Rearview mirror detection system based on image recognition and working method

Abstract

A rearview mirror detection system based on image recognition comprises an acquisition area, a transmission assembly, a camera module, a light source system, an image processing unit, a production parameter optimization module, an adjustment assembly and a data storage module. The transmission assembly is used for driving the rearview mirror to an acquisition area, the light source system provides stable illumination for the acquisition area, the camera module is arranged in the acquisition area, the adjusting assembly is used for adjusting the rearview mirror to a standard posture, and the camera module is used for acquiring rearview mirror images and sending the rearview mirror images to the image processing unit. The image processing unit generates a quality score and then sends the quality score to the production parameter optimization module, and the production parameter optimization module is used for optimizing production parameters to achieve the maximum score. According to the method, image recognition and an optimization algorithm based on Q-learning are introduced, so that rearview mirror detection and production optimization are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent manufacturing, and in particular to a rearview mirror detection system and a working method based on image recognition. Background Art

[0002] Traditional rearview mirror quality inspection methods mainly rely on manual operation. Due to the low efficiency of manual inspection and the strong reliance on subjective judgment, long-term work will lead to a decrease in accuracy and insufficient stability of the inspection results. The inspection standards of different workers may differ, which can easily lead to inconsistent quality. At the same time, manual inspection shows limitations in detecting complex appearance defects. Problems such as fine scratches, color differences and reflections are often difficult to identify with the naked eye. The most critical defect is that the traditional inspection method cannot provide real-time feedback and cannot effectively cooperate with parameter adjustments in the production process, resulting in a lag in subsequent quality control, which leads to a large scrap rate. The above method cannot effectively utilize historical data and experience, and it is difficult to achieve adaptive quantitative production. Therefore, in order to solve the above problems, a rearview mirror inspection system and working method based on image recognition are needed. Summary of the invention

[0003] In view of the deficiencies of the prior art, the present invention proposes a rearview mirror detection system and a working method based on image recognition.

[0004] Among them, a specific technical solution of a rearview mirror detection system based on image recognition is as follows:

[0005] A rearview mirror detection system based on image recognition, characterized in that:

[0006] It includes a collection area, a transmission component, a camera module, a light source system, an image processing unit, a production parameter optimization module, an adjustment component and a data storage module;

[0007] The transmission component is used to drive the rearview mirror to the collection area, the light source system provides stable lighting for the collection area, the camera module is arranged in the collection area, the adjustment component is used to adjust the rearview mirror to a standard posture, and the camera module is used to collect the image of the rearview mirror and send it to the image processing unit;

[0008] The image processing unit generates a quality score and sends it to the production parameter optimization module, and the production parameter optimization module is used to optimize the production parameters to achieve the maximum score.

[0009] To better realize the present invention, it can be further:

[0010] A database is provided, which is used to store image data, quality scores and corresponding production parameters of the rearview mirror in the database to facilitate subsequent optimization of the model.

[0011] Further:

[0012] A regression model is provided in the image processing unit, and the image processing unit predicts the quality score by using the regression model.

[0013] Further:

[0014] A Q-learning model is provided in the production optimization module, and the production optimization module optimizes the production parameter group using the Q-learning model.

[0015] A specific technical solution of a working method of a rearview mirror detection system based on image recognition is:

[0016] A working method of a rearview mirror detection system based on image recognition, characterized in that:

[0017] The steps include:

[0018] S1: Start the light source system, which provides a stable light source for the camera module to ensure that the camera module captures a clear image;

[0019] S2: The transmission component drives the rearview mirror to the collection area, and the adjustment component adjusts the position of the rearview mirror on the collection area to ensure that the rearview mirror is in a standard detection position;

[0020] S3: The camera module collects rearview mirror images in real time, and the camera module transmits the collected images to the image processing unit;

[0021] S4: The image processing unit analyzes the image features of the rearview mirror through an image recognition algorithm to generate a quality score;

[0022] S5: the image processing unit inputs the quality score into the production parameter optimization module, and the production parameter optimization module determines the production parameter adjustment direction based on the Q-learning model;

[0023] S6: The production parameter optimization module stores the image data, quality score and corresponding production parameters of the rearview mirror in a database to facilitate subsequent optimization of the model.

[0024] In order to better realize the present invention, it can be further:

[0025] The S5 comprises the following steps:

[0026] S5-1: The production parameter optimization module adopts the following Q value update formula based on the Q-learning model:

[0027] Q(S t,A t )←Q(S t ,A t )+α[R t +γmax A′ Q(S t+1 ,A′)-Q(S t ,A t )]

[0028] Among them, S t is the current set of production parameter combinations;

[0029] A t It is the current adjustment action on production parameters;

[0030] R t is the current image processing unit-based quality score;

[0031] γ is the discount factor, which represents the weight of future returns;

[0032] α is the learning rate;

[0033] max A′ Q(S t+1 ,A′) is the maximum expected reward of the optimal action in the next state;

[0034] S5-2: Set current production parameters P c , the adjusted parameter is P n , the adjustment amount is ΔP, and the production parameter optimization module adjusts the production line parameters in the following manner after the Q value is updated:

[0035] P n =P c +ΔP.

