An Inertial Vision Landing Navigation Method for Rotary-wing UAVs

Abstract

The present invention provides a method for inertial vision landing navigation of a rotor unmanned aerial vehicle. The specific steps of the navigation method are as follows: First step, cooperative target coordinate system error modeling: Convert the displacement amount resolved by the inertial navigation system on the geographic coordinate system to the cooperative target coordinate system, and use the C n a attitude transfer matrix for conversion. Then, perform error modeling under the cooperative target coordinate system; Second step, fuse the inertial navigation and vision navigation information: Fuse the inertial navigation and vision navigation information through the Kalman filtering algorithm; Third step, system measurement noise matrix fitting method: Calculate the relative distance between the rotor unmanned aerial vehicle and the cooperative target, and fit the measurement noise matrix R k ; Fourth step, position error correction: That is, correct each error of the inertial navigation; The present invention realizes the need not to obtain the cooperative target position information in advance, solves the technical problem that the vision navigation accuracy changes with the distance between the unmanned aerial vehicle and the cooperative target, and improves the vision navigation accuracy of the unmanned aerial vehicle.

Description

Technical Field

[0001] The present invention belongs to the technical field of autonomous landing of unmanned aerial vehicles, and particularly relates to an inertial vision landing navigation method for a rotor unmanned aerial vehicle. Background Art

[0002] Traditional inertial vision integrated navigation models generally use a geographic coordinate system for modeling. This method relies on the complete error equation of inertial navigation and can stimulate various errors with the maneuver of the unmanned aerial vehicle in a large-scale scenario. In the vertical takeoff and landing scenario, the unmanned aerial vehicle flies directly towards the cooperative target, and the attitude error is unobservable. Moreover, the geographic coordinate system modeling needs to convert the relative pose information of visual navigation into the absolute position of longitude and latitude, and the position information of the cooperative target needs to be obtained in advance, which brings more inconvenience to the landing process. Summary of the Invention

[0003] The purpose of the present invention is to provide an inertial vision landing navigation method for a rotor unmanned aerial vehicle, which realizes the need not to obtain the position information of the cooperative target in advance, solves the technical problem that the visual navigation accuracy changes with the distance between the unmanned aerial vehicle and the cooperative target, and improves the visual navigation accuracy of the unmanned aerial vehicle.

[0004] The technical solution of the present invention is that the specific steps of an inertial vision landing navigation method for a rotor unmanned aerial vehicle are as follows:

[0005] The first step, error modeling in the cooperative target coordinate system:

[0006] Convert the displacement amount on the geographic coordinate system calculated by the inertial navigation system to the cooperative target coordinate system, and use the attitude transfer matrix for conversion, and then perform error modeling in the cooperative target coordinate system;

[0007] The cooperative target coordinate system error model includes the continuous state equation of the inertial vision integrated navigation system, the expression of the filtered state quantity, the system state transition matrix, the observation equation of the integrated navigation system, and the attitude transfer matrix;

[0008] The second step, fusion of inertial navigation and visual navigation information:

[0009] According to the error model obtained in the first step, fuse the inertial navigation and visual navigation information through the Kalman filtering algorithm;

[0010] The third step, system measurement noise matrix fitting method: Calculate the relative distance between the rotor unmanned aerial vehicle and the cooperative target, and fit the measurement noise matrix R k to obtain the k accurate value;

[0011] According to the four stages of the landing process of the rotor unmanned aerial vehicle, fit the measurement noise matrix R k in each stage through a quadratic curve as follows:

[0012] 1) When the relative distance to the cooperation target is greater than 200 m, R k = 100;

[0013] 2) When the relative distance to the cooperation target is greater than 10 m and less than 200 m, with x = 10 m as the axis of symmetry, (10 m, 5 m) as the vertex of the parabola, and (5 m, 100 m) as a point on the parabola, the parabola opens upward, and the fitted curve equation is: R k = 0.002631x 2 - 0.052632x + 5.26322;

[0014] 3) When the relative distance to the cooperation target is greater than 5 m and less than 10 m, with x = 10 m as the axis of symmetry, (10 m, 5 m) as the vertex of the parabola, and (200 m, 100 m) as a point on the parabola, the parabola opens upward, and the fitted curve equation is: R k = 3.8x 2 - 76x + 385;

[0015] 4) When the relative distance to the cooperation target is less than 5 m, R k = 100;

[0016] Fourth step, position error correction: That is, correct various errors of inertial navigation:

[0017] a) Position error correction:

