Method for correcting lens distortion of roller binocular optical measurement system

Abstract

The invention relates to a lens distortion correction method for a roller binocular optical measurement system, and the method comprises the steps: defining a coordinate of a feature point on a calibration plate according to a forward neural network RBF network structure, the number of nodes of a hidden layer as a sample number n, a system input vector d = [xy] T, an output vector d '= [x'y'] T, and a weight matrix w as a 2 * n matrix; the system provides n feature point samples on the calibration board, and the network structure can know that the system outputs a corresponding result. Through the above calculation method, the lens imaging distortion of the roller binocular optical measurement system can be corrected, so that the precision of a roller product subjected to three-dimensional reconstruction measurement is higher, and the precision control of the specification and dimension of the roller can be carried out.

Description

Technical Field

[0001] The invention relates to the technical field of roller screws, and in particular to a lens distortion correction method for a roller binocular optical measurement system. Background Art

[0002] In a roller screw, the thread of the core component roller meshes with the thread of the screw, and the gears at both ends are also required to mesh with the inner gear ring of the nut. The parameter matching and modeling of the two are relatively complex, resulting in its design and manufacturing difficulty being much higher than that of a traditional ball screw. In actual processing and manufacturing, the screw is easy to manufacture, while the matching roller is not convenient to manufacture. The roller requires high precision, and the current workload of manual measurement is also large.

[0003] At present, in order to use the visual recognition system to accurately measure the roller in the roller screw and ensure that the thread of the roller meets the meshing requirements, a binocular optical measurement system is currently used to measure it. The system consists of a projector, a camera, and a turntable. When the system is working, the laser emits a line laser, the turntable rotates at the same time, and the camera receives the laser signal and performs three-dimensional reconstruction. However, it is found in actual measurements that the ideal pinhole model of the binocular optical measurement system is only an approximation of the real lens model. The actual camera and projector are different due to the different lens structures and the processing errors and assembly errors that occur during the production process, so the real optical model is not the same as the pinhole model. For ordinary lenses, especially the wide-angle lenses currently used, lens distortion needs to be considered. Among them, radial distortion has the greatest impact on imaging, resulting in a lack of accuracy in the roller products measured by three-dimensional reconstruction. Summary of the invention

[0004] The object of the present invention is to provide a lens distortion correction method for a roller binocular optical measurement system to solve the problems encountered in the above-mentioned background technology.

[0005] To achieve the above object, the technical solution of the present invention is as follows:

[0006] A method for correcting lens distortion of a roller binocular optical measurement system comprises the following steps:

[0007] Step 1: Let the radial distortion parameters of the lens be k1, k 2， Then there is

[0008]

[0009] Among them, (x', y') is the image coordinate obtained in the single-aperture camera model, (x, y) is the actual image coordinate, and r 2 =x 2 +y 2 ;

[0010] Step 2: According to the forward neural network RBF network structure, the coordinates of the feature points on the calibration plate are defined, the number of nodes in the hidden layer is the number of samples n, and the system input vector is d = [xy] T , the output vector is d'=[x'y'] T , the weight matrix w is a 2×n matrix;

[0011] Among them, the element w ij is the weight between the i-th node in the hidden layer and the j-th node in the output layer, and the radial basis function φ(d i ,d p ) takes the Gaussian kernel function, where d i is the i-th input vector, d p is the center point vector, σ is the expansion constant of the radial basis function, and its formula is

[0012]

[0013] Step 3: The system provides n feature point samples on the calibration plate. From the network structure, we can see that the system output is

[0014]

[0015] The expansion constant of the radial basis function is defined as shown in Equation 4

[0016]

[0017] Among them, d max is the maximum distance between samples, and n is the number of samples.

[0018] Compared with the prior art, the beneficial effects of the present invention are: through the above calculation method, the lens imaging distortion of the roller binocular optical measurement system can be corrected, so that the roller product measured by three-dimensional reconstruction has higher accuracy, and the specification size of the roller can also be accurately controlled. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The disclosure of the present invention is described with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. In the accompanying drawings, the same reference numerals are used to refer to the same components. Among them:

[0020] Figure 1 This is a structural diagram of the RBF network for lens distortion correction in the present invention;

[0021] Figure 2 This is a definition diagram of roller cross-section parameters in the present invention. DETAILED DESCRIPTION

[0022] In order to make the technical means, creative features, objectives and effects of the present invention easy to understand, the present invention is now further described in detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the relevant components of the present invention.

[0023] According to the technical solution of the present invention, without changing the essential spirit of the present invention, a person skilled in the art can propose a variety of interchangeable structural modes and implementation modes. Therefore, the following specific implementation modes and the accompanying drawings are only exemplary descriptions of the technical solution of the present invention, and should not be regarded as the entirety of the present invention or as a limitation or restriction to the technical solution of the present invention.

[0024] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0025] like Figure 1 As shown, a lens distortion correction method for a roller binocular optical measurement system is provided.

[0026] Let the radial distortion parameters of the lens be k1, k2, then we have

[0027]

[0028] Among them, (x', y') is the image coordinate obtained in the single-aperture camera model, (x, y) is the actual image coordinate, and r 2 =x 2 +y 2 .

