CN110598834A - Binocular vision detection system structure optimization method - Google Patents

Abstract

The invention discloses a method for optimizing the structure of a binocular vision detection system, comprising the following steps: establishing a performance analysis model of a binocular detection system; analyzing the influence of the system structure on measurement accuracy; analyzing the influence of index parameters; optimizing structural parameters; The structural model of the binocular system is analyzed for accuracy, and the relationship between the structural parameters of the binocular structure and the system index parameters is obtained, and then the univariate numerical analysis of the binocular system index parameters is carried out to determine the influence of each structural parameter on the index parameters. Under the actual industrial requirements and guaranteed effective field of view and resolution, by setting the optimization objective function and applying the improved particle swarm optimization algorithm, the optimal solution of the binocular structure parameters is obtained. Through this method, the original artificial experience design problem is solved. The structural uncertainty caused by the binocular system structure makes the design of the binocular system structure evidence-based and improves the measurement accuracy of the binocular detection system.

Description

A structure optimization method for binocular vision detection system

technical field

The invention belongs to the technical field of binocular vision detection systems, and in particular relates to a method for optimizing the structure of a binocular vision detection system.

Background technique

The binocular vision detection system is a complex mechanical system, and the structural parameters of the detection system have a great influence on the performance of the binocular system. The indicators describing the performance of the binocular system mainly include the effective field of view, resolution, and detection accuracy. The indicators describing the performance of the binocular system mainly include the effective field of view, resolution, and detection accuracy. The angle α with the optical axis and the focal length of the camera f, and the coupling between each parameter is strong, which increases the difficulty of designing a high-precision, large field of view, and high-resolution binocular detection system. With the development of science and technology, Binocular detection has become a major technical means applied to hot tasks such as on-orbit capture of non-cooperative targets, space junk removal, and trajectory planning. Since the detection distance of space targets is relatively long and is greatly affected by external factors, the improvement of binocular detection The measurement accuracy of the system is of great significance.

In terms of intelligent algorithm for structural optimization, Peng D M, Fairfield C A. proposed a structural optimization algorithm combining finite element analysis method and genetic algorithm to optimize the arch bridge structure. Albuquerque A T D, Debs MK E, Melo A M C. Discretize the design variables that need to be optimized, and apply a hybrid genetic algorithm to analyze the discrete variables. Xiao R B, Tao Z W proposed the Interactive Search Algorithm (ISA), which designed two different search schemes and improved the competitiveness of the algorithm. Xiao R B, Tao Z W Whale Optimization Algorithm (WOA) combined with Differential Evolution (de) to improve convergence speed and avoid local optima. Javaid S, Ali I, Mushtaq N provide free-flight constraints to the crow search algorithm, and individual capping strategies, thereby eliminating unnecessary structural analysis. Yashar D A, Wojtusiak J proposed a new integer-permutation-based genetic algorithm (IPGA) to deal with manufacturability constraints by using constraint-controlled sorting techniques in the fitness degree assignment stage. Khoshnoudian F, Talaei S, Fallahian M proposed a heuristic algorithm based on surface plasmon resonance particle swarm optimization (SPR-PSO) and designed different fitness functions for different modulation methods. SeifiH, Rezaee Javan A Seek to improve structural performance and design efficiency using transition segment method and bidirectional evolutionary structural optimization (BESO) method, enabling fast and accurate fabrication of customized structural nodes. Ho-Huu V, Duong-Gia D used multi-objective evolutionary optimization algorithm to solve multi-objective design optimization problems and used reliability analysis method to evaluate the reliability of the solution set. Wang S, Wang M, Xin G U used the response surface optimization method to optimize the multi-objective structure of the baffle. Jing Z, Chen J, Li used genetic algorithm (GA) to solve constrained optimization problems, which has high accuracy and robustness. Wang C, Yu T, Shao G combined extended isogeometric analysis (XIGA) and chaotic particle swarm optimization algorithm, and proposed a new method that can effectively get rid of local optimum.

However, there are still two problems in the above analysis methods that have not been solved, that is, the problem of establishing a multi-parameter structural error model, and how to optimize the strong coupling parameters. These methods are easy to fall into Premature and local optimal situations lead to low accuracy of final structure optimization.