[0018] The estimated value of the position error ΔP in the navigation coordinate system n is:

[0019]

[0020] In the formula, ΔP a is the estimated value of the position error in the cooperation target coordinate system, is the attitude transfer matrix from the cooperation target coordinate system to the navigation coordinate system, and convert its northward error and eastward error into radian coordinates:

[0021]

[0022] Perform position correction:

[0023]

[0024] b) Velocity error correction:

[0025] The estimated value of the velocity error in the navigation target coordinate system is:

[0026]

[0027] Perform velocity correction:

[0028]

[0029] In the first step, the cooperation target uses a QR code; the cooperation target coordinate system is: with the center point of the QR code logo as the origin o a ; x a axis is along the perpendicular line perpendicular to the plane of the QR code logo; y a axis is perpendicular to the x a axis within the plane of the QR code logo, and the upward direction is positive; z a axis is perpendicular to the x a axis within the plane of the QR code logo, and the rightward direction is positive; ox a y a z a constitute a right-handed coordinate system.

[0030] The Kalman filtering method is used to fuse inertial navigation and visual navigation.

[0031] In the first step, the error model of the cooperation target coordinate system is as follows:

[0032] The continuous state equation of the inertial-visual integrated navigation system is

[0033]

[0034] In the formula, F(t) is the state transition matrix of the continuous state equation at time t, is the system random noise vector at time t, is the system error variable to be estimated, is the derivative of the error variable,

[0035] The expression of the filtering state quantity X is

[0036] X T =(s x , v x , a x , s y , v y , a y , s z , v z , a z ) (2)

[0037] The position error s (unit: m), velocity error v (unit: m / s), and acceleration error a (unit: m / s2) in the cooperation target coordinate system. In the formula, the subscripts x, y, and z represent the three directions of the cooperation target coordinate system;

[0038] The system state transition matrix is

[0039]

[0040] Among them, F 1 , F 2 , F 3 represent the sub-transfer matrices in three directions of the coordinate system;

[0041] Among them,

[0042]

[0043] The observation equation of the integrated navigation system is,

[0044]

[0045] is the system observation variable, H(k) is the measurement matrix of the integrated navigation filter, is the observation noise matrix;

[0046] The observation information is:

[0047]

[0048] In the formula, δs represents the system position error, including the x, y, and z directions, s VIS represents the displacement between adjacent frames in the cooperative target coordinate system solved by the visual navigation system, s INS represents the displacement between adjacent frames in the cooperative target coordinate system of the inertial navigation system, s INS The calculation method is as follows:

[0049]

[0050] In the formula:

[0051] is the attitude transfer matrix from the geographic coordinate system to the cooperative target coordinate system, determined by the placement direction of the cooperative target, Lat is the latitude of the inertial navigation system, Hgt is the height of the inertial navigation system, Lon is the longitude of the inertial navigation system, the subscripts k and k - 1 represent the current moment and the previous moment, R m and R n represent the radius of the earth's meridian and the radius of the earth's prime vertical respectively,

[0052] Thus, the observation matrix H can be determined:

[0053] H(k) = [H 1 H 2 H 3 (9)

[0054] Among them:

[0055]

[0056] In the second step, the Kalman filtering calculation method is as follows:

[0057] a) One-step state prediction:

[0058]

[0059] b) One-step prediction mean square error:

[0060]

[0061] c) Filter gain:

[0062]

[0063] d) State estimation:

[0064]

[0065] e) Estimation mean square error:

[0066]

[0067] In the formula:

[0068] One-step prediction estimated state vector;

[0069] Φ k,k-1 : Discretized F matrix;

[0070] Optimal estimated state vector;

[0071] P k,k-1 : One-step prediction covariance matrix;

[0072] P k : Optimal estimated covariance matrix;

[0073] K k : Optimal gain matrix;

[0074] R k : Discretized measurement noise matrix;

[0075] Q: Discretized system noise covariance matrix;

[0076] H k is the measurement matrix; Z k is the observed quantity.

[0077] The subscript k represents the current moment;

[0078] In the third step, the UAV landing process is divided into the following four stages:

[0079] 1) The distance from the cooperation target is relatively far. At this time, the target imaging is not clear, and the visual navigation accuracy is poor.

[0080] 2) As the distance to the cooperation target approaches, the imaging of the cooperation target gradually becomes clear, and the visual navigation accuracy improves as the UAV descends.

[0081] 3) After the UAV descends to a certain height, as it approaches the cooperation target, the cooperation target gradually becomes incomplete in the camera imaging, and the visual navigation error rate increases.

