold, proceed to step S4;

[0012] S4, based on the target tracking technology, the three-dimensional coordinates of the target to be hit are obtained, and the three-dimensional coordinates of the target to be hit and the hit point image are processed to generate the real distance deviation, and based on the ship attitude prediction technology, the water cannon distance deviation between the water cannon at the next water cannon strike and the current water cannon is predicted, and the real distance deviation and the water cannon distance deviation are combined to generate the comprehensive distance deviation;

[0013] S5, iteratively adjusting the yaw angle and pitch angle of the water cannon platform based on the comprehensive distance deviation until the intersection-and-merge ratio is greater than the intersection-and-merge ratio threshold.

[0014] Furthermore, the step S2 includes the following steps:

[0015] S21, build an image segmentation model based on the deeplabv3+ model, select the lightweight network MobileNetV2 as the backbone network, and insert the CBAM attention module and data preprocessing module into the image segmentation model;

[0016] S22, obtaining a historical sample set, the historical sample set including historical video frame images, historical brightness values, historical wave heights, historical wave frequencies, historical air temperatures, and historical visibility, and randomly dividing the historical sample set into a training set and a test set, and marking water columns, strike points, and targets to be struck on the historical video frame images of the training set;

[0017] S23, inputting the historical sample set into a preprocessing module of the image segmentation model for preprocessing, wherein the preprocessing includes denoising the video frame image, and normalizing the video frame image, brightness value, wave height, wave frequency, air temperature, and visibility;

[0018] S24, training the image segmentation model based on the preprocessed training set, and testing the trained image segmentation model based on the preprocessed test set, until a qualified image segmentation model is obtained, and inputting the real-time data collected in step S1 into the qualified image segmentation model to segment the water column image, the impact point image and the target image to be hit.

[0019] Furthermore, the specific logic of using the preprocessed training set to train the image segmentation model is:

[0020] 1.1, extracting features from the video frame image based on the depth-separable convolution technique to generate a first feature map;

[0021] 1.2, based on the pre-processed brightness value, wave height, wave frequency, air humidity and visibility, the environmental impact value is generated, and the calculation formula is as follows:

[0022]

[0023] Wherein, E is the environmental impact value, L, V, RH, H, f are the pre-processed brightness value, visibility, air humidity, wave height and wave frequency respectively, V0, RH0 are the visibility threshold and air humidity threshold respectively, λ1, λ2, λ3, λ4, C1, γ, δ, ω1, ω2 are all unknown coefficients, and ω1+ω2=1;

[0024] 1.3, the first feature map and the environmental impact value are concatenated to form a comprehensive feature map, and the comprehensive feature map is subjected to Atrous convolution to obtain the second feature map, and the formula based on it is as follows:

[0025]

[0026] Where I1 represents the first feature map, Concat(I1,E) represents the concatenation of the first feature map and the environmental impact value, Iz represents the comprehensive feature map, Iz(x+r*k',y) represents the pixel value at the position (x+r*k',y) in the comprehensive feature map, r is the expansion rate, k' is the index of the convolution kernel used in the Atrous convolution, and k'=1,2,…,K, K represents the number of convolution kernels used in the Atrous convolution, W k' represents the weight of the k'th convolution kernel in the Atrous convolution, I2 represents the second feature map, and I2(x, y) represents the pixel value at the position (x, y) in the second feature map;

[0027] 1.4, based on the channel attention mechanism in the CBAM attention module, the second feature map is processed to generate a channel attention map. The formula is as follows:

[0028]

[0029] Among them, F avg (I2) represents the second feature map after the global average pooling operation, F max (I2) represents the second feature map after the global maximum pooling operation, σ is the sigmoid activation function, W c is the weight of the cth channel in the channel attention mechanism, and M is the channel attention map obtained after processing by the channel attention mechanism;

[0030] 1.5, based on the spatial attention mechanism in the CBAM attention module, the comprehensive feature map, channel attention map and second feature map are processed to generate a spatial attention map. The formula is as follows:

[0031] M s =σ(Conv 7×7 (Concat(M*I2,M*Iz)))

[0032] Among them, Concat(·) means using a 7×7 convolution kernel for convolution operation, Concat(M*I2,M*Iz) means concatenating M*I2 and M*Iz, and M s The spatial attention map obtained after processing by the spatial attention mechanism;

[0033] 1.6, combining the channel attention map, the spatial attention map and the second feature map to generate the feature map for segmentation, the formula based on which is as follows:

[0034] O=M s *(M*I2)

[0035] Among them, O represents the feature map for segmentation;

[0036] 1.7. Send the segmentation feature map to the classification layer of the image segmentation model for training to generate a predicted segmentation result, compare the predicted segmentation result with the actual segmentation result of the annotation, calculate the loss function value based on the cross entropy loss function, and optimize the undetermined value used in the generation of the segmentation feature map through the back propagation algorithm. Minimize the loss function value through multiple iterative training until the training is completed.

[0037] Furthermore, the calculation formula of the intersection-over-union ratio is as follows:

[0038]

[0039] Among them, A l Represents the area of ​​the impact point image, A m Indicates the area of ​​the target image to be hit, A j Indicates the intersection area of ​​the impact point image and the target image to be hit.

[0040] Furthermore, the step S4 includes the following steps:

[0041] S41, the target tracking module based on the optoelectronic gimbal obtains the three-dimensional coordinates of the target to be hit in real time, projects the three-dimensional coordinates onto the screen coordinate system, obtains the two-dimensional coordinates of the target to be hit, and determines the center coordinates of the landing point area of ​​the strike landing point image on the same screen coordinate system;

[0042] The projection formula used to project the three-dimensional coordinates of the target to be hit into two-dimensional coordinates is as follows:

[0043]

[0044] Among them, f x ,f y are the focal lengths of the cameras on the photoelectric gimbal, f x Used to indicate the magnification of the camera in the horizontal direction, f y Used to indicate the magnification of the camera in the vertical direction, c x ,c y is the coordinate of the principal point, X m ,Y m ,Z m The three-dimensional coordinates of the target to be hit obtained by the target tracking module in real time, X m Y is the horizontal coordinate of the target to be hit in the camera coordinate system, m Z is the vertical coordinate of the target to be hit in the camera coordinate system, m is the depth coordinate of the target to be hit from the camera. The depth coordinate is used to indicate the distance of the target to be hit in front of the camera. m ,y mis the two-dimensional coordinate projected onto the screen coordinate system, x m is the horizontal coordinate of the target to be hit in the screen coordinate system, y m The vertical coordinate of the target to be hit in the screen coordinate system;