Therefore, in view of this, researches and improvements are made on the existing structures and deficiencies, and a structural optimization method of the binocular vision detection system is provided, in order to achieve more practical purposes.

Contents of the invention

In order to solve the above technical problems, the present invention provides a structural optimization method of binocular vision detection system to solve the problem that the existing binocular detection system tends to fall into premature and local optimum when solving multi-parameter, strong coupling and nonlinear optimization problems question.

The purpose and efficacy of a binocular vision detection system structure optimization method of the present invention are achieved by the following specific technical means:

A method for optimizing the structure of a binocular vision detection system, comprising the following steps:

S1. Establish a binocular detection system performance analysis model:

The main performance indicators for evaluating the binocular detection system are the effective field of view R, the horizontal resolution Δx, the vertical resolution Δy, and the system detection accuracy Δ. Aiming at these performance indicators, the performance analysis models of the binocular detection system are established respectively;

S2. Analyze the impact of system structure on measurement accuracy:

(1) Accuracy analysis of binocular detection system;

(2) Analyze the impact of the effective field of view of the binocular system on the measurement accuracy;

(3) Analyze the impact of binocular system resolution on measurement accuracy;

S3. Impact analysis of index parameters:

The structural parameters of the binocular system include baseline angle α, baseline distance B, and camera focal length f;

(1) Analyze the influence of the baseline angle α in the structural parameters on the index parameters of the binocular system;

(2) Analyze the impact of the baseline B in the structural parameters on the index parameters of the binocular system;

(3) Analyze the impact of the camera focal length f on the index parameters of the binocular system in the structural parameters;

S4. Structural parameter optimization:

On the basis of the basic particle swarm optimization algorithm, the structural parameters are optimized by applying the improved particle swarm optimization algorithm (IEPSO);

S5. Simulation analysis:

Using the improved particle swarm optimization algorithm (IEPSO) to search for the optimal solution globally in multiple groups of feasible solutions, and quickly and efficiently complete the optimal solution of the binocular system structure configuration;

(1) Optimizing the objective function;

The system structure is determined by the optical axis baseline angle α, the focal length f, and the baseline distance B, but the influence of the focal length f and the baseline distance B on the measurement accuracy is a pair of mutually restrictive parameters, which mainly affect the measurement accuracy and detection depth. The optimization of variables can be solved by particle swarm optimization, but the fitness function of this particle swarm optimization must be related to this structural parameter and the constraint variable, and at least two parameters can be coupled with each other as the optimization condition. Based on the above two points, the The representative effective field of view, system resolution, and binocular system measurement accuracy are used as coupling parameters to achieve the purpose of optimization;

(2) Optimize parameter settings;

Using the improved particle swarm optimization algorithm (IEPSO), the optimal solution of the binocular structure parameters is obtained;

(3) Simulation analysis;

Apply Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Improved Particle Swarm Optimization (IEPSO) respectively to obtain the simulation optimization results and analyze them; Find the optimal solution globally, and quickly and efficiently complete the optimization of the structural parameters of the binocular detection system; compare the parameters before and after optimization, and determine that the system structure obtained by applying the improved particle swarm optimization algorithm (IEPSO) has higher structural accuracy.

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

The structure model of binocular system precision analysis is established, and the relationship between the structural parameters of the binocular structure and the system index parameters is obtained, and then the single-variable numerical analysis of the binocular system index parameters is determined to determine the impact of each structural parameter on the index parameters, and finally in By setting the optimization objective function and applying the improved particle swarm optimization algorithm (IEPSO) to meet the actual industrial requirements and ensure the effective field of view and resolution, the optimal solution of the binocular structure parameters is obtained. The effective field of view radius is 597.05mm, With a horizontal resolution of 2.0738mm and a vertical resolution of 0.0960mm, the measurement error of the binocular detection system can reach 0.01mm. This method solves the problem caused by the original binocular system structure designed by human experience. The structural uncertainty of the binocular system makes the design of the binocular system structure evidence-based, improves the measurement accuracy of the binocular detection system, and avoids the problem of the existing binocular detection system when solving multi-parameter, strong coupling, and nonlinear optimization problems. It is easy to fall into the problem of premature and local optimum.

