An image recognition missile tracking intelligent processing method and system

Abstract

The invention provides an image recognition missile tracking intelligent processing method and system, which relates to the technical field of image processing. The method performs spin distortion analysis on first image data of a target object through spin state data to obtain rotational fuzzy displacement, obtains missile aerodynamic lift and missile lateral force through airflow disturbance analysis, performs aerodynamic distortion analysis according to the missile aerodynamic lift, missile lateral force and missile attitude angle to obtain field of view offset, superimposes the field of view offset and the rotational fuzzy displacement to obtain a joint distortion, corrects the first image data of the target object through geometric distortion correction according to the joint distortion to obtain second image data of the target object, identifies the image position of the target object through a fast feature point matching algorithm, and compares the target object with an expected position on a preset flight trajectory, triggers an early warning mechanism when the target object deviates from the expected position range threshold, and executes a preset missile flight adjustment strategy, so as to realize more accurate missile intelligent tracking.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to an intelligent processing method and system for missile tracking based on image recognition. Background Art

[0002] With the continuous increase in the flight speed of missiles and the increasing complexity of external environmental interference factors, in order to ensure the stability of missiles under complex flight conditions, missiles usually perform spin motion during flight. However, the rotation effect and aerodynamic disturbance caused by the missile spin cause significant distortion problems on the images collected by the camera. These image distortions seriously affect the recognition accuracy of the target, thereby interfering with the flight tracking and control of the missile. Although there are currently related studies on image correction in the missile flight state, most of these studies focus on single image correction and cannot effectively deal with the image distortion problem caused by aerodynamic interference in the spinning state. Therefore, how to comprehensively consider aerodynamic factors in the missile spinning state to achieve real-time correction of image distortion and improve the accuracy of the image to achieve more accurate missile intelligent tracking is an urgent problem to be solved. Summary of the invention

[0003] Technical issues to be solved

[0004] The purpose of the present invention is to provide an image recognition missile tracking intelligent processing method and system to solve the problem that the missile is affected by aerodynamic disturbances in a spinning state, resulting in image distortion and thus affecting the missile tracking accuracy.

[0005] Technical Solution

[0006] To achieve the above object, on the one hand, the present invention provides an intelligent processing method for missile tracking based on image recognition, the method comprising:

[0007] The missile's spin state data is acquired through an inertial measurement unit, wherein the spin state data includes a spin speed and a missile attitude angle. The first image data of the target object collected during the missile's flight is acquired through a camera. The first image data of the target object is subjected to spin distortion analysis according to the spin speed to obtain a rotational fuzzy displacement.

[0008] The flight environment parameters including the atmospheric density and the wind speed are obtained through the meteorological sensor, and the missile flight speed is obtained. The missile aerodynamic lift and the missile lateral force are obtained through the airflow disturbance analysis according to the missile flight speed, the spin state data and the flight environment parameters.

[0009] Aerodynamic distortion analysis is performed according to the missile's aerodynamic lift, missile's lateral force and missile's attitude angle to obtain the field of view offset, the field of view offset is superimposed with the rotational blur displacement to obtain the joint distortion, and the first image data of the target object is corrected through geometric distortion correction according to the joint distortion to obtain the second image data of the target object.

[0010] The multi-scale features of the target object in the second image data of the target object are obtained, and the image position of the target object is identified through a fast feature point matching algorithm based on the multi-scale features of the target object. The image position of the target object is compared with the expected position on the preset flight trajectory. When it deviates from the expected position range threshold, the early warning mechanism is triggered and the preset missile flight adjustment strategy is executed.

[0011] Furthermore, the method of performing spin distortion analysis on the first image data of the target object according to the spin speed to obtain the rotation blur displacement includes:

[0012] According to the spin speed, the rotation angle of the image acquired at each moment during the missile flight is obtained for:

[0013] .

[0014] in is the spin speed, It's time. It's in time The image rotation angle at time, When the rotation angle exceeds 360 degrees, the angle is reset to zero and the calculation starts again.

[0015] Extract single-frame images of the first image data at each moment, and extract edge contours and texture features of the target object in the single-frame images through an edge detection algorithm to obtain the position distribution coordinates of key pixel points of the target object in the image.

[0016] The rotation axis coordinates are determined according to the missile attitude angle, and the distance between the key pixel point and the rotation axis is obtained according to the position distribution coordinates of the key pixel point of the target object in the image and the rotation axis coordinates.

[0017] The rotation blur displacement is obtained according to the distance between the key pixel point and the rotation axis and the rotation speed. for:

[0018] .

[0019] in It is The distance between the key pixel point of the target object and the rotation axis, is the rotation blur displacement.

[0020] Furthermore, the missile attitude angle includes a missile pitch angle and a missile yaw angle, and the method of determining the rotation axis coordinates according to the missile attitude angle and obtaining the distance between the key pixel point and the rotation axis according to the position distribution coordinates of the key pixel point of the target object in the image and the rotation axis coordinates includes:

[0021] The vertical offset of the rotation axis is obtained according to the missile pitch angle and the vertical resolution of the image. for:

[0022] .

[0023] in is the missile pitch angle, is the image vertical resolution, It is the vertical offset of the rotation axis.

[0024] The horizontal offset of the rotation axis is obtained according to the missile yaw angle and the horizontal resolution of the image. for:

[0025] .

[0026] in is the missile's yaw angle, is the horizontal resolution of the image, It is the horizontal offset of the rotation axis.