[0045] S42, calculating the pixel distance deviation between the center coordinates of the landing area and the two-dimensional coordinates of the target to be hit, and the calculation formula is as follows:

[0046]

[0047] Among them, x l is the horizontal coordinate of the center of the landing area in the screen coordinate system, y l is the vertical coordinate of the center of the landing area in the screen coordinate system, Δx is the pixel distance deviation in the horizontal direction, and Δy is the pixel distance deviation in the vertical direction;

[0048] S43, calculating the physical distance deviation based on the pixel distance deviation, the calculation formula is as follows:

[0049]

[0050] Among them, d x is the actual distance deviation in the horizontal direction, d y is the actual distance deviation in the vertical direction;

[0051] S44, combines the physical distance deviation and the water monitor distance deviation to generate a comprehensive distance deviation. The calculation formula is as follows:

[0052]

[0053] Among them, dl x is the water monitor distance deviation in the horizontal direction, dl y is the water monitor distance deviation in the vertical direction, dz x is the comprehensive distance deviation in the horizontal direction, dz y is the comprehensive distance deviation in the vertical direction.

[0054] Furthermore, the step S5 comprises the following steps:

[0055] S51, converting the comprehensive distance deviation into an angle deviation, based on the following formula:

[0056]

[0057] Where L is the actual distance from the water cannon platform to the target to be hit, θ x is the yaw angle deviation, θ y is the pitch angle deviation;

[0058] S52, set the single maximum adjustment angle of the yaw angle and the pitch angle, iteratively adjust the yaw angle and the pitch angle of the water cannon platform in combination with the angle deviation and the single maximum adjustment angle, recalculate and generate the intersection-and-combination ratio after each adjustment until the intersection-and-combination ratio is greater than the intersection-and-combination ratio threshold, and compare the absolute values ​​of the yaw angle deviation and the pitch angle deviation during a single adjustment process. If the absolute value of the yaw angle deviation is greater than the absolute value of the pitch angle deviation, adjust the yaw angle first, if not, adjust the pitch angle first.

[0059] Furthermore, the specific logic of adjusting the yaw angle and pitch angle of the water cannon platform in combination with the angle deviation and the single maximum adjustment angle is as follows:

[0060] When the absolute value of the yaw angle deviation is greater than the single maximum adjustment angle of the yaw angle, the yaw angle is adjusted according to the single maximum adjustment angle of the yaw angle, and when the absolute value of the yaw angle deviation is less than the single maximum adjustment angle of the yaw angle, the yaw angle is adjusted according to the yaw angle deviation, when the absolute value of the pitch angle deviation is greater than the single maximum adjustment angle of the pitch angle, the pitch angle is adjusted according to the single maximum adjustment angle of the pitch angle, and when the absolute value of the pitch angle deviation is less than the single maximum adjustment angle of the pitch angle, the pitch angle is adjusted according to the vertical angle deviation.

[0061] A shipborne water cannon platform adjustment device based on jet deviation prediction and ship attitude prediction, used to execute the above-mentioned shipborne water cannon platform adjustment method based on jet deviation prediction and ship attitude prediction, comprising:

[0062] The data acquisition module is used to record the water cannon strike area in real time to obtain real-time video frame images of the water cannon strike area, process the real-time video frame images to obtain the real-time brightness value, real-time wave height and real-time wave frequency of the water cannon strike area, and obtain the real-time air humidity and real-time visibility at the water cannon platform;

[0063] An image segmentation module is used to establish an image segmentation model that introduces brightness value, wave height, wave frequency, air humidity and visibility, and input the data collected in step S1 into the image segmentation model to segment the water column image, the impact point image and the target image to be hit;

[0064] The strike judgment module is used to calculate the intersection-and-union ratio between the strike landing point image and the target object image. When the intersection-and-union ratio is greater than the intersection-and-union ratio threshold, the shooting continues at the current muzzle angle, and when the intersection-and-union ratio is not greater than the intersection-and-union ratio threshold, the process proceeds to step S4;

[0065] The deviation calculation module is used to obtain the three-dimensional coordinates of the target to be hit based on the target tracking technology, and process the three-dimensional coordinates of the target to be hit and the impact point image to generate the real distance deviation. Based on the ship posture prediction technology, the water cannon distance deviation between the water cannon at the next water cannon strike and the current water cannon is predicted, and the comprehensive distance deviation is generated by combining the real distance deviation and the water cannon distance deviation;

[0066] The water cannon platform adjustment module is used to iteratively adjust the yaw angle and pitch angle of the water cannon platform based on the comprehensive distance deviation until the intersection-and-combination ratio is greater than the intersection-and-combination ratio threshold.

[0067] Compared with the prior art, the present invention has the following beneficial effects:

[0068] The shipborne water cannon platform adjustment method and device based on jet deviation prediction and ship attitude prediction of the present invention uses an image segmentation model that introduces brightness value, wave height, wave frequency, air humidity and visibility to segment the image during image segmentation, fully considering the influence of external environmental factors, so that the segmented water column image, impact point image and target image to be hit are more accurate, thereby improving the accuracy of subsequent deviation calculation and water cannon platform angle adjustment, and ensuring that the water cannon platform is accurately adjusted to hit the target. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 It is a flow chart of the shipborne water cannon platform adjustment method based on jet deviation prediction and ship attitude prediction in the present invention;

[0070] Figure 2 It is a module diagram of the shipborne water cannon platform adjustment device based on jet deviation prediction and ship attitude prediction in the present invention. DETAILED DESCRIPTION

[0071] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0072] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0073] Embodiment 1:

[0074] See also Figure 1 The present invention provides a method for adjusting a shipborne water cannon platform based on jet deviation prediction and ship attitude prediction, comprising the following steps:

[0075] S1, real-time video recording of the water cannon strike area to obtain real-time video frame images of the water cannon strike area, and processing the real-time video frame images to obtain real-time brightness value, real-time wave height and real-time wave frequency of the water cannon strike area, and obtain real-time air humidity and real-time visibility at the water cannon platform;

[0076] It should be noted that obtaining video frame images is an existing technology, specifically: using a high-definition and high-frame-rate video camera to record the water cannon strike area, and then extracting and saving the video frame images in the recorded video to obtain the video frame images. To ensure the quality of the video frame images, the resolution of the video camera is at least 1080p, and the frame rate is at least not less than 30 frames per second;

[0077] It should be noted that the real-time air humidity can be obtained by a humidity sensor, and the real-time visibility can be obtained by a single-wavelength visibility meter, which is a prior art and will not be described in detail here;

[0078] It should be noted that the method for obtaining the real-time brightness value is: accumulating the brightness values ​​of all pixels in the real-time video frame image and then calculating the average value, and the average value is the real-time brightness value of the water cannon strike area. This is a prior art and will not be described in detail here;

[0079] It should be noted that the method for obtaining the real-time wave height is: based on the Canny edge detection algorithm, the peaks and troughs in the real-time video frame image are identified, and then the height difference between the peaks and the troughs is calculated. The height difference is the real-time wave height. This is a prior art and will not be described in detail here:

[0080] It should be noted that the method for obtaining the real-time wave frequency is: based on the Canny edge detection algorithm, identify the peaks in all video frame images within a unit time before real-time acquisition, count the number of peaks that appear within the unit time, and the ratio of this number to the unit time is the real-time wave frequency. The value of the unit time can be 1 second, 2 seconds, 3 seconds, etc., which is specifically set by the staff according to the actual situation and is not restricted here.

[0081] S2, establishing an image segmentation model that introduces brightness value, wave height, wave frequency, air humidity and visibility, and inputting the data collected in step S1 into the image segmentation model to segment the water column image, the impact point image and the target image to be hit, including the following steps;

[0082] S21, build an image segmentation model based on the deeplabv3+ model, select the lightweight network MobileNetV2 as the backbone network, and insert the CBAM attention module and data preprocessing module into the image segmentation model;

[0083] S22, obtaining a historical sample set, the historical sample set including historical video frame images, historical brightness values, historical wave heights, historical wave frequencies, historical air temperatures, and historical visibility, and randomly dividing the historical sample set into a training set and a test set, and marking water columns, strike points, and targets to be struck on the historical video frame images of the training set;

[0084] It should be noted that the historical video frame images, historical brightness values, historical wave heights, historical wave frequencies, historical air temperatures and historical visibility in the historical sample set are all obtained using the same method as in step S1 above, and will not be described in detail here;

[0085] S23, inputting the historical sample set into a preprocessing module of the image segmentation model for preprocessing, wherein the preprocessing includes denoising the video frame image, and normalizing the video frame image, brightness value, wave height, wave frequency, air temperature, and visibility;

[0086] It should be noted that the denoising process of the video frame image can adopt the Gaussian blur technology in the prior art. The denoising process of the video frame image can be performed by using the Gaussian blur technology to reduce the noise of the video frame image, making the video frame image smoother and softer, which is convenient for the subsequent segmentation of the video frame image. The normalization process of the video frame image can adopt the existing technologies such as maximum-minimum normalization and mean normalization, so that the pixel value of the video frame image is normalized to the range of 0-1, thereby improving the convergence speed of the subsequent image segmentation model, and then improving the subsequent segmentation rate of the video frame image by the image segmentation model, laying a favorable foundation for the real-time adjustment of the water cannon platform. The brightness value, wave height, wave frequency, air temperature and visibility are normalized by the maximum-minimum normalization in the prior art, so that the data is normalized to the range of 0-1, thereby improving the convergence speed of the subsequent image segmentation model. The maximum and minimum values ​​of the historical data are selected when normalizing the real-time data.

[0087] S24, training the image segmentation model based on the preprocessed training set, and testing the trained image segmentation model based on the preprocessed test set, until a qualified image segmentation model is obtained, and inputting the real-time data collected in step S1 into the qualified image segmentation model to segment the water column image, the impact point image, and the target image to be hit;

[0088] Among them, using the preprocessed training set to train the image segmentation model includes the following steps:

[0089] 1.1, based on MobileNetV2, feature extraction is performed on the video frame image to generate a first feature map. Specifically, the depth separable convolution technology in the prior art can be used to extract features from the video frame image. The formula based on it is as follows:

[0090]

[0091] It should be noted that I represents the preprocessed video frame image, x is the horizontal coordinate of the current pixel, y is the vertical coordinate of the current pixel, k is the radius of the depth convolution kernel, i represents the lateral offset of the current pixel in the depth convolution kernel, and i ranges from -k to k, j represents the longitudinal offset of the current pixel in the depth convolution kernel, and j ranges from -k to k, c represents the index of the image channel, and c = 1, 2, ..., C, C is the number of channels of the image, I (x + i, y + j, c) represents the pixel value of the preprocessed video frame image at position (x + i, y + j) and channel c, the video frame image collected in this implementation is an RGB image, the number of channels C of the image is 3, W c(i, j) represents the weight value of the deep convolution kernel on channel c at position (i, j). Its value is continuously optimized during training and learning to optimize the model performance. W(c) is the weight value of the point-by-point convolution on the cth channel, which is used to linearly combine the feature maps of each channel. Its value is also continuously optimized during training and learning to optimize the model performance. I c is the feature map obtained after the preprocessed video frame image is processed by deep convolution on channel c, and I1 is the first feature map finally obtained after deep separable convolution;

[0092] 1.2, based on the pre-processed brightness value, wave height, wave frequency, air humidity and visibility, the environmental impact value is generated, and the calculation formula is as follows:

[0093]

[0094] It should be noted that E is the environmental impact value, which comprehensively considers the influence of five environmental factors, namely brightness value, wave height, wave frequency, air humidity and visibility. By integrating brightness value, wave height, wave frequency, air humidity and visibility into the setting of environmental impact value, the influence of environmental factors on image segmentation is considered while reducing the amount of calculation and improving the segmentation efficiency of the subsequent image segmentation model. L, V, RH, H, f are the brightness value, visibility, air humidity, wave height and wave frequency after preprocessing, V0, RH0 are the visibility threshold, air humidity threshold, visibility threshold The thresholds of air humidity are set by the staff according to the actual situation, and can be set according to the distance from the target to be hit. For example, the visibility threshold is set to 30m, and the air humidity threshold is set to 40%. There is no restriction here. λ1,λ2,λ3,λ4,C1,γ,δ,ω1,ω2 are all undetermined coefficients, and ω1+ω2=1. These undetermined coefficients are determined by adjusting the feedback of the training during the image segmentation model training process, specifically by using the back propagation algorithm for step-by-step optimization. This is a prior art and will not be described in detail here.