[0082] 4) As the UAV descends to a certain height, the cooperation target is too small in the camera imaging, and the visual navigation accuracy is poor.

[0083] The beneficial effects of the present invention are as follows: The present invention measures the relative pose by identifying the cooperation target and fuses it with the continuous and high-frequency position information of inertial navigation to provide stable and reliable landing navigation information for the UAV. It realizes the online fitting and correction of the measurement noise matrix of the integrated navigation filter according to the characteristics of visual navigation, providing a very broad application prospect for the field of UAV autonomous landing; the present invention conducts research on the landing navigation method based on inertial vision combination for the problem of autonomous landing of the rotor UAV system under satellite denial conditions. First, it completes the error modeling method of the cooperation target coordinate system, and secondly, it completes the fitting method of the measurement noise matrix based on the distance to the cooperation target, supporting the autonomous landing ability of the rotor UAV under satellite denial conditions. Inertial / visual integrated navigation is an important field in the research of autonomous landing methods for rotor UAVs under satellite denial conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] The accompanying drawings included are used to provide a further understanding of the embodiments of the present invention, which form a part of the specification, are used to illustrate the embodiments of the present invention, and are used to explain the principles of the present invention together with the written description. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings without creative efforts based on these drawings.

[0085] Figure 1 It is a schematic diagram of the two-dimensional code coordinate system in an inertial vision landing navigation method for a rotor UAV of the present invention;

[0086] Figure 2 It is the curve of the measurement noise value in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0087] The technical solutions of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0088] The specific steps of an inertial vision landing navigation method for a rotor UAV of the present invention are as follows:

[0089] Step 1, Error Modeling of the Cooperative Target Coordinate System:

[0090] Convert the displacement amounts in the north, up, and east directions of the geodetic coordinate system solved by the inertial navigation system to the cooperative target coordinate system, and use the transfer matrix for conversion, where is calculated from the three attitude angle information of the inertial navigation.

[0091] The cooperative target during the landing process of the rotary-wing UAV is a QR code, a capital "H" logo, or other typical markers, and the landing method is vertical landing. In this scenario, the visual navigation system installed on the rotary-wing UAV provides the relative position information between the UAV and the cooperative target through the recognition of the cooperative target and the calculation of the relative position with the cooperative target. Convert the relative position between the UAV and the cooperative target in the cooperative target coordinate system to the absolute longitude and latitude position information in the geodetic coordinate system and provide it to the UAV flight control system for participation in control. When modeling in the geodetic coordinate system, it is necessary to obtain the position information of the QR code marker in advance, which cannot meet the demand for landing at any time.

[0092] Therefore, error modeling is carried out in the cooperative target coordinate system. Taking the takeoff and landing of the QR code marker of the rotary-wing UAV as an example, the QR code marker coordinate system is defined as follows:

[0093] The specific steps are as follows: As Figure 1 shown, the QR code coordinate system (a system): Take the center point of the QR code marker as the origin o a ; The x a axis is along the perpendicular line perpendicular to the plane of the QR code marker, with the inward direction being positive; the y a axis is perpendicular to the x a axis in the plane of the QR code marker, with the upward direction being positive; the z a axis is perpendicular to the x a axis in the plane of the QR code marker, with the rightward direction being positive; ox a y a z a constitute a right-handed coordinate system; the coordinates of a point in the QR code coordinate system are represented by (x a , y a , z a ).

[0094] Use the Kalman filtering method to fuse the inertial navigation and visual navigation. The continuous state equation of the inertial-visual integrated navigation system is as follows:

[0095]

[0096] In the formula, F(t) is the state transition matrix of the continuous state equation at time t, is the system random noise vector at time t. The system error variable to be estimated, Derivative of error variable.

[0097] The filtered state variable X is established as 9 - dimensional. The filtered state variable includes the position error s (unit: m), velocity error v (unit: m / s), and acceleration error a (unit: m / s²) of the inertial navigation system in the cooperative target coordinate system within the time interval Δt. In the formula, the subscripts x, y, and z represent the three directions of the cooperative target coordinate system.

[0098] X T =(s x , v x , a x , s y , v y , a y , s z , v z , a z )(2)

[0099] According to the following motion equation

[0100]

[0101] The system state transition matrix in the formula can be deduced as:

[0102]

[0103] Among them, F 1 , F 2 , F 3 represent the sub - transition matrices in the three directions of the coordinate system, and their expression methods are all the same as the formula:

[0104]

[0105] In formula (3), Δt is the time interval between two adjacent frames of visual valid information.