[0027] The coordinates of the rotation axis are obtained according to the coordinates of the center point of the image and the vertical offset and horizontal offset of the rotation axis. for:

[0028] ;

[0029] .

[0030] in is the horizontal coordinate of the center point of the image, is the ordinate of the center point of the image, is the horizontal coordinate of the rotation axis, is the ordinate of the rotation axis.

[0031] According to the position distribution coordinates of the key pixel points of the target object in the key frame image and the rotation axis coordinates, the distance between the key pixel points and the rotation axis is obtained. for:

[0032] .

[0033] in The target object The horizontal coordinates of key pixels, The target object The vertical coordinate of the key pixel point.

[0034] Further, the method of obtaining flight environment parameters through meteorological sensors, wherein the flight environment parameters include atmospheric density and wind speed, obtaining missile flight speed, and obtaining missile aerodynamic lift and missile lateral force through airflow disturbance analysis according to the missile flight speed, spin state data and the flight environment parameters includes:

[0035] The atmospheric density and wind speed are obtained by meteorological sensors to obtain the missile flight speed, and the missile angle of attack is obtained according to the missile pitch angle, missile flight speed and wind speed. for:

[0036] .

[0037] in is the wind speed, is the missile flight speed, is the missile's angle of attack.

[0038] The corresponding missile lift coefficient in the preset lift coefficient table is obtained according to the missile attack angle, and the preset lift coefficient table is a mapping table of missile lift coefficients at different attack angles obtained according to wind tunnel experiments or CFD simulations; the basic parameters of the missile are obtained, including the frontal windward area of ​​the missile and the side windward area of ​​the missile, and the aerodynamic lift of the missile is obtained according to the atmospheric density, the missile flight speed, the missile windward area and the missile lift coefficient. for:

[0039] .

[0040] in is the atmospheric density, is the frontal area of ​​the missile, is the missile lift coefficient, It is the aerodynamic lift of the missile.

[0041] The corresponding missile lateral force coefficient in the preset lateral force coefficient table is obtained according to the yaw angle, and the preset lateral force coefficient table is a mapping table of the missile at different yaw angles and the lateral force coefficient obtained according to wind tunnel experiments or CFD simulations; the missile lateral force is obtained according to the atmospheric density, wind speed, missile windward area and missile lateral force coefficient. for:

[0042] .

[0043] in is the wind speed, is the missile's lateral frontal area, is the missile side force coefficient, is the lateral force of the missile.

[0044] Furthermore, the method for obtaining the field of view offset by performing aerodynamic distortion analysis based on the missile aerodynamic lift, the missile lateral force and the missile attitude angle comprises:

[0045] According to the missile aerodynamic lift and missile mass, the vertical fluctuation displacement of the missile at each moment is obtained. for:

[0046] .

[0047] in is the missile's aerodynamic lift, is the missile mass, is the acceleration due to gravity, is the inter-frame time interval, is the vertical fluctuation displacement.

[0048] According to the lateral force and mass of the missile, the horizontal fluctuation displacement of the missile at each moment is obtained. for:

[0049] .

[0050] in is the missile's lateral force, is the horizontal fluctuation displacement.

[0051] The horizontal field of view offset is obtained by pixel conversion through the camera optical system according to the horizontal fluctuation displacement for:

[0052] .

[0053] in is the horizontal resolution of the image, is the distance between the missile and the target object, is the camera horizontal pixel projection scale factor, which is the ratio of the lateral size of the camera sensor to the camera focal length, is the horizontal field of view offset.

[0054] The vertical field of view offset is obtained by pixel conversion through the camera optical system according to the vertical fluctuation displacement and the missile pitch angle. for:

[0055] .

[0056] in is the vertical resolution of the image, is the missile pitch angle, is the camera vertical pixel projection scale factor, which is the ratio of the vertical size of the camera sensor to the camera focal length, is the vertical field of view offset.

[0057] Furthermore, the method of superimposing the field of view offset and the rotation blur displacement to obtain a joint distortion amount, and correcting the first image data of the target object through geometric distortion correction according to the joint distortion amount to obtain the second image data of the target object includes:

[0058] The rotation blur displacement is decomposed into the horizontal component of the rotation blur displacement and the vertical component of the rotation blur displacement.

[0059] The horizontal component of the rotational blur displacement is superimposed with the horizontal field of view offset to obtain the horizontal joint distortion. for:

[0060] .

[0061] in is the rotation blur displacement, The image is The horizontal component of the rotational blur displacement at time , is the horizontal joint distortion.

[0062] The vertical component of the rotational blur displacement is superimposed with the vertical field of view offset to obtain the vertical joint distortion:

[0063] .

[0064] in The image is The vertical component of the rotational blur displacement at time , is the vertical joint distortion.

[0065] The horizontal joint distortion amount and the vertical joint distortion amount of the pixel points at each image frame moment are summarized to obtain a joint distortion amount matrix, and the first image data of the target object is corrected by a geometric distortion correction model according to the joint distortion amount matrix to obtain the second image data of the target object.

[0066] Furthermore, the geometric distortion correction model is a mapping relationship between the first image data of the target object and the second image data of the target object established by inverse compensation and bilinear interpolation method, and the method of correcting the first image data of the target object by the geometric distortion correction model according to the joint distortion matrix to obtain the second image data of the target object includes:

[0067] According to the joint distortion matrix, the corrected pixel points corresponding to each pixel point in the first image data are obtained by inverse compensation:

[0068] .