[0095] It should be noted that the brightness value is an important factor affecting the quality of video frame images, and is particularly important for image segmentation. The logarithmic function can handle the nonlinear characteristics of the brightness value, especially when the brightness value is low, the change of the logarithmic function is more sensitive. Therefore, the logarithmic function ln(L+1) is used to fully reflect the influence of the brightness value on the quality of the video frame image. The undetermined coefficient λ1 is used to adjust the weight of the brightness value in the environmental impact value.

[0096] It should be noted that visibility affects the clarity of the target to be hit in the video frame image. The higher the visibility, the better the segmentation effect of the video frame image. The Sigmoid function is used To process visibility, we can better capture the impact of visibility on segmentation effect. The undetermined coefficient γ is used to control the Sigmoid function. The slope of determines the influence of visibility on the segmentation effect. The visibility threshold V0 defines the Sigmoid function. The midpoint of the image segmentation is obtained by adjusting the unknown coefficient γ and the visibility threshold V0 to adjust the influence of visibility on the image segmentation quality. The unknown coefficient λ2 is used to adjust the weight of visibility in the environmental impact value.

[0097] It should be noted that air humidity affects the propagation of light, especially near the water surface. When the humidity is high, it will cause image blur, which will have an adverse effect on subsequent image segmentation. The power function (RH-RH0) is used. δ To reflect the influence of air humidity on image segmentation, the undetermined coefficient δ determines the influence of air humidity on the segmentation effect. The air humidity threshold RH0 emphasizes the influence of air humidity on the segmentation effect within a specific range. For example, when the air humidity exceeds the air humidity threshold, the influence on the image segmentation effect will increase significantly. The influence of air humidity on the image segmentation quality can be adjusted by adjusting the undetermined coefficient δ and the air humidity threshold RH0. The undetermined coefficient λ3 is used to adjust the weight of air humidity in the environmental impact value.

[0098] It should be noted that wave height and wave frequency are important components of water surface disturbance, and water surface disturbance directly affects the refraction and reflection of light. Higher water surface disturbance will cause image blur and interfere with image segmentation. Therefore, ω1*H+ω2*f is used to quantitatively evaluate water surface disturbance, and the exponential function e(ω1*H+ω2*f) is used to reflect the influence of water surface disturbance on image segmentation. As the water surface disturbance increases, its influence on image segmentation will increase significantly. The undetermined coefficient λ4 is used to adjust the weight of air humidity in the environmental impact value;

[0099] It should be noted that the undetermined coefficient C1 is a correction constant used to make an overall correction to the environmental impact value;

[0100] 1.3, the first feature map and the environmental impact value are concatenated to form a comprehensive feature map, and the comprehensive feature map is subjected to Atrous convolution to obtain the second feature map, and the formula based on it is as follows:

[0101]

[0102] It should be noted that Concat(I1,E) represents concatenating the first feature map and the environmental impact value. Specifically, the first feature map and the environmental impact value can be concatenated using TensorFlow technology or PyTorch technology. When using TensorFlow technology for concatenation, the environmental impact value is first expanded using tf.expand_dims technology, so that the environmental impact value is expanded to the same spatial dimension as the first feature map, and then the two are concatenated. When using PyTorch technology for concatenation, the environmental impact value is first expanded using unsqueeze technology, so that the environmental impact value is expanded to the same spatial dimension as the first feature map, and then the two are concatenated. Iz represents a comprehensive feature map formed by concatenating the first feature map and the environmental impact value. By concatenating the first feature map and the environmental impact value to form a comprehensive feature map that takes environmental impact into consideration, environmental impact factors are considered in subsequent image segmentation to improve the accuracy of subsequent image segmentation.

[0103] It should be noted that Iz(x+r*k',y) represents the pixel value at the position (x+r*k',y) in the comprehensive feature map, r is the expansion rate, such as when r=2, it means skipping 1 pixel for convolution, k' is the index of the convolution kernel used in the Atrous convolution, and k'=1,2,…,K, K represents the number of convolution kernels used in the Atrous convolution, W k' represents the weight of the k'th convolution kernel in the Atrous convolution, which is continuously optimized during training and learning to optimize the model performance. I2 represents the second feature map, and I2(x, y) represents the pixel value at the position (x, y) in the second feature map;

[0104] 1.4, based on the channel attention mechanism in the CBAM attention module, the second feature map is processed to generate a channel attention map. The formula is as follows:

[0105]

[0106] It should be noted that F avg (I2) represents the second feature map after the global average pooling operation, F max (I2) represents the second feature map after the global maximum pooling operation. The global average pooling and global maximum pooling of the second feature map can be performed using existing technologies, which will not be described in detail here. σ is the sigmoid activation function, W c is the weight of the cth channel in the channel attention mechanism, and its value is continuously optimized during training to optimize the model performance, and M is the channel attention map obtained after processing by the channel attention mechanism;

[0107] 1.5, based on the spatial attention mechanism in the CBAM attention module, the comprehensive feature map, channel attention map and second feature map are processed to generate a spatial attention map. The formula is as follows:

[0108] M s =σ(Conv 7×7 (Concat(M*I2,M*Iz)))

[0109] It should be noted that Concat(·) means using a 7×7 convolution kernel for convolution operation, M s The spatial attention map obtained after the spatial attention mechanism is processed firstly weights the second feature map and the comprehensive feature map through the channel attention map to form two weighted feature maps, then concatenates the two weighted feature maps, and finally convolves the concatenated weighted feature maps through the spatial attention mechanism, so that the final spatial attention map comprehensively considers the channel and environmental influencing factors;

[0110] 1.6, combining the channel attention map, the spatial attention map and the second feature map to generate the feature map for segmentation, the formula based on which is as follows:

[0111] O=M s *(M*I2)

[0112] It should be noted that O represents a feature map for segmentation. The second feature map is adjusted by the channel attention map and the spatial attention map to obtain the final feature map for segmentation. The image segmentation model subsequently analyzes and identifies different regions according to the feature map for segmentation to improve the convergence speed of the image segmentation model during training.