[0036] Further: in S5-1, a regression model is established between the production parameter vector and the quality score, and the production parameter vector in the training set is predicted by the regression model to obtain a combination of production parameter vectors with higher prediction scores. The Q-learning is trial-produced in the screened production parameter vector combinations with higher prediction scores, and the Q value is updated to find the optimal production parameter combination.

[0037] Furthermore: the Q-learning model adopts the Q-learning model in the Stable Baselines3 reinforcement learning library.

[0038] Further: S4-1: the image processing unit pre-processes the image collected by the camera module through a Gaussian filter:

[0039]

[0040] in,

[0041] I(x,y) is the pixel value of the original image at coordinate (x,y);

[0042] G(i,j) is the Gaussian kernel, I′(x,y) is the filtered image;

[0043] S4-2: The image processing unit extracts features related to the quality of the rearview mirror through CNN to obtain a quality feature vector X = [x1, x2, ..., x n ], where x i represents attributes related to quality;

[0044] S4-3: The image processing unit is for x i Perform normalization processing;

[0045] S4-4: The image processing unit inputs the normalized quality feature vector X into the prediction model y=f(X), and the prediction model outputs a quality score.

[0046] Further: the prediction model f(X) adopts a regression model:

[0047] S=ω1x 1′ +ω2x 2′ +...+ω n x n′

[0048] ω1, ω2...ω n is the weight of each feature.

[0049] The beneficial effects of the present invention are as follows: the present invention significantly improves rearview mirror detection and production optimization by introducing image recognition and an optimization algorithm based on Q-learning. The image processing unit performs image feature analysis on the collected rearview mirror image to generate an accurate quality score. The learning process based on the Q-learning model determines the optimal production parameter adjustment direction based on the associated data of historical production parameters and quality scores, reduces the trial and error cost, and realizes the gradual optimization of the production process. By combining the quality score with the Q-learning intelligent optimization model, only a small number of trial samples are needed to adjust a near-optimal combination of production parameters, reducing the waste of resources and time costs caused by a large number of trial productions. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is the overall block diagram of the present invention;

[0051] Figure 2 It is the work flow chart of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] In the description of the present invention, it should be noted that the terms "vertical", "upper", "lower", "horizontal", etc. indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, which are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific position, be constructed and operated in a specific position, and therefore cannot be understood as limiting the present invention. In addition, "first", "second", "third", and "fourth" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0054] like Figure 1 As shown:

[0055] A specific technical solution of a rearview mirror detection system based on image recognition is as follows:

[0056] A rearview mirror detection system based on image recognition includes a collection area, a transmission component, a camera module, a light source system, an image processing unit, a production parameter optimization module, an adjustment component and a database.

[0057] The transmission component is used to drive the rearview mirror to the collection area, the light source system provides stable lighting for the collection area, the camera module is arranged in the collection area, the adjustment component is used to adjust the rearview mirror to a standard posture, the camera module is used to collect the rearview mirror image and send it to the image processing unit, the image processing unit is provided with a regression model, and the image processing unit uses the regression model to predict the quality score.

[0058] The image processing unit generates a quality score and then sends it to the production parameter optimization module. The production parameter optimization module is used to optimize the production parameters to achieve the maximum score. A Q-learning model is set in the production optimization module, and the production optimization module optimizes the production parameter group using the Q-learning model.

[0059] The database is used to store the image data, quality score and corresponding production parameters of the rearview mirror in the database to facilitate subsequent optimization of the model.

[0060] like Figure 2 As shown,

[0061] A specific technical solution of a working method of a rearview mirror detection system based on image recognition is:

[0062] A working method of a rearview mirror detection system based on image recognition comprises the following steps:

[0063] S1: Start the light source system, which provides a stable light source for the camera module to ensure that the camera module captures a clear image;

[0064] S2: The transmission component drives the rearview mirror to the collection area, and the adjustment component adjusts the position of the rearview mirror on the collection area to ensure that the rearview mirror is in a standard detection position;

[0065] S3: The camera module collects rearview mirror images in real time, and the camera module transmits the collected images to the image processing unit;

[0066] S4: The image processing unit analyzes the image features of the rearview mirror through an image recognition algorithm to generate a quality score.

[0067] S4-1: The image processing unit pre-processes the image collected by the camera module through a Gaussian filter:

[0068]

[0069] in,

[0070] I(x,y) is the pixel value of the original image at coordinate (x,y);

[0071] G(i,j) is the Gaussian kernel, I′(x,y) is the filtered image;

[0072] S4-2: The image processing unit extracts features related to the quality of the rearview mirror through CNN to obtain a quality feature vector X = [x1, x2, ..., x n ], where x i represents attributes related to quality;

[0073] S4-3: The image processing unit is for x i Perform normalization processing;

[0074] S4-4: The image processing unit inputs the normalized quality feature vector X into the prediction model y=f(X), and the prediction model outputs a quality score. Specifically, the prediction model f(X) adopts a regression model:

[0075] S=ω1x 1′ +ω2x 2′ +...+ω n x n′

[0076] ω1, ω2...ωn is the weight of each feature.