[0069] in It's a pixel The horizontal joint distortion of It's a pixel The vertical joint distortion of It's a pixel The corresponding corrected pixel.

[0070] Determine whether the correction pixel point is located at an integer coordinate point. If the correction pixel point is located at an integer coordinate point, use the correction pixel point as the second image data. Otherwise, correct the pixel value of the correction pixel point through a bilinear interpolation algorithm to obtain a corrected pixel point, and use the corrected pixel point as the second image data.

[0071] Furthermore, the method of correcting the pixel value of the correction pixel point by using a bilinear interpolation algorithm to obtain the correction pixel point includes:

[0072] Get the first corrected pixel The pixel value of the corrected pixel is obtained by weighted averaging the pixel values ​​of the four adjacent integer pixels:

[0073] = .

[0074] in are the pixel values ​​of the four adjacent integer pixels, They are the first correction pixels The integer pixel coordinates of its upper left corner The horizontal and vertical distances.

[0075] Furthermore, the method for acquiring multi-scale features of a target object in the second image data of the target object includes:

[0076] According to the second image data of the target object, a multi-scale image pyramid of the target object is obtained by gradually reducing the full-resolution image level to the low-resolution image level.

[0077] The multi-scale features of the target object are obtained by extracting features from images of each scale in the multi-scale image pyramid of the target object through a scale-invariant feature transformation algorithm.

[0078] Based on the same inventive concept, the present invention provides an image recognition missile tracking intelligent processing system, the system comprising: a spin distortion processing module, an airflow disturbance analysis module, a joint distortion correction module, and an image monitoring and early warning module. The spin distortion processing module and the airflow disturbance analysis module are respectively connected to the joint distortion correction module, and the joint distortion correction module is connected to the image monitoring and early warning module.

[0079] The spin distortion processing module is used to obtain the missile's spin state data through an inertial measurement unit, the spin state data including the spin speed and the missile attitude angle, obtain the first image data of the target object collected during the missile's flight through a camera, and perform spin distortion analysis on the first image data of the target object according to the spin speed to obtain the rotational fuzzy displacement.

[0080] The airflow disturbance analysis module is used to obtain flight environment parameters through meteorological sensors, and the flight environment parameters include atmospheric density and wind speed, obtain the missile flight speed, and obtain the missile aerodynamic lift and missile lateral force through airflow disturbance analysis according to the missile flight speed, spin state data and the flight environment parameters.

[0081] The joint distortion correction module is used to perform aerodynamic distortion analysis according to the missile aerodynamic lift, the missile lateral force and the missile attitude angle to obtain the field of view offset, superimpose the field of view offset with the rotational blur displacement to obtain the joint distortion amount, and correct the first image data of the target object through geometric distortion correction according to the joint distortion amount to obtain the second image data of the target object.

[0082] The image monitoring and early warning module is used to obtain multi-scale features of the target object in the second image data of the target object, identify the image position of the target object through a fast feature point matching algorithm based on the multi-scale features of the target object, compare the image position of the target object with the expected position on the preset flight trajectory, and trigger the early warning mechanism and execute the preset missile flight adjustment strategy when it deviates from the expected position range threshold.

[0083] Beneficial Effects

[0084] Compared with the prior art, the present invention has the following beneficial effects: the rotational blur displacement and the field of view offset are obtained by spin state analysis and airflow disturbance analysis and are superimposed to form a joint distortion; image correction is achieved by performing geometric distortion correction through inverse compensation and bilinear difference method, thereby achieving more accurate intelligent tracking of the missile in the spin state. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0086] Figure 1 This is a flowchart of an intelligent processing method for missile tracking based on image recognition according to Embodiment 1 of the present invention;

[0087] Figure 2This is a schematic diagram of the module composition of an image recognition missile tracking intelligent processing system according to Embodiment 2 of the present invention. DETAILED DESCRIPTION

[0088] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0089] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0090] Example 1: Figure 1 As shown, this embodiment provides an image recognition missile tracking intelligent processing method, the method comprising:

[0091] S1. Obtaining the missile's spin state data through an inertial measurement unit, wherein the spin state data includes a spin speed and a missile attitude angle, obtaining the first image data of the target object collected during the missile's flight through a camera, and performing a spin distortion analysis on the first image data of the target object according to the spin speed to obtain a rotational fuzzy displacement; the inertial measurement unit is a sensor group integrating a gyroscope, an accelerometer, and a magnetometer, and is used to monitor the missile's spin speed, pitch angle, yaw angle and other rolling motion parameters.

[0092] The missile pitch angle and the missile yaw angle are calculated by detecting the acceleration change of the missile in the longitudinal direction through the inertial measurement unit. The missile pitch angle is obtained according to the acceleration of the missile in the horizontal forward direction, the horizontal lateral direction and the vertical direction detected by the inertial measurement unit:

[0093] .

[0094] in is the horizontal acceleration in the forward direction, is the horizontal lateral acceleration, is the vertical acceleration, is the missile pitch angle.

[0095] According to the missile's horizontal lateral acceleration and vertical acceleration, the missile's yaw angle is:

[0096] .

[0097] in is the missile's yaw angle.

[0098] S2. Acquire flight environment parameters through meteorological sensors, wherein the flight environment parameters include atmospheric density and wind speed, acquire the missile flight speed, and obtain the missile aerodynamic lift and missile lateral force through airflow disturbance analysis according to the missile flight speed, spin state data and the flight environment parameters.