Description of drawings

Fig. 1 is a performance analysis model diagram of the binocular detection system of the present invention;

Fig. 2 is a schematic diagram of the effective field of view of the binocular system of the present invention;

Fig. 3 is a schematic diagram of coordinate transformation of the binocular system of the present invention;

Fig. 4 is a graph showing the influence of the included baseline angle α on the effective field of view radius R of the present invention;

Fig. 5 is a graph showing the influence of the baseline angle α on the horizontal resolution Δx in the present invention;

Fig. 6 is a graph showing the influence of the included baseline angle α on the vertical resolution Δy of the present invention;

Fig. 7 is a distribution diagram of the influence of the baseline angle α on the measurement error in the present invention;

Fig. 8 is a graph showing the influence of the baseline distance B on the effective field of view radius R of the present invention;

Fig. 9 is a graph showing the influence of the baseline distance B on the horizontal resolution Δx in the present invention;

Fig. 10 is a graph showing the influence of the baseline distance B on the vertical resolution Δy in the present invention;

Fig. 11 is the impact distribution diagram of the structure parameter B/Z of the present invention on the measurement error;

Fig. 12 is a graph showing the influence of the focal length f of the camera of the present invention on the radius R of the effective field of view;

Fig. 13 is a graph showing the influence of the focal length f of the camera of the present invention on the horizontal resolution Δx;

Fig. 14 is a graph showing the influence of the focal length f of the camera of the present invention on the vertical resolution Δy;

Figure 15 is a distribution diagram of the influence of the baseline angle α on the measurement error Δ in the present invention;

Fig. 16 is a flowchart of the improved particle swarm algorithm of the present invention;

Fig. 17 (a) is the convergence curve of genetic algorithm, particle swarm optimization algorithm and improved particle swarm optimization algorithm objective function;

Fig. 17 (b) is genetic algorithm, particle swarm optimization algorithm and improved particle swarm optimization algorithm measurement error optimization curve;

Fig. 18 (a) is a convergence curve diagram of the focal length f of the camera of the present invention;

Fig. 18 (b) is the convergence curve diagram of the baseline distance B of the present invention;

Fig. 18 (c) is a convergence curve diagram of baseline included angle α of the present invention;

Fig. 18 (d) is an optimization curve diagram of the effective field of view radius R of the present invention;

Fig. 18 (e) is a horizontal resolution optimization curve diagram of the present invention;

Fig. 18 (f) is a vertical resolution optimization curve diagram of the present invention;

Fig. 18 (g) is the iteration graph of objective function tz of the present invention;

Fig. 18(h) is a measurement error optimization diagram of the present invention.

Detailed ways

The present invention will be further described below in conjunction with accompanying drawing:

Example:

As shown in accompanying drawing 1 to accompanying drawing 18 (h): the present invention provides a kind of binocular visual detection system structure optimization method, and its working principle is specifically as follows:

1. Establish performance analysis model of binocular detection system

The main performance indicators for evaluating the binocular detection system include effective field of view R, horizontal resolution Δx, vertical resolution Δy, and system detection accuracy Δ. According to these performance indicators, the performance analysis models of the binocular detection system are respectively established, as follows:

1.1 Principle of binocular detection system

The binocular system structure is a strong coupling model. In order to analyze the influence of the system structure on the measurement accuracy, this paper establishes a system performance analysis model, as shown in Figure 1. The coordinate origin of the measurement system is established at the projection center oc1 of one of the cameras, xc1 is along the direction of the optical center oc1, oc2, yc1 is vertically downward, zc1 and xc1, yc1 form a right-handed spiral coordinate system. The final deduced relationship between the structural accuracy of the binocular system is shown in formulas (1)-(5).

The position of the spatial point p(xc1, yc1, zc1) can be represented by α1, ω1, φ1;

Since f1=f2=f, get:

In the formula,

xc1, yc1, zc1 are the position coordinates of any point p(xc1, yc1, zc1) in the space;

B represents the baseline, which is the length of the line segment from optical center oc1 to oc2;

ω1 and ω2 represent the horizontal projection angles, which are the angles between the projection of a point p(xc1, yc1, zc1) on the xc1, zc1 plane and the optical axis in space;

φ1, φ2 represent the vertical projection angle;

α1, α2 are the angles between the baseline B and the optical axes of the two cameras;

f1, f2 represent the focal length of the two cameras;

(x1, y1), (x2, y2) represent the position coordinates of a point p(xc1, yc1, zc1) in the space in the image coordinate system.

