CN105169979B - A kind of automotive paints mixing color method - Google Patents

Abstract

The invention relates to a method for blending colors of automobile paint, comprising the steps of: S1: collecting a surface image of a vehicle to be repainted, and identifying and obtaining surface color information according to the collected surface image; S2: generating a corresponding rough color matching formula according to the surface color information , and obtain the specified proportion of paint color masterbatch based on the color formula, mix and stir to obtain the required paint through anti-collapse compensation. Compared with the prior art, the present invention realizes the automatic modulation of the paint color on the surface of the vehicle to be repainted, overcomes the disadvantages of low efficiency and insufficient experience of manual modulation based on experience, and greatly reduces the damage caused by paint volatilization to the human body.

Description

A kind of automobile paint color blending method

technical field

The invention relates to the field of automotive paint repairs, in particular to a color blending method for automotive paint.

Background technique

In recent years, with the rapid development of the automobile industry, the popularization of automobiles is rapid, especially for family cars. However, with the rapid increase of automobile production, the use of automobile surface paint is also increasing. On the one hand, automobile manufacturers need to use specific color paint to treat the surface after the production of automobiles. On the other hand, as the number of automobiles continues to increase, traffic accidents also rise sharply, and collision damage to the surface of automobiles is inevitable. Lots of body shops need to treat them with a specific color of paint. However, the current plan is that each manufacturer or its 4S shop can only obtain the required color according to the ratio of the color formula. To adjust the color, the ratio of each component has been determined in advance, and each repair shop uses a large number of experienced painters. To artificially prepare the desired color.

There are serious deficiencies in the above approach. The color prepared by the 4S shop according to the formula ratio often deviates from the actual car color, because the surface color of the car is exposed to wind and sun during use, and there is a certain deviation from the factory color. Experience manual debugging. On the one hand, the manual preparation of paint for many years will cause harm to the human body; on the other hand, there are many paint suppliers, and the paint characteristics of each manufacturer are inconsistent and there are various types. Therefore, it is difficult for painters to accurately grasp the deployment ratio based on experience, there are technical problems, and the preparation process is slow and the efficiency is low.

Contents of the invention

Purpose of the present invention is exactly to provide a kind of automobile paint color blending method in order to overcome the defective that above-mentioned prior art exists.

The purpose of the present invention can be achieved through the following technical solutions:

A kind of automobile paint color blending method, comprises steps:

S1: Collect the surface image of the vehicle to be repainted, and obtain surface color information based on the collected surface image recognition;

S2: Generate the corresponding rough color formula according to the surface color information, and obtain the specified proportion of paint masterbatch based on the rough color formula, mix and stir, and obtain the required paint through feedback compensation.

Described step S2 specifically comprises the steps:

S21: Coarse adjustment: Based on the rough adjustment formula, obtain the paint color masterbatch in a specified ratio, mix and stir;

S22: Feedback: get a paint sample from the paint obtained by stirring, and judge whether the color difference between the sample color and the surface color is less than a threshold value, if yes, the blending ends, if no, then execute step S23;

S23: Fine-tuning: Generate a compensation formula according to the color difference between the sample color and the surface color, and obtain a specified proportion of paint color masterbatch based on the compensation formula, mix and stir.

Described step S22 specifically comprises the steps:

S221: Apply the stirred paint on a sampling plate made of the same material as the car shell to obtain a paint sample;

S222: The color sensor captures the light emitted by the paint sample under the irradiation of the base light source, and learns the color of the sample according to the captured light;

S223: Judging whether the color difference between the sample color and the surface color is less than a threshold, if yes, then the blending ends, if no, then execute step S23.

After the required paint is obtained in the step S2, the modulation log is uploaded to the host computer and saved.

In the step S1, the SVM classifier is used to identify and obtain surface color information based on the collected surface image, and the kernel function of the SVM classifier is an RBF kernel function.