[0099] S3. Perform aerodynamic distortion analysis according to the missile's aerodynamic lift, the missile's lateral force, and the missile's attitude angle to obtain a field of view offset, superimpose the field of view offset with the rotational blur displacement to obtain a joint distortion, and correct the first image data of the target object through geometric distortion correction according to the joint distortion to obtain second image data of the target object.

[0100] S4. Acquire multi-scale features of the target object in the second image data of the target object, identify the image position of the target object through a fast feature point matching algorithm based on the multi-scale features of the target object, compare the image position of the target object with the expected position on the preset flight trajectory, and trigger an early warning mechanism and execute a preset missile flight adjustment strategy when it deviates from the expected position range threshold.

[0101] It should be noted that the method of performing spin distortion analysis on the first image data of the target object according to the spin speed to obtain the rotation blur displacement includes:

[0102] S11. Obtain the rotation angle of the image acquired at each moment during the missile flight according to the spin speed for:

[0103] .

[0104] in is the spin speed, It's time. It's in time The image rotation angle at time, When the rotation angle exceeds 360 degrees, the angle is reset to zero and the calculation starts again.

[0105] For example: missile spin speed Degrees / second, time seconds, the image rotation angle .

[0106] S12, extracting single-frame images at each moment of the first image data, and extracting edge contours and texture features of the target object in the single-frame images through an edge detection algorithm to obtain position distribution coordinates of key pixel points of the target object in the image.

[0107] S13, determining the rotation axis coordinates according to the missile attitude angle, and obtaining the distance between the key pixel point and the rotation axis according to the position distribution coordinates of the key pixel point of the target object in the image and the rotation axis coordinates.

[0108] S14, obtain the rotational blur displacement according to the distance between the key pixel point and the rotation axis and the rotation speed for:

[0109] .

[0110] in It is The distance between the key pixel point of the target object and the rotation axis, is the rotation blur displacement.

[0111] It should be noted that the missile attitude angle includes the missile pitch angle and the missile yaw angle. The method of determining the rotation axis coordinates according to the missile attitude angle and obtaining the distance between the key pixel point and the rotation axis according to the position distribution coordinates of the key pixel point of the target object in the image and the rotation axis coordinates includes:

[0112] S131. Obtain the vertical offset of the rotation axis according to the missile pitch angle and the image vertical resolution. for:

[0113] .

[0114] in is the missile pitch angle, is the image vertical resolution, It is the vertical offset of the rotation axis.

[0115] S132: Obtain the horizontal offset of the rotation axis according to the missile yaw angle and the image horizontal resolution for:

[0116] .

[0117] in is the missile's yaw angle, is the horizontal resolution of the image, It is the horizontal offset of the rotation axis.

[0118] For example: Assume the image resolution is Pixels, Pixels, if the missile pitch angle , missile yaw angle When , the vertical offset can be obtained Pixels, horizontal offset Pixel.

[0119] S133, obtaining the rotation axis coordinates according to the image center coordinates and the vertical offset of the rotation axis and the horizontal offset of the rotation axis for:

[0120] ;

[0121] .

[0122] in is the horizontal coordinate of the center point of the image, is the ordinate of the center point of the image, is the horizontal coordinate of the rotation axis, is the ordinate of the rotation axis.

[0123] When there is no influence of the attitude angle, the rotation axis is the center point of the image. When there is an influence of the attitude angle, the rotation axis will be offset under the influence of the attitude angle. Therefore, the coordinates of the image center point are corrected by calculating the offset to obtain the accurate position of the rotation axis.

[0124] For example: When the image resolution is Pixels, pixels, the coordinates of the center point of the image are , the horizontal coordinate of the rotation axis Pixels, rotation axis ordinate pixels, the rotation axis coordinates are .

[0125] S134, obtaining the distance between the key pixel point and the rotation axis according to the position distribution coordinates of the key pixel point of the target object in the key frame image and the rotation axis coordinates for:

[0126] .

[0127] in The target object The horizontal coordinates of key pixels, The target object The vertical coordinate of the key pixel point.

[0128] For example: If the key pixel points of the target object have , , then the distance between the first key pixel of the target object and the rotation axis is , the distance between the second key pixel of the target object and the rotation axis .

[0129] It should be noted that the method of obtaining flight environment parameters through meteorological sensors, wherein the flight environment parameters include atmospheric density and wind speed, obtaining missile flight speed, and obtaining missile aerodynamic lift and missile lateral force through airflow disturbance analysis according to the missile flight speed, spin state data and the flight environment parameters includes:

[0130] S21, obtain the atmospheric density and wind speed through the meteorological sensor, obtain the missile flight speed, and obtain the missile angle of attack according to the missile pitch angle, missile flight speed and wind speed. for:

[0131] .

[0132] in is the wind speed, is the missile flight speed, is the missile's angle of attack.

[0133] The missile's angle of attack is the angle between the missile's longitudinal axis and the airflow relative to the missile's flight direction. , wind speed The missile's flight speed is When the missile angle of attack .

[0134] S22, according to the missile's angle of attack, obtain the corresponding missile lift coefficient in the preset lift coefficient table, the preset lift coefficient table is a mapping table of missile lift coefficients at different angles of attack obtained by wind tunnel experiments or CFD simulations; obtain the missile's basic parameters including the missile's front windward area and the missile's side windward area, and obtain the missile's aerodynamic lift according to the atmospheric density, the missile's flight speed, the missile's windward area and the missile's lift coefficient. for:

[0135] .