[0113] 1.7, sending the segmentation feature map to the classification layer of the image segmentation model for training to generate a predicted segmentation result, comparing the predicted segmentation result with the annotated actual segmentation result, calculating the loss function value based on the cross entropy loss function, and optimizing the undetermined value used in generating the segmentation feature map through the back propagation algorithm, and minimizing the loss function value through multiple iterative training until the training is completed;

[0114] Among them, the undetermined values ​​include but are not limited to W c (i,j),W(c),V0,RH0,λ1,λ2,λ3,λ4,C1,γ,δ,ω1,ω2,W k' , W c ;

[0115] It should be noted that the existing technology can be used to train the image segmentation model using features extracted from the original data (such as the segmentation feature map in this embodiment). For example, the classification layer of the image segmentation model can use a fully connected layer, and the loss function value is calculated based on the cross entropy loss function. It is an existing technology, and the optimization parameters based on the back propagation algorithm are also an existing technology. The termination condition of the training can be reaching a predetermined number of iterations or the loss function value is less than a preset threshold. The specific setting is made by the staff according to the actual situation and is not limited here.

[0116] Among them, the standard for passing the image segmentation model test is: the accuracy of the water column image, the impact point image and the target image segmented by the image segmentation model is higher than 92%, 95% or 97%, etc., which is specifically set by the staff according to the actual situation and is not restricted here. If the test fails, return to the training step for retraining until the test passes.

[0117] S3, calculating the intersection-and-union ratio between the impact point image and the target object image, and when the intersection-and-union ratio is greater than the intersection-and-union ratio threshold, continue shooting at the current muzzle angle, and when the intersection-and-union ratio is not greater than the intersection-and-union ratio threshold, proceed to step S4;

[0118] The calculation formula of intersection-over-union ratio is as follows:

[0119]

[0120] It should be noted that A l Represents the area of ​​the impact point image, A m Indicates the area of ​​the target image to be hit, A j A represents the intersection area of ​​the impact point image and the target image to be hit. l ,A m ,A j All of them can be obtained through image processing software, which will not be described in detail here. IoU stands for intersection over union, and the larger the IoU, the higher the overlap between the impact point and the target to be hit, which means that the water cannon's impact is more accurate;

[0121] As an implementation method, the intersection-over-union ratio threshold is set to 0.7. When the intersection-over-union ratio IoU=0, it means that the water column has not hit the target to be hit. When the intersection-over-union ratio IoU is greater than 0.7, it is considered that the target to be hit is completely hit, and the current muzzle angle is continued to be fired. When the intersection-over-union ratio IoU is less than 0.7, it is considered that the target to be hit is not completely hit or is not hit, and step S4 is entered to calculate the physical distance deviation. Of course, the intersection-over-union ratio threshold can also be set to other values ​​between 0-1, which is specifically set by the staff according to the actual situation and is not limited here;

[0122] S4, based on the target tracking technology, the three-dimensional coordinates of the target to be hit are obtained, and the three-dimensional coordinates of the target to be hit and the hit point image are processed to generate a physical distance deviation, and based on the ship attitude prediction technology, the water cannon distance deviation between the water cannon at the next water cannon strike and the current water cannon is predicted, and the physical distance deviation and the water cannon distance deviation are combined to generate a comprehensive distance deviation, including the following steps:

[0123] S41, the target tracking module based on the optoelectronic gimbal obtains the three-dimensional coordinates of the target to be hit in real time, projects the three-dimensional coordinates onto the screen coordinate system, obtains the two-dimensional coordinates of the target to be hit, and determines the center coordinates of the landing point area of ​​the strike landing point image on the same screen coordinate system;

[0124] It should be noted that determining the center coordinates of the landing area is an existing technology, which will not be described in detail here. The optoelectronic pan-tilt can use a ship-borne dual-spectrum thermal imaging night vision pan-tilt camera. The target tracking module is a built-in module of the optoelectronic pan-tilt. The target tracking module based on the optoelectronic pan-tilt obtains the three-dimensional coordinates of the target to be hit in real time, which is an existing technology and will not be described in detail here.

[0125] Among them, projecting the three-dimensional coordinates of the target to be hit into two-dimensional coordinates is an existing technology, and the projection formula used is as follows:

[0126]

[0127] It should be noted that f x ,f y are the focal lengths of the cameras on the photoelectric gimbal, where f x Used to indicate the magnification of the camera in the horizontal direction, f y Used to indicate the magnification of the camera in the vertical direction, f x ,f y All of them can be obtained through the parameter table of the camera on the photoelectric pan / tilt. x ,c y The coordinates of the principal point, specifically the coordinates of the camera coordinate system origin projected on the screen coordinate system, X m ,Y m ,Z m The target tracking module obtains the three-dimensional coordinates of the target to be hit in real time, where X m Y is the horizontal coordinate of the target to be hit in the camera coordinate system, m Z is the vertical coordinate of the target to be hit in the camera coordinate system, m is the depth coordinate of the target to be hit from the camera. The depth coordinate is used to indicate the distance of the target to be hit in front of the camera. m ,y m is the two-dimensional coordinate projected onto the screen coordinate system, where x m is the horizontal coordinate of the target to be hit in the screen coordinate system, ym The vertical coordinate of the target to be hit in the screen coordinate system;

[0128] S42, calculating the pixel distance deviation between the center coordinates of the landing area and the two-dimensional coordinates of the target to be hit, and the calculation formula is as follows:

[0129]