The optimization process of the support vector machine parameter of described SVM classifier adopts the hybrid particle swarm algorithm based on genetic and simulated annealing to carry out, specifically comprises steps:

S11: Set the population size, crossover rate and mutation rate of GA, the maximum evolutionary generation and the minimum fitness threshold of GA;

S12: Generate an initial population of GA, where each GA chromosome represents a set of PSO acceleration coefficients;

S13: Calculate the individual fitness value of GA;

S14: Calculate the selection probability of each individual from the obtained individual fitness value, implement selection, crossover and mutation, and generate a new generation of GA population;

S15: Determine whether at least one of the following conditions is established:

a. The evolution algebra reaches the maximum evolution algebra,

b. The individual fitness value reaches the GA minimum fitness threshold,

If no, then return to step S13, if yes, then perform step S16;

S16: Output the optimal value, which is the optimal solution of the hybrid particle swarm optimization algorithm.

The step S13 specifically includes the steps of:

S131: Set the PSO population size, the maximum number of iterations, and the PSO minimum fitness threshold, where each PSO particle represents a set of optimized parameters for the SVM classifier;

S132: Initialize the acceleration coefficient and inertia weight coefficient, calculate the initial velocity and position of the particle, and initialize the initial value of the simulated annealing algorithm;

S133: Update the particle velocity and position, and calculate the fitness value difference ΔE before and after the update;

S134: Determine whether the fitness value difference is greater than 0, if yes, execute step S136, if not, execute step S135;

S135: Determine whether exp(ΔE/T)>rand(0,1) is established, if yes, execute step S136, if no, execute step S137;

S136: accept the update result, and execute step S138;

S137: Refuse to update the result, roll back particle velocity and position, and execute step S138;

S138: Calculate the adaptive weight coefficient, and update the individual extremum and the global extremum according to the adaptive value;

S139: Determine whether at least one of the following conditions is established:

c. The number of iterations reaches the maximum number of iterations,

d. Reach the PSO minimum fitness threshold,

If no, return to step S133; if yes, find the optimal value, and the optimal value found by PSO is the individual fitness of GA.

Compared with the prior art, the present invention has the following advantages:

1) The automatic modulation of the paint color on the surface of the vehicle to be repainted is realized, which overcomes the disadvantages of low efficiency and insufficient experience of manual modulation based on experience modulation, and greatly reduces the damage caused by paint volatilization to the human body.

2) The device can be adjusted according to the actual situation of the color of the car watch, instead of relying solely on the existing color scheme, the color modulated is more accurate and reliable, reducing the phenomenon of rework due to inaccurate paint color matching, saving a lot of manpower and material resources .

3) In the special case of paint toning, the complementary method of image processing and sensor feedback compensation can not only ensure the accuracy of the modulation method, but also reduce the expenditure of hardware system resources and improve the modulation efficiency.

Description of drawings

Fig. 1 is a schematic flow chart of the main steps of the present invention;

Fig. 2 is a schematic flow sheet of the present invention;

Fig. 3 is a method control flow chart of the present invention;

Wherein 1, deployment mechanism, 2, microprocessor.

detailed description

The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is carried out on the premise of the technical solution of the present invention, and detailed implementation and specific operation process are given, but the protection scope of the present invention is not limited to the following embodiments.

A kind of automobile paint color blending method, as shown in Figure 1 and Figure 2, comprises steps:

S1: Collect the surface image of the vehicle to be repainted, and obtain surface color information based on the collected surface image recognition, specifically including steps:

In step S1, an SVM classifier is used to identify surface color information based on the collected surface image, and the kernel function of the SVM classifier is an RBF kernel function.

The optimization process of the support vector machine parameters of the SVM classifier is carried out using a hybrid particle swarm optimization algorithm based on genetics and simulated annealing, which specifically includes steps:

S11: Set the population size, crossover rate and mutation rate of GA (genetic algorithm), the maximum evolutionary generation and the minimum fitness threshold of GA;

S12: Generate an initial population of GA, wherein each GA chromosome represents a set of PSO (particle swarm optimization) acceleration coefficients;

S13: Calculating the individual fitness value of GA, specifically including steps:

S131: Set the PSO population size, the maximum number of iterations, and the PSO minimum fitness threshold, wherein each PSO particle represents a set of optimization parameters of the SVM classifier, and the optimization parameters specifically refer to RBF kernel function parameters and penalty factors;