1.2 Accuracy analysis of binocular detection system

From the above binocular system performance analysis model, it can be seen that the position information of a point p in space is determined by B, α1, α2, f1, f2, x1, x2, y1, y2 factors, that is, P(xc1, yc1, zc1)=f( B, α1, α2, f1, f2, x1, x2, y1, y2), then the system structure accuracy of point p is expressed as:

In the formula,

i=B, α1, α2, f1, f2;

x1, x2, y1, y2 are the influence factors of the measurement system;

δi is the extraction accuracy of each influencing factor, that is, the inherent error of the system.

In order to simplify the research process, this paper does not consider the influence of system inherent errors.

From the formula (6), the measurement error transmission coefficient of the coordinate of the space point p is:

In the formula,

θ1=ω1+α1;

θ2=ω2+α2.

The errors in the directions of the three coordinate axes are:

The measurement error of the whole system is:

1.3 Effective field of view of binocular system

The effective field of view of the binocular system is described by the effective field of view radius. The inscribed circle of the effective field of view in Figure 2 is the circle HEFG, and the radius of the inscribed circle is calculated from the area of any quadrilateral, which is the effective field of view radius.

According to the area formula of any quadrilateral, when the four sides are known,

In the formula,

s: area of any quadrilateral;

R: effective field of view radius;

k=1/2(AB+BC+CD+AD).

In the formula, θ is the field of view angle of the camera CCD, which is determined by the focal length f of the camera and the size of the CCD.

1.4 binocular system resolution

The resolution of the binocular system refers to the ability to distinguish the smallest change in the x and y coordinates of a point p in the effective field of view. Assuming that the CCD parameter we use is 2048×2048, and the pixel size is 5.5μm×5.5μm, it is defined by resolution: the resolution of an object is the size represented by the unit pixel:

In the formula,

x1: measure the length of the object in the x direction on the CCD;

x: measure the length of the object in the x direction in the structural precision coordinate system;

Δx: the length in the x direction represented by a pixel, which is the resolution in the x direction;

Δω: CCD pixel size.

In the same way,

From the analysis of the system structure, it can be seen that when the binocular system is a symmetrical structure (α1=α2,), the measurement accuracy is relatively high, and it can be known from the conversion between the binocular structure model coordinate system and the image coordinate system:

Then the resolution expressions in the x and y directions are:

In the formula,

Δx, Δy are the system resolutions in the x-direction and y-direction in the structural precision coordinate system, respectively;

ω is the horizontal projection angle of the binocular vision system.

It can be obtained from the geometric relationship:

The solution of the ω angle in the formula is a nonlinear solution process. This equation has multiple solutions, and the unqualified solution set can be filtered out by the actual constraint range of the ω angle in the optimization program.

2. Analysis of the influence of index parameters

The structural parameters of the binocular system include baseline angle α, baseline distance B, and camera focal length f. In order to understand the influence of structural parameters on the index parameters of the binocular system, that is, the effective field of view radius, resolution, and detection error, this paper respectively analyzes Three parameters are analyzed, and the numerical analysis method of single variable is applied for the convenience of the research, so as to find out the experimental law.

2.1 Baseline angle α

The size of the system structure parameter α angle directly affects the structure of the binocular system. The larger the α angle, the larger the detection field of view and the farther the detection depth. Therefore, it is of great significance to explore the value of the α angle. From formulas (10)-(18), it can be obtained that the effective field of view radius R of the binocular system is related to the system resolution and the three structural parameters. For the convenience of research, a single variable numerical analysis is set. It can be seen from Fig. 4 that the α angle has an effect on R There are multiple extreme points, but the maximum value in this range is too large to meet the actual situation, and additional constraints are required. Select the extreme points that meet the constraints to obtain the maximum value of R. It can be seen from Figure 5 and Figure 6 that the larger the α angle, the greater the horizontal resolution, while the vertical resolution has no linear relationship.

From formulas (6)-(9), the detection errors caused by α1 and α2 are respectively:

The variation curve of measurement error with angle α is shown in Fig. 7. It can be seen from the figure that the measurement error is the smallest when α1=α2, that is, when the structure of the binocular system is symmetrical (α1=α2), the measurement accuracy of the binocular system is higher.