[0130] It should be noted that x l is the horizontal coordinate of the center of the landing area in the screen coordinate system, y l is the vertical coordinate of the center of the landing area in the screen coordinate system, Δx is the pixel distance deviation in the horizontal direction, which is used to indicate the number of pixels that need to be adjusted in the horizontal direction, and Δy is the pixel distance deviation in the vertical direction, which is used to indicate the number of pixels that need to be adjusted in the vertical direction;

[0131] S43, calculating the physical distance deviation based on the pixel distance deviation, the calculation formula is as follows:

[0132]

[0133] It should be noted that d x is the actual distance deviation in the horizontal direction, which is used to indicate the actual distance that needs to be adjusted in the horizontal direction, d y It is the actual distance deviation in the vertical direction, which is used to indicate the actual distance that needs to be adjusted in the vertical direction;

[0134] S44, combines the physical distance deviation and the water monitor distance deviation to generate a comprehensive distance deviation. The calculation formula is as follows:

[0135]

[0136] It should be noted that dl x is the water monitor distance deviation in the horizontal direction, dl y is the water cannon distance deviation in the vertical direction, and the water cannon distance deviation in the horizontal and vertical directions dl x and dl y The coordinates of the water cannon platform at the next water cannon strike and the current water cannon platform coordinates can be obtained by using the ship attitude prediction technology. Specifically, according to the patented technology of ship attitude prediction and shipborne water cannon muzzle angle adjustment method with publication number "CN115950395A", the coordinates of the water cannon platform at the next water cannon strike and the current water cannon platform coordinates can be obtained. The water cannon platform coordinates at the next water cannon strike minus the current water cannon platform coordinates can be obtained to obtain the water cannon distance deviation dl in the horizontal and vertical directions. x and dly, which is a prior art and will not be described in detail here, and the transformation of the three-dimensional coordinates and the two-dimensional coordinates used in this process can adopt the technology in this embodiment or the prior art, and will not be limited here, dzx is the comprehensive distance deviation in the horizontal direction, dz y is the comprehensive distance deviation in the vertical direction.

[0137] S5, iteratively adjusting the yaw angle and pitch angle of the water cannon platform based on the comprehensive distance deviation until the intersection-and-union ratio is greater than the intersection-and-union ratio threshold, including the following steps:

[0138] S51, converting the comprehensive distance deviation into an angle deviation, based on the following formula:

[0139]

[0140] It should be noted that L is the actual distance from the water cannon platform to the target to be hit, which can be obtained by ranging equipment such as laser rangefinders, θ x is the yaw angle deviation, θ y is the pitch angle deviation;

[0141] S52, setting the single maximum adjustment angle of the yaw angle and the pitch angle, iteratively adjusting the yaw angle and the pitch angle of the water cannon platform in combination with the angle deviation and the single maximum adjustment angle, recalculating and generating the intersection-and-combination ratio after each adjustment until the intersection-and-combination ratio is greater than the intersection-and-combination ratio threshold, and comparing the absolute values ​​of the yaw angle deviation and the pitch angle deviation during a single adjustment process, if the absolute value of the yaw angle deviation is greater than the absolute value of the pitch angle deviation, the yaw angle is adjusted first, if not, the pitch angle is adjusted first;

[0142] Among them, the method for adjusting the yaw angle and pitch angle of the water cannon platform in combination with the angle deviation and the single maximum adjustment angle is: when the absolute value of the yaw angle deviation is greater than the single maximum adjustment angle of the yaw angle, the yaw angle is adjusted according to the single maximum adjustment angle of the yaw angle, and when the absolute value of the yaw angle deviation is less than the single maximum adjustment angle of the yaw angle, the yaw angle is adjusted according to the yaw angle deviation; when the absolute value of the pitch angle deviation is greater than the single maximum adjustment angle of the pitch angle, the pitch angle is adjusted according to the single maximum adjustment angle of the pitch angle; and when the absolute value of the pitch angle deviation is less than the single maximum adjustment angle of the pitch angle, the pitch angle is adjusted according to the vertical angle deviation. Since there is a certain error in the conversion of two-dimensional coordinate and three-dimensional coordinate information, a large rotation angle at one time will lead to inaccurate results. Here, an iterative adjustment algorithm is used to effectively improve the adjustment accuracy;

[0143] It should be noted that the single maximum adjustment angle of the yaw angle and the pitch angle can be 3 degrees, 4 degrees, 5 degrees or 7 degrees, etc., which are specifically set by the staff according to the actual situation and are not limited here.

[0144] Embodiment 2:

[0145] See also Figure 2The present invention provides a shipborne water cannon platform adjustment device based on jet deviation prediction and ship attitude prediction, which is used to execute the shipborne water cannon platform adjustment method based on jet deviation prediction and ship attitude prediction in the above-mentioned embodiment 1, comprising:

[0146] The data acquisition module is used to record the water cannon strike area in real time to obtain real-time video frame images of the water cannon strike area, process the real-time video frame images to obtain the real-time brightness value, real-time wave height and real-time wave frequency of the water cannon strike area, and obtain the real-time air humidity and real-time visibility at the water cannon platform;

[0147] An image segmentation module is used to establish an image segmentation model that introduces brightness value, wave height, wave frequency, air humidity and visibility, and input the data collected in step S1 into the image segmentation model to segment the water column image, the impact point image and the target image to be hit;

[0148] The strike judgment module is used to calculate the intersection-and-union ratio between the strike landing point image and the target object image. When the intersection-and-union ratio is greater than the intersection-and-union ratio threshold, the shooting continues at the current muzzle angle, and when the intersection-and-union ratio is not greater than the intersection-and-union ratio threshold, the process proceeds to step S4;

[0149] The deviation calculation module is used to obtain the three-dimensional coordinates of the target to be hit based on the target tracking technology, and process the three-dimensional coordinates of the target to be hit and the impact point image to generate the real distance deviation. Based on the ship posture prediction technology, the water cannon distance deviation between the water cannon at the next water cannon strike and the current water cannon is predicted, and the comprehensive distance deviation is generated by combining the real distance deviation and the water cannon distance deviation;

[0150] The water cannon platform adjustment module is used to iteratively adjust the yaw angle and pitch angle of the water cannon platform based on the comprehensive distance deviation until the intersection-and-combination ratio is greater than the intersection-and-combination ratio threshold.

