Quick high-precision biomimetic focusing method for aerial camera

Abstract

The present application relates to a kind of aviation camera quick high-precision bionic focusing method, aviation camera focusing precision and speed are restricted by focusing mechanism and focusing method.The present application is based on human eye lens deformation and fish eyeball telescopic focusing mechanism, establishes the focusing system model of linear Lorentz force magnetic bearing zoom lens;On this basis, the neural network focusing method of combination of coarse focusing and fine focusing is designed, lens displacement is driven by Lorentz force magnetic bearing for coarse focusing, and fine focusing is carried out by driving lens deformation;The model of lens deformation and focal length is obtained using neural network training, so that the quick high-precision focusing of camera is realized.The present application belongs to the field of novel focusing mechanism and method, and can be applied to the quick high-precision focusing system of optical load of aviation aircraft.

Description

Technical Field

[0001] This invention relates to a fast and high-precision bionic focusing method for aerial cameras, applicable to fast and high-precision focusing systems for optical payloads of aircraft. Technical Background

[0002] The focusing accuracy and speed of aerial cameras are constrained by the focusing mechanism and focusing method. Traditional focusing mechanisms can often only change the lens position or lens deformation, and the two cannot be done simultaneously. Moreover, the relationship between the focal length and deformation of zoom lenses is difficult to determine, resulting in slow focusing speed and poor focusing accuracy.

[0003] This invention establishes a focusing system model for a zoom lens with dual linear Lorentz magnetic bearings based on the focusing mechanisms of human eye lens deformation and fish eyeball expansion and contraction. On this basis, a neural network focusing method combining coarse and fine focusing is designed. The Lorentz magnetic bearings drive lens displacement for coarse focusing and lens deformation for fine focusing. A model of lens deformation and focal length is obtained using neural network training, thereby achieving fast and high-precision focusing of the camera. Summary of the Invention

[0004] The technical problem solved by this invention is that traditional focusing mechanisms can only change the lens position or lens deformation, and the two cannot be done simultaneously; moreover, the relationship between the focal length and deformation of the zoom lens is difficult to determine, leading to slow focusing speed and poor focusing accuracy. This invention proposes a fast and high-precision bionic focusing method for aerial cameras. This method uses a zoom lens with dual linear Lorentz magnetic bearings to simultaneously drive the displacement and deformation of the zoom lens. A neural network is used for training to derive the relationship between deformation and focal length, ultimately producing a coarse-to-fine focusing equation, thereby improving the speed of focusing.

[0005] Specifically, the following steps are included:

[0006] (1) Model of double linear Lorentz force magnetic bearing

[0007] First, a model is created for a single linear Lorentz force actuator: when a DC power supply i is applied to the Lorentz force magnetic bearing coil, the output force of the Lorentz force magnetic bearing can be obtained as shown in equation (1);

[0008] F i =NBLi

[0009] Where F i The force provided to the Lorentz force magnetic bearing, N is the number of coil turns, B is the magnetic field strength at the coil, L is the effective length of the coil, and i is the current in the coil;

[0010] Assuming that the two coils have the same number of turns and effective length, the model of the double linear Lorentz force actuator can be represented as:

[0011] F L=NBL(i1+i2) (2)

[0012] Where F L For the forces provided by the dual linear Lorentz magnetic bearing, where i1 and i2 are the currents in coil 1 and coil 2 respectively, the force balance equations for the dual linear Lorentz magnetic bearing model system are:

[0013]

[0014] Where m is the mass of the mover, l1 and l2 are the displacements of coil 1 and coil 2 respectively, c is the friction damping coefficient, and k is the elastic damping coefficient;

[0015] (2) Focusing method based on coarse-fine combination of neural networks

[0016] The focal length of the zoom lens is:

[0017]

[0018] Where f is the focal length, r is the radius of the sphere, and n is the refractive index, let the relationship between the displacement difference of the two coils and r be:

[0019] r=Φ(l1-l2) (5)

[0020] Where Φ is the function relating displacement and radius, and a neural network is used to fit r:

[0021]

[0022] Where n is the number of nodes, C i Let σ be the center vector of the i-th node in the neural network. i Let w be the product width vector of the i-th node in the neural network. i Let be the weight vector from the hidden layer to the output layer of the i-th node in the neural network; lens imaging satisfies:

[0023]

[0024] Where u is the object distance and v is the image distance, from equation (2), it can be seen that when i1=i2, it is displacement focusing, and when i1≠i2, it is deformation focusing. Therefore:

[0025]

[0026] Using sharpness as the evaluation metric, the neural network focusing method can be expressed as:

[0027]

[0028] Where v0 is the initial image distance, and F is the image sharpness evaluation function, with F being closer to 1 indicating a sharper image.

[0029] Compared with existing solutions, the main advantages of the present invention are: traditional focusing mechanisms can often only change the lens position or lens deformation, and the two cannot be performed simultaneously; moreover, the relationship between the focal length and deformation of the zoom lens is difficult to determine, resulting in slow focusing speed and poor focusing accuracy.