2.2 Baseline distance B

The change of the binocular baseline distance will lead to the change of the detection depth z of the binocular system and the angle α1, α2 between the optical axis and the baseline. Therefore, the influence of the baseline distance B on the index parameters is very complicated. Figure 8 shows the relationship between the baseline B and the R Influence curve, the longer the baseline, the larger the effective field of view radius. Figure 9 and Figure 10 are the impact curves of baseline B on the system resolution. The longer the baseline, the lower the system resolution. Therefore, how The setting meets the requirements of larger field of view and higher resolution, and the baseline setting is very important.

In order to find out more clearly the influence of baseline B on detection accuracy, we assume that α1=α2, ω1=ω2=0, That is, it is assumed that the binocular camera is a parallel binocular system. Then the error transfer coefficient can be expressed as:

In the formula,

The total measurement error of this system with respect to B and z can be expressed as:

In the formula,

The measurement error of this system changes with the change of k. Its measurement error curve is shown in Figure 11. It can be seen from the figure that when the value of k is around 1.3, the error can reach the minimum value, which shows the importance of system structure design , but if the baseline distance is too large, it will directly lead to a decrease in measurement accuracy, and will also have a greater impact on the effective field of view of binocular measurement, and the size of the baseline distance is directly restricted by the weight, volume, and system space of the camera. Considering the design of system structure under the condition of factors

2.3 Camera focal length f

The focal length of the camera is an important parameter of the optical lens. The size of the focal length has a great influence on the measurement accuracy, field of view and resolution. Figure 12 shows the influence curve of the focal length f on R. Design according to the actual situation. Figure 13 and Figure 14 are the influence curves of the focal length f on the system resolution. The larger the focal length f is, the lower the horizontal resolution is, and the higher the vertical resolution is.

From (6)-(9), the error transfer function of the focal length can be obtained as:

The measurement error distribution of the focal length is shown in Figure 15. It can be seen from Figure 15 that the system measurement error decreases with the increase of the focal length, that is, the greater the effective focal length of the camera in the binocular vision system, the higher the measurement accuracy, but the optical Camera principle, the larger the focal length of the camera, the camera with too large focal length will inevitably increase the load, which is not only poor in maneuverability, but also not in line with the actual situation.

To sum up, when the binocular system has a symmetrical structure, that is, α1=α2, the measurement accuracy is higher. From the analysis of the baseline distance B, it can be known that the smaller the baseline distance is, the higher the measurement accuracy is, but the smaller the baseline distance is, the smaller the effective field of view radius R is. Small, the detection range is small. From the analysis of the focal length, it can be seen that the longer the focal length, the higher the measurement accuracy and the higher the resolution, but the larger the focal length of the camera, the greater the load, which does not meet the actual requirements. Therefore, if you want to have a larger detection depth and a larger detection range, you must A large baseline is required, which will reduce the measurement accuracy. At the same time, the focal length of the camera must be within a certain range, and it is impossible to achieve infinity. Therefore, how to design the system structure of binocular detection so that it can meet the conditions of structural design In this case, meeting the requirements of effective field of view, resolution and measurement accuracy is the top priority. At this time, single-variable numerical analysis can no longer meet this requirement. In order to solve this problem, this paper applies the improved particle swarm optimization method to the The structure of the binocular detection system is optimized, and the coupling condition of the binocular system is used as the fitness function of the algorithm, and the focal length and baseline range are used as constraints, so as to achieve the purpose of finding the optimal solution.

3. Structural parameter optimization method

The particle swarm optimization algorithm is proposed under the inspiration of the behavior rules of bird flocks, fish flocks and human society. The Boids model is mainly used to simulate the flight behavior of flocks of birds. In this model, the behavior of each individual is only close to his surroundings. It is related to individual behavior and needs to follow three principles, namely avoiding collisions, consistent speed, and gathering toward the center. For a feasible solution of the optimization problem, the quality of the particle is determined by a preset fitness function. Each particle will move in the feasible solution space, and its direction and distance are determined by a speed variable. Usually, the particle follows the current The optimal particle of , and the optimal solution is obtained after generation-by-generation search. The particle swarm algorithm uses a random initial population, and updates the position and speed of the particle swarm through adaptive learning, as shown in equations (26) and (27).