[0151] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0152] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0153] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0154] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for adjusting a shipborne water cannon platform based on jet deviation prediction and ship attitude prediction, characterized in that: The specific steps include: S1, real-time video recording of the water cannon strike area to obtain real-time video frame images of the water cannon strike area, and processing the real-time video frame images to obtain real-time brightness value, real-time wave height and real-time wave frequency of the water cannon strike area, and obtain real-time air humidity and real-time visibility at the water cannon platform; S2, establishing an image segmentation model that introduces brightness value, wave height, wave frequency, air humidity and visibility, and inputting the data collected in step S1 into the image segmentation model to segment the water column image, the impact point image and the target image to be hit; S3, calculating the intersection-and-union ratio between the impact point image and the target object image, and when the intersection-and-union ratio is greater than the intersection-and-union ratio threshold, continue shooting at the current muzzle angle, and when the intersection-and-union ratio is not greater than the intersection-and-union ratio threshold, proceed to step S4; S4, based on the target tracking technology, the three-dimensional coordinates of the target to be hit are obtained, and the three-dimensional coordinates of the target to be hit and the hit point image are processed to generate the real distance deviation, and based on the ship attitude prediction technology, the water cannon distance deviation between the water cannon at the next water cannon strike and the current water cannon is predicted, and the real distance deviation and the water cannon distance deviation are combined to generate the comprehensive distance deviation; S5, iteratively adjusting the yaw angle and pitch angle of the water cannon platform based on the comprehensive distance deviation until the intersection-and-merge ratio is greater than the intersection-and-merge ratio threshold.

2. The method for adjusting a shipborne water cannon platform based on jet deviation prediction and ship attitude prediction according to claim 1 is characterized in that: The step S2 comprises the following steps: S21, build an image segmentation model based on the deeplabv3+ model, select the lightweight network MobileNetV2 as the backbone network, and insert the CBAM attention module and data preprocessing module into the image segmentation model; S22, obtaining a historical sample set, the historical sample set including historical video frame images, historical brightness values, historical wave heights, historical wave frequencies, historical air temperatures, and historical visibility, and randomly dividing the historical sample set into a training set and a test set, and marking water columns, strike points, and targets to be struck on the historical video frame images of the training set; S23, inputting the historical sample set into a preprocessing module of the image segmentation model for preprocessing, wherein the preprocessing includes denoising the video frame image, and normalizing the video frame image, brightness value, wave height, wave frequency, air temperature, and visibility; S24, training the image segmentation model based on the preprocessed training set, and testing the trained image segmentation model based on the preprocessed test set, until a qualified image segmentation model is obtained, and inputting the real-time data collected in step S1 into the qualified image segmentation model to segment the water column image, the impact point image and the target image to be hit.

3. The method for adjusting a shipborne water cannon platform based on jet deviation prediction and ship attitude prediction according to claim 2 is characterized in that: The specific logic of training the image segmentation model using the preprocessed training set is: 1.1, extracting features from the video frame image based on the depth-separable convolution technique to generate a first feature map; 1.2, based on the pre-processed brightness value, wave height, wave frequency, air humidity and visibility, the environmental impact value is generated, and the calculation formula is as follows: Wherein, E is the environmental impact value, L, V, RH, H, f are the pre-processed brightness value, visibility, air humidity, wave height and wave frequency respectively, V0, RH0 are the visibility threshold and air humidity threshold respectively, λ1, λ2, λ3, λ4, C1, γ, δ, ω1, ω2 are all unknown coefficients, and ω1+ω2=1; 1.3, the first feature map and the environmental impact value are concatenated to form a comprehensive feature map, and the comprehensive feature map is subjected to Atrous convolution to obtain the second feature map, and the formula based on it is as follows: Where I1 represents the first feature map, Concat(I1,E) represents the concatenation of the first feature map and the environmental impact value, Iz represents the comprehensive feature map, Iz(x+r*k',y) represents the pixel value at the position (x+r*k',y) in the comprehensive feature map, r is the expansion rate, k' is the index of the convolution kernel used in the Atrous convolution, and k'=1,2,…,K, K represents the number of convolution kernels used in the Atrous convolution, W k' represents the weight of the k'th convolution kernel in the Atrous convolution, I2 represents the second feature map, and I2(x, y) represents the pixel value at the position (x, y) in the second feature map; 1.4, based on the channel attention mechanism in the CBAM attention module, the second feature map is processed to generate a channel attention map. The formula is as follows: Among them, F avg (I2) represents the second feature map after the global average pooling operation, F max (I2) represents the second feature map after the global maximum pooling operation, σ is the sigmoid activation function, W c is the weight of the cth channel in the channel attention mechanism, and M is the channel attention map obtained after processing by the channel attention mechanism; 1.5, based on the spatial attention mechanism in the CBAM attention module, the comprehensive feature map, channel attention map and second feature map are processed to generate a spatial attention map. The formula is as follows: M s =σ(Conv 7×7 (Concat(M*I2,M*Iz))) Among them, Concat(·) means using a 7×7 convolution kernel for convolution operation, Concat(M*I2,M*Iz) means concatenating M*I2 and M*Iz, and M s The spatial attention map obtained after processing by the spatial attention mechanism; 1.6, combining the channel attention map, the spatial attention map and the second feature map to generate the feature map for segmentation, the formula based on which is as follows: O=M s *(M*I2) Among them, O represents the feature map for segmentation; 1.

7. Send the segmentation feature map to the classification layer of the image segmentation model for training to generate a predicted segmentation result, compare the predicted segmentation result with the actual segmentation result of the annotation, calculate the loss function value based on the cross entropy loss function, and optimize the undetermined value used in the generation of the segmentation feature map through the back propagation algorithm. Minimize the loss function value through multiple iterative training until the training is completed.

4. The method for adjusting a shipborne water cannon platform based on jet deviation prediction and ship attitude prediction according to claim 1, characterized in that: The calculation formula of the intersection-over-union ratio is as follows: Among them, A l Represents the area of ​​the impact point image, A m Indicates the area of ​​the target image to be hit, A j Indicates the intersection area of ​​the impact point image and the target image to be hit.