[0029] Because the RBF network is an efficient forward neural network, the relationship between its input layer and hidden layer is nonlinear, and the relationship between the hidden layer and the output layer is linear weighted. This structure avoids the tedious and lengthy calculation of the BP network. While having good nonlinear approximation ability, it also has fast calculation ability, which is particularly suitable for nonlinear prediction from n-dimensional space to m-dimensional space.

[0030] The lens distortion correction RBF network structure is as follows Figure 1 As shown, its input signal is the coordinates (x, y) of the image taken by the real lens, and the output signal is the actual image coordinates (x', y'), which can be defined here as the coordinates of the feature points on the calibration plate, and the number of nodes in the hidden layer is the number of samples n.

[0031] The system input vector is d = [xy] T , the output vector is d'=[x'y'] T , the weight matrix w is a 2×n matrix, where the element w ij is the weight between the i-th node in the hidden layer and the j-th node in the output layer. i ,dp ) takes the Gaussian kernel function, where d i is the i-th input vector, d p is the center point vector, and σ is the expansion constant of the radial basis function.

[0032]

[0033] The system provides n feature point samples on the calibration plate. From the network structure, we can see that the system output is

[0034]

[0035] In order to avoid each radial basis function being too sharp or too flat, the expansion constant of the radial basis function is defined as shown in Equation 4

[0036]

[0037] Among them, d max is the maximum distance between samples, and n is the number of samples.

[0038] Calculating a reasonable σ can effectively improve the speed of the RBF network while maintaining the necessary selectivity, thereby calculating the image coordinates in the camera model more quickly.

[0039] As a preferred solution, the learning of the roller binocular optical measurement system is divided into two stages. The first stage is unsupervised learning, and the specific work is to solve the center and variance in the hidden layer. The second stage is supervised learning, and the specific work is to solve the weight matrix from the hidden layer to the output layer. The adjustment of the weight can be achieved by the minimum mean square error calculation, and the weight adjustment formula is:

[0040]

[0041] Among them, d' j is the jth expected value; w j =[w 1j w 2j ......w nj ] T ;

[0042]

[0043] Due to the existence of processing errors and assembly errors, the optical models of different lenses are also different and cannot be represented by a unified mathematical model. The RBF network can be used to quickly calculate the image coordinates in the camera model based on the coordinate values ​​of the actual image, thereby reconstructing the three-dimensional model of the roller more accurately.

[0044] In addition, during implementation, it is also necessary to measure and accurately control basic parameters such as the roller's diameter, roundness, surface roughness, element line straightness, and the arc connection condition between the element line and the end face.

[0045] Taking roundness as an example, the roller section is defined as Figure 2 As shown:

[0046] If the cross section is expressed in polar coordinates, the target curve to be measured is ρ(θ).

[0047]

[0048] Where i is the harmonic order, α i is the amplitude of the i-th harmonic, φ i is the initial phase of the i-th harmonic, iθ+φ i is the phase angle. If the cross section is a perfect circle, ρ should be a constant.

[0049] The measurement of basic parameters such as roller diameter, roundness, surface roughness, element line straightness, and arc connection between element line and end face has no direct relationship with distortion correction. Distortion is caused by the lens, so this solution uses RBF network to correct it. What is described here are the parameters that need to be measured. In this solution, not only the basic parameters of the roller need to be measured, but also the distortion needs to be corrected, so as to achieve better 3D reconstruction.

[0050] Through the above calculation method, the lens imaging distortion of the roller binocular optical measurement system can be corrected, so that the roller product measured by three-dimensional reconstruction has higher accuracy, and the specification and size of the roller can also be accurately controlled.

[0051] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for correcting lens distortion of a roller binocular optical measurement system, characterized in that: The following methods are included: Step 1: Let the radial distortion parameters of the lens be k1, k 2， Then there is Among them, (x', y') is the image coordinate obtained in the single-aperture camera model, (x, y) is the actual image coordinate, and r 2 =x 2 +y 2 ; Step 2: According to the forward neural network RBF network structure, the coordinates of the feature points on the calibration plate are defined, the number of nodes in the hidden layer is the number of samples n, and the system input vector is d = [xy] T , the output vector is d'=[x'y'] T , the weight matrix w is a 2×n matrix; Among them, the element w ij is the weight between the i-th node in the hidden layer and the j-th node in the output layer, and the radial basis function φ(d i ,d p ) takes the Gaussian kernel function, where d i is the i-th input vector, d p is the center point vector, σ is the expansion constant of the radial basis function, and its formula is Step 3: The system provides n feature point samples on the calibration plate. From the network structure, we can see that the system output is The expansion constant of the radial basis function is defined as shown in Equation 4 Among them, d max is the maximum distance between samples, and n is the number of samples.

2. The lens distortion correction method for a roller binocular optical measurement system according to claim 1, characterized in that: In step 2, the weight adjustment can be achieved by using the minimum mean square error calculation method. The weight adjustment formula is: Among them, d' j is the jth expected value, w j =[w 1j w 2j ......w nj ] T , 3. The lens distortion correction method for a roller binocular optical measurement system according to claim 1, characterized in that: It also includes the measurement of the basic parameters of the roller. Taking roundness as an example, the roller section parameters are defined and the section is expressed in polar coordinates. The target curve to be measured is ρ(θ) Where i is the harmonic order, α i is the amplitude of the i-th harmonic, φ i is the initial phase of the i-th harmonic, iθ+φ i is the phase angle; if the cross section is a standard circle, ρ should be a constant.

Images