[0136] in is the atmospheric density, is the frontal area of ​​the missile, is the missile lift coefficient, It is the aerodynamic lift of the missile.

[0137] For example: When the atmospheric density , missile frontal area , missile lift coefficient , then the missile aerodynamic lift .

[0138] S23, according to the yaw angle, obtain the corresponding missile lateral force coefficient in the preset lateral force coefficient table, wherein the preset lateral force coefficient table is a mapping table of the missile at different yaw angles and the lateral force coefficient obtained by wind tunnel experiment or CFD simulation; according to the atmospheric density, wind speed, missile windward area and missile lateral force coefficient, obtain the missile lateral force for:

[0139] .

[0140] in is the wind speed, is the missile's lateral frontal area, is the missile side force coefficient, is the lateral force of the missile.

[0141] For example: When the atmospheric density , missile surface windward area , missile lateral force coefficient , then the missile lateral force .

[0142] It should be noted that the method for obtaining the field of view offset by performing aerodynamic distortion analysis based on the missile aerodynamic lift, the missile lateral force and the missile attitude angle includes:

[0143] S31. Obtain the vertical fluctuation displacement of the missile at each moment based on the missile's aerodynamic lift and missile mass For:

[0144] .

[0145] in is the missile's aerodynamic lift, is the missile mass, is the acceleration due to gravity, is the inter-frame time interval, is the vertical fluctuation displacement.

[0146] S32. Obtain the horizontal fluctuation displacement of the missile at each moment based on the lateral force and mass of the missile for:

[0147] .

[0148] in is the missile's lateral force, is the horizontal fluctuation displacement.

[0149] For example: Assuming the missile mass kg,Inter-frame time interval seconds, the vertical fluctuation displacement Horizontal Fluctuation Displacement .

[0150] S33, performing pixel conversion through the camera optical system according to the horizontal fluctuation displacement to obtain the horizontal field of view offset for:

[0151] .

[0152] in is the horizontal resolution of the image, is the distance between the missile and the target object, is the camera horizontal pixel projection scale factor, which is the ratio of the lateral size of the camera sensor to the camera focal length, is the horizontal field of view offset.

[0153] S34, according to the vertical fluctuation displacement and the missile pitch angle, the vertical field of view offset is obtained by pixel conversion through the camera optical system. for:

[0154] .

[0155] in is the vertical resolution of the image, is the missile pitch angle, is the camera vertical pixel projection scale factor, which is the ratio of the vertical size of the camera sensor to the camera focal length, is the vertical field of view offset.

[0156] For example: If the camera horizontal pixel projection scale factor , vertical pixel projection scale factor , the distance between the missile and the target object , then the horizontal field of view offset Pixels, vertical field of view offset Pixel.

[0157] The vertical field of view offset is not only the offset caused by the missile's aerodynamic lift, but also the vertical offset caused by the missile's pitch angle. Therefore, when converting the vertical fluctuation displacement into pixels, the vertical offset caused by the missile's pitch angle needs to be added to form the final vertical field of view offset.

[0158] It should be noted that the method of superimposing the field of view offset and the rotation blur displacement to obtain a joint distortion amount, and correcting the first image data of the target object through geometric distortion correction according to the joint distortion amount to obtain the second image data of the target object includes:

[0159] S35. Decompose the rotational blur displacement to obtain a horizontal component of the rotational blur displacement and a vertical component of the rotational blur displacement.

[0160] S36, superimposing the horizontal component of the rotational blur displacement and the horizontal field of view offset to obtain a horizontal joint distortion for:

[0161] .

[0162] in is the rotation blur displacement, The image is The horizontal component of the rotational blur displacement at time , is the horizontal joint distortion.

[0163] S37, superimposing the vertical component of the rotational blur displacement and the vertical field of view offset to obtain a vertical joint distortion amount:

[0164] .

[0165] in The image is The vertical component of the rotational blur displacement at time , is the vertical joint distortion.

[0166] For example: According to the above example, when the time Seconds, the image rotates by an angle , then the horizontal joint distortion Pixel, vertical joint distortion Pixels. That is, the key pixel in this example The joint distortion is ( ), indicating that the point is offset to the left Pixels, offset upward Pixel.

[0167] S38, summarizing the horizontal joint distortion and the vertical joint distortion of the pixel points at each image frame moment to obtain a joint distortion matrix, and correcting the first image data of the target object through a geometric distortion correction model according to the joint distortion matrix to obtain the second image data of the target object.

[0168] It should be noted that the geometric distortion correction model is a mapping relationship between the first image data of the target object and the second image data of the target object established by inverse compensation and bilinear interpolation method. The method of correcting the first image data of the target object by the geometric distortion correction model according to the joint distortion matrix to obtain the second image data of the target object includes:

[0169] S381, obtaining the corrected pixel points corresponding to each pixel point in the first image data by inverse compensation according to the joint distortion matrix:

[0170] .

[0171] in It's a pixel The horizontal joint distortion of It's a pixel The vertical joint distortion of It's a pixel The corresponding corrected pixel.

[0172] For example: When , , , then the corresponding correction pixel point .