[0030] This invention establishes a focusing system model for a zoom lens with dual linear Lorentz magnetic bearings based on the focusing mechanisms of human eye lens deformation and fish eyeball expansion and contraction. On this basis, a neural network focusing method combining coarse and fine focusing is designed. The Lorentz magnetic bearings drive lens displacement for coarse focusing and lens deformation for fine focusing. A model of lens deformation and focal length is obtained using neural network training, thereby achieving fast and high-precision focusing of the camera. Attached Figure Description

[0031] Figure 1 This is a flowchart of the present invention;

[0032] Figure 2 It is a component of a fast and high-precision bionic focusing system for aerial cameras;

[0033] Figure 3 Simulation comparing the lens displacement following error of the embodiments of the present invention with that of the traditional method;

[0034] Figure 4 Simulation comparing the focusing speed of the embodiments of the present invention with that of the conventional method; Detailed Implementation Plan

[0035] The present invention relates to a fast and high-precision bionic focusing system for aerial cameras, and the specific implementation scheme is as follows: Figure 1 As shown, the implementation target is as follows Figure 2 As shown, the specific implementation steps are as follows:

[0036] (1) Model of double linear Lorentz force magnetic bearing

[0037] First, a model is created for a single linear Lorentz force actuator: when a DC power supply i is applied to the Lorentz force magnetic bearing coil, the output force of the Lorentz force magnetic bearing can be obtained as shown in equation (10);

[0038] F i =NBLi (10)

[0039] Where F i The force provided to the Lorentz force magnetic bearing, N is the number of coil turns, B is the magnetic field strength at the coil, L is the effective length of the coil, and i is the current in the coil;

[0040] Assuming that the two coils have the same number of turns and effective length, the model of the double linear Lorentz force actuator can be represented as:

[0041] FL =NBL(i1+i2) (11)

[0042] Where F L For the forces provided by the dual linear Lorentz magnetic bearing, where i1 and i2 are the currents in coil 1 and coil 2 respectively, the force balance equations for the dual linear Lorentz magnetic bearing model system are:

[0043]

[0044] Where m is the mass of the mover, l1 and l2 are the displacements of coil 1 and coil 2 respectively, c is the friction damping coefficient, and k is the elastic damping coefficient;

[0045] (2) Focusing method based on coarse-fine combination of neural networks

[0046] The focal length of the zoom lens is:

[0047]

[0048] Where f is the focal length, r is the radius of the sphere, and n is the refractive index, let the relationship between the displacement difference of the two coils and r be:

[0049] r=Φ(l1-l2) (14)

[0050] Where Φ is the function relating displacement and radius, and a neural network is used to fit r:

[0051]

[0052] Where n is the number of nodes, C i Let σ be the center vector of the i-th node in the neural network. i Let w be the product width vector of the i-th node in the neural network. i Let be the weight vector from the hidden layer to the output layer of the i-th node in the neural network; lens imaging satisfies:

[0053]

[0054] Where u is the object distance and v is the image distance, from equation (11), it can be seen that when i1=i2, it is displacement focusing, and when i1≠i2, it is deformation focusing. Therefore:

[0055]

[0056] Using sharpness as the evaluation metric, the neural network focusing method can be expressed as:

[0057]

[0058] Where v0 is the initial image distance, and F is the image sharpness evaluation function, with F being closer to 1 indicating a sharper image.

[0059] Simulation results of lens displacement tracking using the focusing method described in this invention and traditional methods are as follows: Figure 3 As shown, the left figure shows the lens displacement controlled by the traditional focusing method, and the right figure shows the lens displacement controlled by the focusing method of the present invention. It can be seen that the actual position of the lens controlled by the traditional method has a phase difference with the command, which is not conducive to accurate imaging. However, the lens displacement control of the dual linear Lorentz force actuator combined with the neural network described in the present invention can track the displacement command more accurately and effectively improve the focusing accuracy.

[0060] A simulation comparing the focusing speed of the embodiments of the present invention with that of the conventional method is shown below. Figure 4 As shown in the figure, the left figure shows the focusing time of the traditional focusing method, and the right figure shows the focusing time controlled by the focusing method of the present invention. It can be seen that the traditional focusing method completes focusing in 9 seconds, while the focusing method of the present invention, which combines a dual linear Lorentz force actuator with a neural network, completes focusing in 5 seconds, thus improving the focusing speed.

[0061] The contents not described in detail in this invention are existing technologies known to those skilled in the art.

Claims

1. A quick high-precision biomimetic focusing method for aerial cameras, characterized in that: Based on the deformation of human eye lens and the focusing mechanism of fish eye, a double linear Lorenz force magnetic bearing model is established, and a coarse-fine combination focusing system model of zoom lens is established. A coarse-fine combination neural network focusing method is designed, which includes the following steps: (1) Double linear Lorenz force magnetic bearing model Assuming that the two coil turns and effective length are the same, the double linear Lorenz force actuator model is represented as: F L = NBL(i1+i2) (1) where F L is the force provided by the double linear Lorentz force magnetic bearing, N is the number of turns of the coil, B is the magnetic field strength at the coil, L is the effective length of the coil, i1 and i2 are the currents in the two coils, and the force balance equation of the double linear Lorentz force magnetic bearing model system is: Where m is the mass of the mover, l1 and l2 are the displacements of coil 1 and coil 2 respectively, c is the friction damping coefficient, k is the elastic damping coefficient, and t is the time. (2) Coarse-fine combination focusing method based on neural network The focal length of zoom lens is: Where f is the focal length, r is the spherical radius, n is the refractive index, and the relationship between the displacement difference of the two coils and r is: Where Φ is the displacement and radius function, and the neural network is used to fit r: where n is the number of nodes, C i is the center vector of the i-th node of the neural network, σ i is the width vector of the i-th node of the neural network, w i is the weight vector from the hidden layer to the output layer of the i-th node of the neural network; the lens imaging satisfies: Where u is the object distance, v is the image distance, and from equation (2), when i1 = i2, it is displacement focusing, and when i1 ≠ i2, it is deformation focusing, so we can get: Taking the clarity as the evaluation index, the neural network focusing method is represented as: Where v0 is the initial image distance, F is the image clarity evaluation function, and the closer F is to 1, the clearer the image is.

Images