S132: Initialize the acceleration coefficient and inertia weight coefficient, calculate the initial velocity and position of the particle, and initialize the initial value of the simulated annealing algorithm;

S133: Update the particle velocity and position, and calculate the fitness value difference ΔE before and after the update;

S134: Determine whether the fitness value difference is greater than 0, if yes, execute step S136, if not, execute step S135;

S135: Determine whether exp(ΔE/T)>rand(0,1) is established, if yes, execute step S136, if no, execute step S137;

S136: accept the update result, and execute step S138;

S137: Refuse to update the result, roll back particle velocity and position, and execute step S138;

S138: Calculate the adaptive weight coefficient, and update the individual extremum and the global extremum according to the adaptive value;

S139: Determine whether at least one of the following conditions is established:

c. The number of iterations reaches the maximum number of iterations,

d. Reach the PSO minimum fitness threshold,

If no, return to step S133; if yes, find the optimal value, and the optimal value found by PSO is the individual fitness of GA.

S14: Calculate the selection probability of each individual from the obtained individual fitness value, implement selection, crossover and mutation, and generate a new generation of GA population;

S15: Determine whether at least one of the following conditions is established:

a. The evolution algebra reaches the maximum evolution algebra,

b. The individual fitness value reaches the GA minimum fitness threshold,

If no, then return to step S13, if yes, then perform step S16;

S16: Output the optimal value, which is the optimal solution of the hybrid particle swarm optimization algorithm.

S2: Generate the corresponding rough color formula according to the surface color information, and obtain the specified proportion of paint masterbatch based on the rough color formula, mix and stir, and obtain the required paint through feedback compensation, specifically including steps:

S21: Coarse adjustment: Based on the rough adjustment formula, obtain the paint color masterbatch in a specified ratio, mix and stir;

S22: Feedback: get a paint sample from the paint obtained by stirring, and judge whether the color difference between the sample color and the surface color is less than the threshold value, if yes, the blending ends, if no, then execute step S23, specifically including steps :

S221: Apply the stirred paint on a sampling plate made of the same material as the car shell to obtain a paint sample;

S222: The color sensor captures the light emitted by the paint sample under the irradiation of the base light source, and learns the color of the sample according to the captured light;

S223: Judging whether the color difference between the sample color and the surface color is less than a threshold, if yes, then the blending ends, if no, then execute step S23.

S23: fine-tuning: generate a compensation formula according to the color difference between the sample color and the surface color, and obtain a specified proportion of paint masterbatch based on the compensation formula, mix and stir;

S24: Determine whether the modulation result is added to the blending scheme library, if yes, proceed to step S25, if not, proceed to step S26;

S25: The blending plan is obtained after the coarse color-matching formula and the compensation formula are fused, and added to the blending plan library, and step S26 is executed;

S26: Report to the host computer, generate a deployment log, and upload it to the host computer for storage.

As shown in Figure 3, the implementation of the method is as follows:

In the process of implementing the method, the microprocessor 2 can control the batching mechanism 1 to work in coordination, and the batching mechanism 1 includes the link mechanisms such as rough adjustment, fine adjustment, and sampling in the method. The paint color information of the car surface collected by the camera is processed by the microprocessor 2. The acquisition process uses a certain method to eliminate or greatly reduce the external environment factors such as light intensity and surface stains that may affect the modulation process. The microprocessor generates a blending plan through image preprocessing, SVM pattern recognition and other processes. The blending mechanism feeds the material for rough modulation. After the rough adjustment is completed, the modulated sample is sampled and judged. In order to eliminate interference factors, the sample is sprayed on the sampling plate. , put the base light source on the sampling plate, and analyze the color composition through the sensor, instead of directly irradiating the base light source on the sample solution. If the analysis result shows that a certain ingredient is missing, then according to the composition ratio obtained by the analysis, it will be re-modulated by the blending mechanism 1, and the process will be repeated until the modulation result is correct or reaches the allowable error range, then the blending is completed. The scheme is added to the scheme library, reported to the host computer, and the deployment log is generated. During the whole brewing process, the microprocessor is responsible for controlling the blending mechanism and the coordination of each peripheral unit, such as feeding, stirring, touch screen display, communication, etc. It should be noted that the microprocessor also strictly controls the combination behavior of paints with different chemical characteristics, and the optimal deployment scheme and other mechanisms.