[0173] S382, determine whether the correction pixel point is located at an integer coordinate point, if the correction pixel point is located at an integer coordinate point, use the correction pixel point as the second image data, otherwise use the bilinear interpolation algorithm to correct the pixel value of the correction pixel point to obtain the correction pixel point, and use the correction pixel point as the second image data. Non-integer coordinate pixel points do not directly exist in the pixel grid of the image. If the correction pixel point is not located at an integer coordinate point, it will cause the image to produce a sawtooth effect or discontinuous jump, thereby causing the edge of the image to be blurred and distorted. The pixel value at this position is calculated by the bilinear interpolation algorithm to further correct the discontinuity or distortion in the reverse compensation process.

[0174] It should be noted that the method of correcting the pixel value of the correction pixel point by using the bilinear interpolation algorithm to obtain the correction pixel point includes:

[0175] S3821, obtaining the first correction pixel point The pixel value of the corrected pixel is obtained by weighted averaging the pixel values ​​of the four adjacent integer pixels:

[0176] = .

[0177] in are the pixel values ​​of the four adjacent integer pixels, They are the first correction pixels The integer pixel coordinates of its upper left corner The horizontal and vertical distances.

[0178] For example: When the first correction pixel , since it is located at a non-integer coordinate, the pixel values ​​of the four neighboring integer pixels are selected and weighted averaged. The four neighboring integer pixels are , , , , , , if their intensity values ​​are , then the corrected pixel value is .

[0179] The pixel values ​​of non-integer pixels are obtained through the bilinear interpolation algorithm to ensure the continuity and smoothness of the image.

[0180] It should be noted that the method for obtaining multi-scale features of the target object in the second image data of the target object includes:

[0181] S41 , according to the second image data of the target object, gradually reducing the full-resolution image level to the low-resolution image level to obtain a multi-scale image pyramid of the target object.

[0182] S42. Extracting features of each scale image in the multi-scale image pyramid of the target object by using a scale-invariant feature transformation algorithm to obtain multi-scale features of the target object.

[0183] Embodiment 2: Based on the same inventive concept, Figure 2 As shown, this embodiment provides an image recognition missile tracking intelligent processing system, the system includes: a spin distortion processing module, an airflow disturbance analysis module, a joint distortion correction module, and an image monitoring and early warning module. The spin distortion processing module and the airflow disturbance analysis module are respectively connected to the joint distortion correction module, and the joint distortion correction module is connected to the image monitoring and early warning module.

[0184] The spin distortion processing module is used to obtain the missile's spin state data through an inertial measurement unit, the spin state data including the spin speed and the missile attitude angle, obtain the first image data of the target object collected during the missile's flight through a camera, and perform spin distortion analysis on the first image data of the target object according to the spin speed to obtain the rotational fuzzy displacement.

[0185] The airflow disturbance analysis module is used to obtain flight environment parameters through meteorological sensors, and the flight environment parameters include atmospheric density and wind speed, obtain the missile flight speed, and obtain the missile aerodynamic lift and missile lateral force through airflow disturbance analysis according to the missile flight speed, spin state data and the flight environment parameters.

[0186] The joint distortion correction module is used to perform aerodynamic distortion analysis according to the missile aerodynamic lift, the missile lateral force and the missile attitude angle to obtain the field of view offset, superimpose the field of view offset with the rotational blur displacement to obtain the joint distortion amount, and correct the first image data of the target object through geometric distortion correction according to the joint distortion amount to obtain the second image data of the target object.

[0187] The image monitoring and early warning module is used to obtain multi-scale features of the target object in the second image data of the target object, identify the image position of the target object through a fast feature point matching algorithm based on the multi-scale features of the target object, compare the image position of the target object with the expected position on the preset flight trajectory, and trigger the early warning mechanism and execute the preset missile flight adjustment strategy when it deviates from the expected position range threshold.

[0188] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in Embodiment 1 of the method, and will not be elaborated here.

[0189] Finally, it should be noted that: Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent processing method for missile tracking based on image recognition, characterized in that: The method comprises: The missile's spin state data is obtained through an inertial measurement unit, wherein the spin state data includes a spin speed and a missile attitude angle, and first image data of a target object collected during the missile's flight is obtained through a camera. A spin distortion analysis is performed on the first image data of the target object according to the spin speed to obtain a rotational fuzzy displacement; Acquire flight environment parameters through meteorological sensors, the flight environment parameters including atmospheric density and wind speed, acquire missile flight speed, and obtain missile aerodynamic lift and missile lateral force through airflow disturbance analysis based on missile flight speed, spin state data and the flight environment parameters; Performing aerodynamic distortion analysis according to the missile's aerodynamic lift, the missile's lateral force, and the missile's attitude angle to obtain a field of view offset, superimposing the field of view offset with the rotational blur displacement to obtain a joint distortion amount, and correcting the first image data of the target object through geometric distortion correction according to the joint distortion amount to obtain second image data of the target object; The multi-scale features of the target object in the second image data of the target object are obtained, and the image position of the target object is identified through a fast feature point matching algorithm based on the multi-scale features of the target object. The image position of the target object is compared with the expected position on the preset flight trajectory. When it deviates from the expected position range threshold, the early warning mechanism is triggered and the preset missile flight adjustment strategy is executed.