In the formula,

ω is the inertia weight;

C1 and C2 are acceleration items;

R1 and R2 are random numbers between [0,1];

Pgt is the global optimal position;

Pit is the historical optimal position found by the particle;

xid is the current iteration particle position;

vidt+1 is the next iteration speed.

Particle swarm optimization algorithm reduces the number of optimization iterations by virtue of the memory characteristics of particles, and can find the optimal solution in a short time, but at the same time, it also produces premature convergence, poor local optimization ability, and it is easy to fall into the local optimum. question. This paper makes improvements on the basis of the basic particle swarm optimization algorithm, and uses the improved particle swarm optimization algorithm to optimize the structural parameters. The improved particle swarm optimization algorithm (IEPSO) proposed in this paper not only retains the local development ability of the traditional particle swarm optimization algorithm, but also increases the local-global information sharing item. To improve the global exploration ability of the algorithm. Moreover, based on the idea of genetic algorithm population variation, the principle of last elimination is adopted to maintain the diversity of the population. The global optimization performance of the particle swarm optimization algorithm is improved through the above-mentioned improved method. Figure 16 shows the specific process of the IEPSO algorithm implementation.

Initialize the position and velocity of population particles randomly, calculate the fitness value of individual particles, and retain the position and fitness value of individual particles and global best particles in the current iteration. Then carry out the particle swarm operation. Added local-global information sharing items It is the information exchange between the local optimal particle and the global optimal particle obtained by the current iteration, which is used to balance the exploration and development capabilities of the particle in the global optimization process of the algorithm. The IEPSO algorithm uses equations (27) and (28) to update the speed and position.

Equation (29) consists of three parts, the first part is the "inheritance" of the previous velocity, the second part is the particle's "self-cognition", and the third part is "local information sharing". The fourth part is "local-global information sharing".

The IEPSO algorithm is no longer limited to the one-way communication between global particles and individual particles, the increased local-global information sharing item Information exchange between the locally optimal particles found for particles and the globally optimal particles obtained in the current iteration. The population velocity is updated according to formula (3), and the particles in the early stage of the algorithm search the entire search space at a relatively high speed to determine the approximate range of the optimal solution, which is conducive to the global search; the search space of most particles in the later stage gradually decreases and concentrates on In-depth search in the neighborhood of the optimal value is beneficial to local search.

Particles whose velocity does not exceed the preset range after the velocity update continue to retain their original velocity. When the particle velocity exceeds the velocity boundary, the maximum value of the velocity is assigned to the particle. Particles whose positions do not exceed the preset range after the position update continue to retain their original positions. When the particles exceed the preset range, the inferior particles are eliminated, and new particles are added within the preset range to form a new population. Recalculate the fitness value of the new population, save individual particles and the global optimal particle position and fitness value information obtained by the current iteration.

As the diversity of the population decreases, the algorithm can easily converge near the local optimum. In order to maintain the diversity of the population, the fitness value is used as the evaluation standard, the particles with poor objective function values are eliminated, and the particles are selected within the preset range to supplement the population, and the particle swarm operation is re-executed, and it stops when the convergence condition reaches the convergence accuracy. Iterate to obtain the global optimum.

4. Simulation analysis

The IEPSO algorithm is used to search for the optimal solution globally among multiple groups of feasible solutions, and quickly and efficiently complete the optimal solution for the configuration of the binocular system structure. Take the design of a certain type of binocular system as an example for simulation optimization, where B=1000mm, α=450, horizontal resolution Δx=1.1564mm, vertical resolution Δy=4.5505mm, f=10mm, binocular system detection error Δ= 10mm.

4.1 Optimizing the objective function

The system structure is determined by the optical axis baseline angle α, the focal length f, and the baseline distance B, but the influence of the focal length f and the baseline distance B on the measurement accuracy is a pair of mutually restrictive parameters, which mainly affect the measurement accuracy and detection depth. The optimization of variables can be solved by particle swarm optimization, but the fitness function of this particle swarm optimization must be related to this structural parameter and the constraint variable, and at least two parameters can be coupled with each other as the optimization condition. Based on the above two points, the The representative effective field of view, system resolution, and binocular system measurement accuracy are used as coupling parameters to achieve the purpose of optimization. The larger the effective field of view, the larger the detection range, the higher the resolution and measurement accuracy. The better the effect, the larger the effective field of view, the larger the baseline distance, the higher the resolution and the higher the measurement accuracy, the larger the required focal length, which is in line with the coupling parameter standard. The purpose of this paper is to improve the measurement accuracy of the binocular system structure as much as possible, and to minimize the system measurement error while maximizing the effective field of view and system resolution.