[0106] The observation equation of the integrated navigation system is defined as follows:

[0107]

[0108] is the system observation variable, H(k) is the measurement matrix of the integrated navigation filter, is the observation noise matrix.

[0109] Taking the difference between the displacement of the inertial navigation system and the displacement of the visual navigation system in the x, y, and z directions of the cooperative target coordinate system within the time Δt as the observation information (3 - dimensional):

[0110]

[0111] In Equation (6), δs represents the system position error, including the x, y, and z directions, where s VIS represents the displacement between adjacent frames in the cooperative target coordinate system solved by the visual navigation system. s INS represents the displacement between adjacent frames in the cooperative target coordinate system of the inertial navigation system, and the calculation method is as follows:

[0112]

[0113] In Equation (7):

[0114] is the attitude transfer matrix from the geographic coordinate system to the cooperative target coordinate system, which is determined by the placement direction of the cooperative target. Lat is the latitude of the inertial navigation system, Hgt is the altitude of the inertial navigation system, Lon is the longitude of the inertial navigation system, and the subscripts k and k - 1 represent the current moment and the previous moment. R m and R n represent the radius of the Earth's meridian and the radius of the Earth's prime vertical circle, respectively.

[0115] Thus, the observation matrix H can be determined as follows:

[0116] H(k) = [H 1 H 2 H 3 (9)

[0117] where:

[0118]

[0119] Second, fuse the inertial navigation and visual navigation information:

[0120] After completing the error modeling, according to the error model obtained in the first step, fuse the inertial navigation and visual navigation information through the Kalman filter algorithm. The Kalman filter calculation method can be divided into five steps as follows:

[0121] f) One-step state prediction:

[0122]

[0123] g) One-step prediction mean square error:

[0124]

[0125] h) Filter gain:

[0126]

[0127] i) State estimation:

[0128]

[0129] j) Estimated mean square error:

[0130]

[0131] In Equations 10 to 14:

[0132] One-step prediction estimated state vector;

[0133] Φ k,k-1 : Discretized F matrix in the equation;

[0134] Optimal estimated state vector;

[0135] P k,k-1 : One-step prediction covariance matrix;

[0136] P k : Optimal estimated covariance matrix;

[0137] K k : Optimal gain matrix;

[0138] R k : Discretized measurement noise matrix;

[0139] Q: Discretized system noise covariance matrix.

[0140] Among them, the system measurement noise matrix R k reflects the accuracy of the observation information, that is, the position accuracy of visual navigation. The combined navigation correction amount decreases with the increase of the R k value. Therefore, the system measurement noise matrix is fitted according to the characteristics of visual navigation.

[0141] The third step, the method for fitting the system measurement noise matrix: Calculate the relative distance between the rotary-wing UAV and the cooperative target, fit the measurement noise matrix, and obtain the accurate k value of R.

[0142] During the landing process of the UAV, as the distance to the cooperative target approaches, the imaging of the cooperative target gradually becomes clear, and the visual navigation accuracy also increases. Therefore, the traditional method of fixed measurement noise matrix is not applicable to the inertial / visual integrated navigation during the autonomous landing process of the rotary-wing UAV. At different altitudes relative to the ground, the measurement noise matrix needs to be linearly fitted.

[0143] The landing process of the rotary-wing UAV is divided into the following four stages:

[0144] 1. Far from the cooperative target, at this time the target imaging is not clear and the visual navigation accuracy is poor;

[0145] 2. As the distance to the cooperation target approaches, the imaging of the cooperation target gradually becomes clear, and the visual navigation accuracy improves as the UAV descends;

[0146] 3. After the UAV descends to a certain height, as it approaches the cooperation target, the cooperation target gradually becomes incomplete in the camera imaging, and the visual navigation error rate increases;

[0147] 4. As the UAV descends to a certain height, the cooperation target is too small in the camera imaging, and the visual navigation accuracy is poor.