2. The method for intelligent missile tracking based on image recognition according to claim 1, characterized in that: The method for performing spin distortion analysis on the first image data of the target object according to the spin speed to obtain the rotation blur displacement comprises: According to the spin speed, the image rotation angle θ(t) obtained at each moment during the missile flight is: θ(t)=(ω·t)mod360°; Where ω is the spin speed, t is the time, θ(t) is the image rotation angle at time t, and mod360° is when the rotation angle exceeds 360 degrees, the angle is reset to zero and the calculation is restarted; Extracting single-frame images at each moment of the first image data, and extracting edge contours and texture features of the target object in the single-frame images by an edge detection algorithm to obtain position distribution coordinates of key pixel points of the target object in the image; Determine the rotation axis coordinates according to the missile attitude angle, and obtain the distance between the key pixel point and the rotation axis according to the position distribution coordinates of the key pixel point of the target object in the image and the rotation axis coordinates; The rotational blur displacement R is obtained according to the distance between the key pixel point and the rotation axis and the rotation speed. i for: R i =r i ·θ(t); where r i is the distance between the key pixel of the i-th target object and the rotation axis, R i is the rotation blur displacement.

3. The missile tracking intelligent processing method based on image recognition according to claim 2 is characterized in that: The missile attitude angle includes a missile pitch angle and a missile yaw angle. The method of determining the rotation axis coordinates according to the missile attitude angle and obtaining the distance between the key pixel point and the rotation axis according to the position distribution coordinates of the key pixel point of the target object in the image and the rotation axis coordinates includes: According to the missile pitch angle and the image vertical resolution, the vertical offset Δy of the rotation axis is obtained as follows: Δy=H·tanα f ; where α f is the missile pitch angle, H is the image vertical resolution, and Δy is the vertical offset of the rotation axis; According to the missile yaw angle and the image horizontal resolution, the horizontal offset Δx of the rotation axis is obtained as: Δx=W·tanα p ; where α p is the missile yaw angle, W is the image horizontal resolution, and Δx is the horizontal offset of the rotation axis; According to the coordinates of the center point of the image and the vertical offset and horizontal offset of the rotation axis, the coordinates of the rotation axis (x0, y0) are obtained as follows: x0=x c +Δx; y0=y c +Δy; where x c is the horizontal coordinate of the center point of the image, y c is the ordinate of the center of the image, x0 is the abscissa of the rotation axis, and y0 is the ordinate of the rotation axis; According to the position distribution coordinates of the key pixel points of the target object in the key frame image and the rotation axis coordinates, the distance r between the key pixel points and the rotation axis is obtained. i for: where x i is the horizontal coordinate of the i-th key pixel point of the target object, y i is the ordinate of the i-th key pixel of the target object.

4. The missile tracking intelligent processing method based on image recognition according to claim 3 is characterized in that: The method of obtaining flight environment parameters by means of meteorological sensors, wherein the flight environment parameters include atmospheric density and wind speed, obtaining the missile flight speed, and obtaining the missile aerodynamic lift and the missile lateral force by airflow disturbance analysis according to the missile flight speed, spin state data and the flight environment parameters comprises: The atmospheric density and wind speed are obtained by the meteorological sensor to obtain the missile flight speed, and the missile angle of attack α is obtained according to the missile pitch angle, missile flight speed and wind speed. g for: Where V w is the wind speed, V m is the missile flight speed, α g is the missile angle of attack, α f is the missile pitch angle; According to the missile's angle of attack, the corresponding missile lift coefficient in the preset lift coefficient table is obtained, and the basic missile parameters including the missile's front windward area and the missile's side windward area are obtained. According to the atmospheric density, missile flight speed, missile windward area and missile lift coefficient, the missile's aerodynamic lift F is obtained. L for: where ρ is the atmospheric density, S z is the frontal windward area of ​​the missile, C L is the missile lift coefficient, F L is the missile's aerodynamic lift, V m is the missile flight speed; The corresponding missile lateral force coefficient in the preset lateral force coefficient table is obtained according to the yaw angle, and the missile lateral force F is obtained according to the atmospheric density, wind speed, missile frontal area and missile lateral force coefficient. y for: Where V w is the wind speed, S c is the missile's lateral frontal area, C y is the missile side force coefficient, F y is the lateral force of the missile.

5. The image recognition missile tracking intelligent processing method according to claim 4 is characterized in that: The method for obtaining the field of view offset by performing aerodynamic distortion analysis based on the missile aerodynamic lift, the missile lateral force and the missile attitude angle comprises: According to the missile aerodynamic lift and missile mass, the missile's vertical fluctuation displacement S at each moment is obtained. L for: where F L is the missile aerodynamic lift, m is the missile mass, g is the gravitational acceleration, t′ is the time interval between frames, S L is the vertical wave displacement; According to the lateral force and mass of the missile, the horizontal fluctuation displacement S of the missile at each moment is obtained. y for: where F y is the missile's lateral force, S y is the horizontal fluctuation displacement; According to the horizontal fluctuation displacement, the horizontal field of view offset W is obtained by pixel conversion through the camera optical system. P for: Where W is the image horizontal resolution, D is the distance between the missile and the target object, and K L is the camera horizontal pixel projection scale factor, which is the ratio of the lateral size of the camera sensor to the camera focal length, W P is the horizontal field of view offset; According to the vertical fluctuation displacement and the missile pitch angle, the vertical field of view offset H is obtained by pixel conversion through the camera optical system. p for: Where H is the vertical resolution of the image, α f is the missile pitch angle, K w is the camera vertical pixel projection scale factor, which is the ratio of the vertical size of the camera sensor to the camera focal length, H p is the vertical field of view offset.