Based on the above analysis, the optimization objective function of this paper is:

tz=μ1Δ+μ2Δx+μ3Δy-μ4R

In the formula,

Δ: Binocular system measurement error, the smaller the measurement error, the higher the system accuracy;

Δx: Horizontal resolution of the binocular system, Δy: Vertical resolution of the binocular system, the smaller Δx and Δy, the stronger the resolution of the system;

R: The radius of the effective field of view of the binocular system, the larger R is, the wider the detection range of the binocular system;

μ1, μ2, μ3, μ4 are the weight coefficients of each parameter μ1+μ2+μ3+μ4=1, they are set after normalization processing, that is, to find the binocular system whose particles are within the parameter range and make the optimization objective function the smallest Excellent structure.

4.2 Optimize parameter settings

From the analysis of the impact of the above index parameters, it can be concluded that the structural detection accuracy of the symmetrical binocular system is the highest, so in this optimization process, the structural parameters of the binocular system are set as α1=α2=α; f1=f2=f; and the baseline length B , these three parameters are the structural parameters of particle swarm optimization. The specific process is to set the three-dimensional particle parameter range of α, f, B. Within this parameter range, the population size of each optimization is 500, and the optimization function is used to find the global minimum value, the system measurement error Δ has the largest weight, that is, to find the three-variable optimal solution that minimizes the measurement error, which is the optimal solution of the binocular system structure. The specific parameter settings are shown in Table 1.

Table 1: Parameter Settings

4.3 Simulation analysis

In this paper, the simulation optimization results obtained by applying genetic algorithm (GA), particle swarm optimization algorithm (PSO) and improved particle swarm optimization algorithm (IEPSO) are shown in Figure 17(a) and Figure 17(b).

Figure 17(a) is the convergence curve of the objective function of the three intelligent optimization algorithms. It can be seen from the figure that the genetic algorithm (GA) and the particle swarm optimization algorithm (PSO) converge in the 25th generation, while the improved particle swarm optimization algorithm (IEPSO) converges in the 14th generation , indicating that the improved particle swarm optimization algorithm has a faster convergence speed, and it can be seen from the figure that the objective function tz of the improved particle swarm optimization algorithm is the smallest, indicating that the improved particle swarm optimization algorithm has the best optimization effect. Figure 17(b) is the optimization curve of the measurement error of the three algorithms , it can be seen from the figure that in terms of measurement error, the measurement error of genetic algorithm (GA) and particle swarm optimization (PSO) can reach 0.03mm, while the measurement error of improved particle swarm optimization (IEPSO) can reach 0.01mm, that is, the application of improved particle swarm optimization The algorithm has the highest detection accuracy.

Based on the above analysis, the IEPSO algorithm is used to globally search for the optimal solution among multiple groups of feasible solutions, and quickly and efficiently complete the structural parameter optimization of the binocular detection system. The simulation results are shown in Figure 18.

Fig. 18 is a simulation analysis diagram of a system using the improved particle swarm optimization algorithm. Fig. 18(a), Fig. 18(b), and Fig. 18(c) are structural parameter convergence curves. This result converges, and the optimization result has practical significance. Figure 18(d), Figure 18(e), and Figure 18(f) are index parameter optimization curves, and Figure 18(d) is the effective field of view radius. Compared with the baseline distance optimization curve in Figure 18(b), the maximum effective The radius of the field of view is roughly half of the baseline distance. Figure 18(e) and Figure 18(f) show the horizontal resolution and vertical resolution respectively. The system resolution set according to this parameter can distinguish at least 3.2mm moving objects. Figure 18(g) is the iterative curve of the objective function tz, which converges at about the 14th generation, runs fast, and has a good convergence effect, indicating that the fitness function has strong coupling, meets the convergence condition, and meets the optimization standard of the particle swarm function. Figure 18(g) 18(h) is the measurement error optimization diagram. It can be seen from the figure that the measurement error can converge to 0.01mm. Compared with the pure empirical production system, the design accuracy of this system is higher.