[0148] According to the above four stages, during the landing process of the rotary-wing UAV, the fixed measurement noise matrix cannot accurately correct the inertial navigation. Therefore, a method for fitting the measurement noise matrix is designed. Taking the lateral direction as an example, the noise value curve is as Figure 2 shown:

[0149] Among them, the measurement noise matrix is determined according to multiple factors such as the actual flight height of the UAV, the accuracy of the inertial navigation system, and the field of view angle of the visual sensor. Taking the common UAV landing height (200m) as an example for the four stages described above, the measurement noise matrix R k is fitted by a quadratic curve as follows:

[0150] 1) When the relative distance to the cooperation target is greater than 200m, R k = 100;

[0151] 2) When the relative distance to the cooperation target is greater than 10m and less than 200m, with x = 10m as the axis of symmetry, (10m, 5m) as the vertex of the parabola, and (5m, 100m) as a point on the parabola, the parabola opens upward, and the fitting curve equation is: R k = 0.002631x 2 - 0.052632x + 5.26322;

[0152] 3) When the relative distance to the cooperation target is greater than 5m and less than 10m, with x = 10m as the axis of symmetry, (10m, 5m) as the vertex of the parabola, and (200m, 100m) as a point on the parabola, the parabola opens upward, and the fitting curve equation is: R k = 3.8x 2 - 76x + 385;

[0153] 4) When the relative distance to the cooperation target is less than 5m, R k = 100;

[0154] The method for fitting the measurement noise matrix proposed in this section can effectively improve the landing navigation accuracy of the rotary-wing UAV based on the inertial / visual combination under satellite denial conditions.

[0155] Fourth step, position error correction:

[0156] Before the visual navigation position solution is completed, the Kalman filter time update is performed according to Equations 10 to 14. After the visual navigation solution is completed and the position information is obtained, the observation quantity Z of the integrated navigation system is calculated according to the equation k , and the Kalman filter measurement update is performed according to Equations 10 to 14. After the filtering is completed, the estimated value of the system error variable is obtained in the equation, and various errors of the inertial navigation are corrected:

[0157] 1. Position error correction:

[0158] The estimated value of the position error ΔP in the navigation coordinate system n is:

[0159]

[0160] In the formula, ΔP a is the estimated value of the position error in the cooperative target coordinate system, is the attitude transfer matrix from the cooperative target coordinate system to the navigation coordinate system, and its northward error and eastward error are converted into radian coordinates:

[0161]

[0162] Perform position correction:

[0163]

[0164] 2. Velocity error correction:

[0165] The estimated value of the velocity error in the navigation target coordinate system is:

[0166]

[0167] Perform velocity correction:

[0168]

Claims

1. A method for inertial vision landing navigation of a rotor UAV, characterized in that: The specific steps of this navigation method are as follows: First step, cooperation target coordinate system error modeling: Convert the displacement calculated by the inertial navigation system in the geographical coordinate system to the cooperative target coordinate system, and use the attitude transfer matrix for conversion, and then perform error modeling in the cooperative target coordinate system; The cooperation target coordinate system error model includes the continuous state equation of the inertial vision integrated navigation system, the filtered state variables, the system state transition matrix, the observation equation of the integrated navigation system, and the attitude transition matrix; Second step, fuse the inertial navigation and vision navigation information: According to the error model obtained in the first step, fuse the inertial navigation and vision navigation information through the Kalman filter algorithm; Step 3, System measurement noise matrix fitting method: Calculate the relative distance between the rotary-wing UAV and the cooperative target, and fit the measurement noise matrix R k to obtain the k accurate value; According to the four stages of the landing process of the rotary-wing UAV, the measurement noise matrix R for each stage is fitted by a quadratic curve as follows: k as follows: 1) When the relative distance to the cooperation target is greater than 200m, R k = 100; 2) When the relative distance to the cooperation target is greater than 10m and less than 200m, with x = 10m as the axis of symmetry, (10m, 5m) is the vertex of the parabola, (5m, 100m) is a point on the parabola, the parabola opens upward, and the fitting curve equation is: R k = 0.002631x 2 - 0.052632x + 5.26322, where x is the distance of the rotary-wing UAV relative to the cooperative target; 3) When the relative distance to the cooperation target is greater than 5m and less than 10m, with x = 10m as the axis of symmetry, (10m, 5m) is the vertex of the parabola, (200m, 100m) is a point on the parabola, the parabola opens upward, and the fitting curve equation is: R k = 3.8x 2 - 76x + 385; 4) When the relative distance to the cooperation target is less than 5m, R k = 100; Fourth step, position error correction: that is, correct various errors of inertial navigation: a) Position error correction: The estimated value of the position error ΔP in the navigation coordinate system n is as follows: where $X_{k}[1]$, $X_{k}[4]$, and $X_{k}[7]$ are the position errors in the $x$, $y$, and $z$ directions, respectively, in the cooperative target coordinate system, and $\Delta P$ a is the estimated value of the position error in the cooperative target coordinate system, is the attitude transfer matrix from the cooperative target coordinate system to the navigation coordinate system, and its northward error and eastward error are converted into radian coordinates: Execute position correction: where, ΔP n [1], ΔP n [2], ΔP n [3] are respectively the estimated values of position errors in the north, up, and east directions in the navigation coordinate system. Velocity error correction: The estimated value of the velocity error in the navigation target coordinate system is: Where, Xk[2], Xk[5], Xk[8] are the velocity errors in the x, y, and z directions in the cooperation target coordinate system respectively, Execute velocity correction: Among them, ΔV n [1], ΔV n [2], ΔV n [3] are respectively the estimated values of velocity errors in the north, up, and east directions in the navigation target coordinate system.