6. The image recognition missile tracking intelligent processing method according to claim 5 is characterized in that: The method of superimposing the field of view offset and the rotation blur displacement to obtain a joint distortion amount, and correcting the first image data of the target object through geometric distortion correction according to the joint distortion amount to obtain the second image data of the target object includes: Decomposing the rotation blur displacement to obtain the horizontal component of the rotation blur displacement and the vertical component of the rotation blur displacement; The horizontal component of the rotational blur displacement is superimposed with the horizontal field of view offset to obtain the horizontal joint distortion J. W for: J W =W P +R i cos(θ(t)); Where R i is the rotational blur displacement, R i cos(θ(t)) is the horizontal component of the rotational blur displacement of the image at time t, J W is the horizontal joint distortion; The vertical component of the rotational blur displacement is superimposed with the vertical field of view offset to obtain the vertical joint distortion: J H =H P +R i sin(θ(t)); Where R i sin(θ(t)) is the vertical component of the rotational blur displacement of the image at time t, J H is the vertical joint distortion; The horizontal joint distortion amount and the vertical joint distortion amount of the pixel points at each image frame moment are summarized to obtain a joint distortion amount matrix, and the first image data of the target object is corrected by a geometric distortion correction model according to the joint distortion amount matrix to obtain the second image data of the target object.

7. The image recognition missile tracking intelligent processing method according to claim 6 is characterized in that: The geometric distortion correction model is a mapping relationship between the first image data of the target object and the second image data of the target object established by inverse compensation and bilinear interpolation method. The method of correcting the first image data of the target object by the geometric distortion correction model according to the joint distortion matrix to obtain the second image data of the target object includes: According to the joint distortion matrix, the corrected pixel points corresponding to each pixel point in the first image data are obtained by inverse compensation: (x′,y′)=(xJ W (x,y),yJ H (x,y)); Among them J W (x,y) is the horizontal joint distortion of the pixel point (x,y), J H (x,y) is the vertical joint distortion of pixel point (x,y), (x′,y′) is the corrected pixel point corresponding to pixel point (x,y); Determine whether the correction pixel point is located at an integer coordinate point. If the correction pixel point is located at an integer coordinate point, use the correction pixel point as the second image data. Otherwise, correct the pixel value of the correction pixel point through a bilinear interpolation algorithm to obtain a corrected pixel point, and use the corrected pixel point as the second image data.

8. The image recognition missile tracking intelligent processing method according to claim 7 is characterized in that: The method of correcting the pixel value of the correction pixel point by using a bilinear interpolation algorithm to obtain the correction pixel point includes: The pixel values ​​of the four integer pixels adjacent to the first correction pixel (x′, y′) are obtained and weighted averaged to obtain the pixel value of the correction pixel: I(x′,y′)=(1-Δx)(1-Δy)I(x1,y1)+Δx(1-Δy)I(x2,y1)+(1-Δx)ΔyI(x1,y2)+ΔxΔyI(x2,y2); Where I(x1,y1), I(x2,y1), I(x1,y2), I(x2,y2) are the pixel values ​​of the four adjacent integer pixels, and Δx, Δy are the horizontal and vertical distances between the first correction pixel point (x′,y′) and the coordinates of its upper left integer pixel point (x1,y1), respectively.

9. The missile tracking intelligent processing method based on image recognition according to claim 1 is characterized in that: The method for obtaining multi-scale features of a target object in the second image data of the target object comprises: According to the second image data of the target object, a multi-scale image pyramid of the target object is obtained by gradually reducing the full-resolution image level to the low-resolution image level; The multi-scale features of the target object are obtained by extracting features from images of each scale in the multi-scale image pyramid of the target object through a scale-invariant feature transformation algorithm.

10. An image recognition missile tracking intelligent processing system, used to execute the method according to any one of claims 1 to 9, characterized in that: The system comprises: a spin distortion processing module, an airflow disturbance analysis module, a joint distortion correction module, and an image monitoring and early warning module. The spin distortion processing module and the airflow disturbance analysis module are respectively connected to the joint distortion correction module, and the joint distortion correction module is connected to the image monitoring and early warning module. The spin distortion processing module is used to obtain the spin state data of the missile through the inertial measurement unit, the spin state data including the spin speed and the missile attitude angle, obtain the first image data of the target object collected during the flight of the missile through the camera, and perform spin distortion analysis on the first image data of the target object according to the spin speed to obtain the rotation fuzzy displacement; The airflow disturbance analysis module is used to obtain flight environment parameters through meteorological sensors, the flight environment parameters include atmospheric density and wind speed, obtain the missile flight speed, and obtain the missile aerodynamic lift and missile lateral force through airflow disturbance analysis according to the missile flight speed, spin state data and the flight environment parameters; The joint distortion correction module is used to perform aerodynamic distortion analysis according to the missile aerodynamic lift, the missile lateral force and the missile attitude angle to obtain a field of view offset, superimpose the field of view offset with the rotational blur displacement to obtain a joint distortion amount, and correct the first image data of the target object through geometric distortion correction according to the joint distortion amount to obtain the second image data of the target object; The image monitoring and early warning module is used to obtain multi-scale features of the target object in the second image data of the target object, identify the image position of the target object through a fast feature point matching algorithm based on the multi-scale features of the target object, compare the image position of the target object with the expected position on the preset flight trajectory, and trigger the early warning mechanism and execute the preset missile flight adjustment strategy when it deviates from the expected position range threshold.

Images