Table 2: Comparison of parameters before and after optimization

Table 2 is a comparison of a set of parameters before and after the optimization based on experience and the application of the particle swarm optimization algorithm. It can be obtained that the effective field of view R increases by 98.1371mm when the structural parameters (B, α, f) meet the actual industrial requirements. The horizontal resolution is reduced by 0.9174mm, the vertical resolution is increased by 4.4545mm, and the measurement error is from 10mm to 0.01mm. Therefore, when the gap between the effective field of view and the resolution is small, the improved particle swarm algorithm is used to optimize the structural accuracy of the system. higher.

5. Summary

The structure model of binocular system precision analysis is established, and the relationship between the structural parameters of the binocular structure and the system index parameters is obtained, and then the single-variable numerical analysis of the binocular system index parameters is determined to determine the impact of each structural parameter on the index parameters, and finally in By setting the optimization objective function and applying the improved particle swarm optimization algorithm (IEPSO) to meet the actual industrial requirements and ensure the effective field of view and resolution, the optimal solution of the binocular structure parameters is obtained. The effective field of view radius is 597.05mm, With a horizontal resolution of 2.0738mm and a vertical resolution of 0.0960mm, the measurement error of the binocular detection system can reach 0.01mm. This method solves the problem caused by the original binocular system structure designed by human experience. The structural uncertainty of the binocular system makes the design of the binocular system structure evidence-based and improves the measurement accuracy of the binocular detection system.

Utilize the technical solution described in the present invention, or those skilled in the art design a similar technical solution under the inspiration of the technical solution of the present invention, and achieve the above-mentioned technical effects, all fall into the protection scope of the present invention.

Claims (1)

1. A binocular vision detection system structural optimization method, is characterized in that, comprises the following steps: S1. Establish a binocular detection system performance analysis model: The main performance indicators for evaluating the binocular detection system include the effective field of view, horizontal resolution, vertical resolution, and system detection accuracy. Aiming at these performance indicators, the performance analysis models of the binocular detection system are established respectively; S2. Analyze the impact of system structure on measurement accuracy: (1) Accuracy analysis of binocular detection system; (2) Analyze the impact of the effective field of view of the binocular system on the measurement accuracy; (3) Analyze the impact of binocular system resolution on measurement accuracy; S3. Impact analysis of index parameters: The structural parameters of the binocular system include baseline angle, baseline distance, and camera focal length; (1) Analyze the influence of the baseline clamp α in the structural parameters on the index parameters of the binocular system; (2) Analyze the influence of the baseline in the structural parameters on the index parameters of the binocular system; (3) Analyze the impact of the camera focal length on the index parameters of the binocular system in the structural parameters; S4. Structural parameter optimization: On the basis of the basic particle swarm optimization algorithm, the structural parameters are optimized by applying the improved particle swarm optimization algorithm; S5. Simulation analysis: The improved particle swarm optimization algorithm is used to search for the optimal solution globally among multiple sets of feasible solutions, and quickly and efficiently complete the optimal solution for the structural configuration of the binocular system; (1) Optimizing the objective function; The system structure is determined by the optical axis baseline angle, focal length, and baseline distance. The influence of focal length and baseline distance on measurement accuracy is a pair of mutually restrictive parameters. They mainly affect measurement accuracy and detection depth. Particle swarm algorithm is used to solve this mutually restrictive variable. The optimization of the particle swarm optimization algorithm, the fitness function of the particle swarm optimization algorithm must be related to the structural parameters and the constraint variables, and at least two parameters can be coupled with each other as the optimization condition. Based on the above two points, a representative effective field of view is set. System resolution and binocular system measurement accuracy are used as coupling parameters to achieve the purpose of optimization; (2) Optimize parameter settings; Using the improved particle swarm optimization algorithm, the optimal solution of the binocular structure parameters is obtained; (3) Simulation analysis; The genetic algorithm, particle swarm optimization algorithm, and improved particle swarm optimization algorithm were used to obtain the simulation optimization results and analyzed; based on the above analysis, the improved particle swarm optimization algorithm was used to search for the optimal solution globally in multiple groups of feasible solutions, and quickly and efficiently complete the dual The structural parameters of the visual detection system are optimized; the parameters before and after optimization are compared, and it is determined that the system structure obtained by applying the improved particle swarm optimization algorithm is of high precision.