2. The method for inertial vision landing navigation of a rotor UAV according to claim 1, characterized in that: In the first step, the cooperation target adopts a QR code; the cooperation target coordinate system is: with the center point of the QR code logo as the origin o a ; x a axis along the perpendicular line perpendicular to the plane of the QR code logo; y a axis is perpendicular to x in the plane of the QR code logo a axis, with the upward direction being positive; z a axis is perpendicular to x in the plane of the QR code logo a axis, with the rightward direction being positive; ox a y a z a constitute a right-handed coordinate system; Fuse inertial navigation and vision navigation using the Kalman filter algorithm.

3. The method for inertial vision landing navigation of a rotor UAV according to claim 1, characterized in that: In the first step, the cooperation target coordinate system error model is as follows: The continuous state equation of the inertial vision integrated navigation system is, where \(F(t)\) is the state transition matrix of the continuous state equation at time \(t\), is the system random noise vector at time \(t\), the system error variable to be estimated, the derivative of the error variable The expression of the filtered state variable X is, X T = (s x , v x , a x , s y , v y , a y , s z , v z , a z )(2) The position error s in the cooperation target coordinate system, unit: m, the velocity error v, unit: m / s, the acceleration error a, unit: m / s2, where the subscripts x, y, z represent the three directions of the cooperation target coordinate system; The system state transition matrix is, Among them, F 1 , F 2 , F 3 represent the sub-transfer matrices in three directions of the coordinate system; Where, The observation equation of the integrated navigation system is, is the system observation variable, and H(k) is the measurement matrix of the integrated navigation filter, is the observation noise matrix; Observation information is as follows: Wherein, δs represents the system position error, including the x, y, and z directions, and s VIS represents the displacement between adjacent frames in the cooperative target coordinate system solved by the visual navigation system, and s INS represents the displacement between adjacent frames in the cooperative target coordinate system of the inertial navigation system, and s INS The calculation method is as follows: In the formula: is the attitude transfer matrix from the geographic coordinate system to the cooperative target coordinate system, which is determined by the placement direction of the cooperative target. Lat is the latitude of the inertial navigation system, Hgt is the altitude of the inertial navigation system, Lon is the longitude of the inertial navigation system, and the subscripts k and k-1 represent the current moment and the previous moment, respectively. R m and R n represent the radius of the Earth's meridian and the radius of the Earth's prime vertical, respectively. From this, the observation matrix H can be determined: H(k) = [H 1 H 2 H 3 (9) Where:

4. The method for inertial vision landing navigation of a rotor UAV according to claim 1, characterized in that: In the second step, the steps of the Kalman filter calculation method are as follows: a) State one-step prediction: b) One-step prediction mean square error: c) Filter gain: d) State estimation: e) Estimation mean square error: In the formula: One-step prediction estimated state vector; Φ k,k-1 : The F matrix after discretization; Optimal estimated state vector; P k,k-1 : One-step prediction covariance matrix; P k : Optimal estimation covariance matrix; K k : Optimal gain matrix; R k : The measured noise matrix after discretization; Q: The system noise covariance matrix after discretization; H k is the measurement matrix; Z k is the observable quantity; The subscript k represents the current moment.

5. The method for inertial vision landing navigation of a rotor UAV according to claim 1, characterized in that: In the third step, the landing process of the UAV is divided into the following four stages: 1) The distance from the cooperation target is far, at this time the target imaging is not clear and the vision navigation accuracy is poor; 2) As the distance to the cooperation target approaches, the cooperation target imaging gradually becomes clear and the vision navigation accuracy improves as the UAV descends; 3) After the UAV descends to a certain height, as it approaches the cooperation target, the cooperation target gradually becomes incomplete in the camera imaging and the vision navigation error rate increases; 4) As the UAV descends to a certain height, the cooperation target is too small in the camera imaging and the vision navigation accuracy is poor.

Images