Claims (5)

1. a kind of automotive paints mixing color method, it is characterised in that including step：

S1：The surface image of touch-up paint vehicle is treated in collection, and obtains surface color information according to the identification of the surface image of collection,

S2：Corresponding coarse adjustment color formula is generated according to surface color information, and designated ratio is obtained based on the coarse adjustment color formula Color masterbatch is painted, mixes and required paint is obtained by feedback compensation；

Surface color information, the SVM points are obtained according to the identification of the surface image of collection using SVM classifier in the step S1 Class device kernel function is RBF kernel functions；

The optimization process of the SVMs parameter of the SVM classifier uses the hybrid particle swarm based on heredity with simulated annealing Algorithm is carried out, and specifically includes step：

S11：Setting GA population scale, crossing-over rate and aberration rate, the minimum fitness threshold value of maximum evolutionary generation and GA,

S12：GA initial population is generated, wherein, each GA chromosome represents one group of PSO accelerator coefficient；

S13：GA individual fitness is calculated,

S14：The select probability of each individual is calculated by obtained individual fitness, implements selection, intersect and make a variation, generation new one For GA populations,

S15：Judge whether at least one sets up following condition：

A. evolutionary generation reaches maximum evolutionary generation,

B. individual fitness reaches the minimum fitness threshold values of GA,

If it has not, then return to step S13, if it has, then step S16 is performed,

S16：Optimal value is exported, for the optimal solution of the Hybrid Particle Swarm.

2. a kind of automotive paints mixing color method according to claim 1, it is characterised in that the step S2 is specifically wrapped Include step：

S21：Coarse adjustment：Based on coarse adjustment color formula, the paint color masterbatch of designated ratio is obtained, is mixed and stirred for；

S22：Feedback：Sample and obtained between paint sample, and judgement sample color and surface color in the paint obtained from stirring Aberration whether be less than threshold value, if it has, then allotment terminates, if it has not, then performing step S23；

S23：Fine setting：According to the aberration generation compensation formula between color sample and surface color, and obtained based on compensation formula The paint color masterbatch of fetching certainty ratio, is mixed and stirred for.

3. a kind of automotive paints mixing color method according to claim 2, it is characterised in that the step S22 is specific Including step：

S221：By the paint application being stirred in obtaining paint sample on the sampling plate of automobile case same material；

S222：The light that paint sample is sent under the irradiation of base light source is caught by color sensor, learned according to the light of seizure Color sample；

S223：Whether the aberration between judgement sample color and surface color is less than threshold value, if it has, then allotment terminates, if It is no, then perform step S23.

4. a kind of automotive paints mixing color method according to claim 1, it is characterised in that obtained in the step S2 After required paint, modulation daily record is uploaded into host computer and preserved.

5. a kind of automotive paints mixing color method according to claim 1, it is characterised in that the step S13 is specific Including step：

S131：PSO population scales, maximum iteration are set, PSO minimum fitness threshold values, each of which PSO particles are represented The Optimal Parameters of one group of SVM classifier；

S132：Accelerator coefficient, inertia weight coefficient are initialized, particle initial velocity and position is calculated, and initialize simulated annealing Algorithm initial value；

S133：Particle rapidity and position are updated, and calculates adaptation value difference Δ E before and after renewal；

S134：Judge to adapt to whether value difference is more than 0, if it has, then step S136 is performed, if it is not, then performing step S135；

S135：Judge whether exp (Δ E/T) ＞ rand (0,1) set up, if it has, then step S136 is performed, if it has not, then holding Row step S137；

S136：Receive to update result, and perform step S138；

S137：Refusal updates result, rollback particle rapidity and position, and performs step S138；

S138：Adaptive weighting coefficient is calculated, and according to adaptive value more new individual extreme value and global extremum；

S139：Judge whether at least one sets up following condition：

C. iterations reaches maximum iteration,

D. the minimum fitness threshold values of PSO are reached,

If it has not, then return to step S133；It is GA individual by the PSO optimal values found if it has, then obtaining optimal value Fitness.