5. The method for adjusting a shipborne water cannon platform based on jet deviation prediction and ship attitude prediction according to claim 1, characterized in that: The step S4 comprises the following steps: S41, the target tracking module based on the optoelectronic gimbal obtains the three-dimensional coordinates of the target to be hit in real time, projects the three-dimensional coordinates onto the screen coordinate system, obtains the two-dimensional coordinates of the target to be hit, and determines the center coordinates of the landing point area of ​​the strike landing point image on the same screen coordinate system; The projection formula used to project the three-dimensional coordinates of the target to be hit into two-dimensional coordinates is as follows: Among them, f x ,f y are the focal lengths of the cameras on the photoelectric gimbal, f x Used to indicate the magnification of the camera in the horizontal direction, f y Used to indicate the magnification of the camera in the vertical direction, c x ,c y is the coordinate of the principal point, X m ,Y m ,Z m The three-dimensional coordinates of the target to be hit obtained by the target tracking module in real time, X m Y is the horizontal coordinate of the target to be hit in the camera coordinate system, m Z is the vertical coordinate of the target to be hit in the camera coordinate system, m is the depth coordinate of the target to be hit from the camera. The depth coordinate is used to indicate the distance of the target to be hit in front of the camera. m ,y m is the two-dimensional coordinate projected onto the screen coordinate system, x m is the horizontal coordinate of the target to be hit in the screen coordinate system, y m The vertical coordinate of the target to be hit in the screen coordinate system; S42, calculating the pixel distance deviation between the center coordinates of the landing area and the two-dimensional coordinates of the target to be hit, and the calculation formula is as follows: Among them, x l is the horizontal coordinate of the center of the landing area in the screen coordinate system, yl is the vertical coordinate of the center of the landing area in the screen coordinate system, Δx is the pixel distance deviation in the horizontal direction, and Δy is the pixel distance deviation in the vertical direction; S43, calculating the physical distance deviation based on the pixel distance deviation, the calculation formula is as follows: Among them, d x is the actual distance deviation in the horizontal direction, d y is the actual distance deviation in the vertical direction; S44, combines the physical distance deviation and the water monitor distance deviation to generate a comprehensive distance deviation. The calculation formula is as follows: Among them, dl x is the water monitor distance deviation in the horizontal direction, dl y is the water monitor distance deviation in the vertical direction, dz x is the comprehensive distance deviation in the horizontal direction, dz y is the comprehensive distance deviation in the vertical direction.

6. The method for adjusting a shipborne water cannon platform based on jet deviation prediction and ship attitude prediction according to claim 1, characterized in that: The step S5 comprises the following steps: S51, converting the comprehensive distance deviation into an angle deviation, based on the following formula: Where L is the actual distance from the water cannon platform to the target to be hit, θ x is the yaw angle deviation, θ y is the pitch angle deviation; S52, set the single maximum adjustment angle of the yaw angle and the pitch angle, iteratively adjust the yaw angle and the pitch angle of the water cannon platform in combination with the angle deviation and the single maximum adjustment angle, recalculate and generate the intersection-and-combination ratio after each adjustment until the intersection-and-combination ratio is greater than the intersection-and-combination ratio threshold, and compare the absolute values ​​of the yaw angle deviation and the pitch angle deviation during a single adjustment process. If the absolute value of the yaw angle deviation is greater than the absolute value of the pitch angle deviation, adjust the yaw angle first, if not, adjust the pitch angle first.

7. The method for adjusting a shipborne water cannon platform based on jet deviation prediction and ship attitude prediction according to claim 6, characterized in that: The specific logic of adjusting the yaw angle and pitch angle of the water cannon platform in combination with the angle deviation and the single maximum adjustment angle is: When the absolute value of the yaw angle deviation is greater than the single maximum adjustment angle of the yaw angle, the yaw angle is adjusted according to the single maximum adjustment angle of the yaw angle, and when the absolute value of the yaw angle deviation is less than the single maximum adjustment angle of the yaw angle, the yaw angle is adjusted according to the yaw angle deviation, when the absolute value of the pitch angle deviation is greater than the single maximum adjustment angle of the pitch angle, the pitch angle is adjusted according to the single maximum adjustment angle of the pitch angle, and when the absolute value of the pitch angle deviation is less than the single maximum adjustment angle of the pitch angle, the pitch angle is adjusted according to the vertical angle deviation.

8. A shipborne water cannon platform adjustment device based on jet deviation prediction and ship attitude prediction, used to execute the shipborne water cannon platform adjustment method based on jet deviation prediction and ship attitude prediction as described in any one of claims 1 to 7, characterized in that: include: The data acquisition module is used to record the water cannon strike area in real time to obtain real-time video frame images of the water cannon strike area, process the real-time video frame images to obtain the real-time brightness value, real-time wave height and real-time wave frequency of the water cannon strike area, and obtain the real-time air humidity and real-time visibility at the water cannon platform; An image segmentation module is used to establish an image segmentation model that introduces brightness value, wave height, wave frequency, air humidity and visibility, and input the data collected in step S1 into the image segmentation model to segment the water column image, the impact point image and the target image to be hit; The strike judgment module is used to calculate the intersection-and-union ratio between the strike landing point image and the target object image. When the intersection-and-union ratio is greater than the intersection-and-union ratio threshold, the shooting continues at the current muzzle angle, and when the intersection-and-union ratio is not greater than the intersection-and-union ratio threshold, the process proceeds to step S4; The deviation calculation module is used to obtain the three-dimensional coordinates of the target to be hit based on the target tracking technology, and process the three-dimensional coordinates of the target to be hit and the impact point image to generate the real distance deviation. Based on the ship posture prediction technology, the water cannon distance deviation between the water cannon at the next water cannon strike and the current water cannon is predicted, and the comprehensive distance deviation is generated by combining the real distance deviation and the water cannon distance deviation; The water cannon platform adjustment module is used to iteratively adjust the yaw angle and pitch angle of the water cannon platform based on the comprehensive distance deviation until the intersection-and-combination ratio is greater than the intersection-and-combination ratio threshold.

Images