[0077] S5: The image processing unit inputs the quality score to the production parameter optimization module, and the production parameter optimization module determines the production parameter adjustment direction based on the Q-learning model.

[0078] Specifically, S5-1: a regression model is established between the production parameter vector and the quality score, and the production parameter vector in the training set is predicted by the regression model to obtain a combination of production parameter vectors with higher prediction scores. The Q-learning is trial-produced in the screened production parameter vector combinations with higher prediction scores, and the Q value is updated to find the optimal production parameter combination. At the same time, the Q-learning model adopts the Q-learning model in the Stable Baselines3 reinforcement learning library.

[0079] The production parameter optimization module adopts the following Q value update formula based on the Q-learning model:

[0080] Q(S t ,A t )←Q(S t ,A t )+α[R t +γmax A′ Q(S t+1 ,A′)-Q(S t ,A t )]

[0081] Among them, S t is the current set of production parameter combinations;

[0082] A t It is the current adjustment action on production parameters;

[0083] R t is the current image processing unit-based quality score;

[0084] γ is the discount factor, which represents the weight of future returns;

[0085] α is the learning rate;

[0086] max A′ Q(S t+1 ,A′) is the maximum expected reward of the optimal action in the next state;

[0087] S5-2: Set current production parameters P c , the adjusted parameter is P n , the adjustment amount is ΔP, and the production parameter optimization module adjusts the production line parameters in the following manner after the Q value is updated:

[0088] P=P+ΔP.

[0089] nc

[0090] S6: The production parameter optimization module stores the image data, quality score and corresponding production parameters of the rearview mirror in a database to facilitate subsequent optimization of the model.

[0091] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0092] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A rearview mirror detection system based on image recognition, characterized in that: It includes a collection area, a transmission component, a camera module, a light source system, an image processing unit, a production parameter optimization module, an adjustment component and a data storage module; The transmission component is used to drive the rearview mirror to the collection area, the light source system provides stable lighting for the collection area, the camera module is arranged in the collection area, the adjustment component is used to adjust the rearview mirror to a standard posture, and the camera module is used to collect the image of the rearview mirror and send it to the image processing unit; The image processing unit generates a quality score and sends it to the production parameter optimization module, and the production parameter optimization module is used to optimize the production parameters to achieve the maximum score.

2. The rearview mirror detection system based on image recognition according to claim 1, characterized in that: A database is provided, which is used to store image data, quality scores and corresponding production parameters of the rearview mirror in the database to facilitate subsequent optimization of the model.

3. A rearview mirror detection system based on image recognition according to claim 2, characterized in that: A regression model is provided in the image processing unit, and the image processing unit predicts the quality score by using the regression model.

4. The rearview mirror detection system based on image recognition according to claim 3, characterized in that: A Q-learning model is provided in the production optimization module, and the production optimization module optimizes the production parameter group using the Q-learning model.

5. The working method of the rearview mirror detection system based on image recognition according to claim 4 is characterized in that: The steps include: S1: Start the light source system, which provides a stable light source for the camera module to ensure that the camera module captures a clear image; S2: The transmission component drives the rearview mirror to the collection area, and the adjustment component adjusts the position of the rearview mirror on the collection area to ensure that the rearview mirror is in a standard detection position; S3: The camera module collects rearview mirror images in real time, and the camera module transmits the collected images to the image processing unit; S4: The image processing unit analyzes the image features of the rearview mirror through an image recognition algorithm to generate a quality score; S5: the image processing unit inputs the quality score into the production parameter optimization module, and the production parameter optimization module determines the production parameter adjustment direction based on the Q-learning model; S6: The production parameter optimization module stores the image data, quality score and corresponding production parameters of the rearview mirror in a database to facilitate subsequent optimization of the model.

6. A working method of a rearview mirror detection system based on image recognition according to claim 5, characterized in that: The S5 comprises the following steps: S5-1: The production parameter optimization module adopts the following Q value update formula based on the Q-learning model: Q(S t ,A t )←Q(S t ,A t )+α[R t +γmax A′ Q(S t+1 ,A′)-Q(S t ,A t )] Among them, S t is the current set of production parameter combinations; A t It is the current adjustment action of production parameters; R t is the current image processing unit-based quality score; γ is the discount factor, which represents the weight of future returns; α is the learning rate; max A′ Q(S t+1 ,A′) is the maximum expected reward of the optimal action in the next state; S5-2: Set current production parameters P c , the adjusted parameter is P n , the adjustment amount is ΔP, and the production parameter optimization module adjusts the production line parameters in the following manner after the Q value is updated: P n =P c +ΔP。 7. The working method of the rearview mirror detection system based on image recognition according to claim 6, characterized in that: In S5-1, a regression model is established between the production parameter vector and the quality score. The production parameter vector in the training set is predicted by the regression model to obtain a combination of production parameter vectors with a higher prediction score. The Q-learning is trial-produced in the screened production parameter vector combination with a higher prediction score, and the Q value is updated to find the optimal production parameter combination.

8. The working method of the rearview mirror detection system based on image recognition according to claim 7, characterized in that: The Q-learning model adopts the Q-learning model in the Stable Baselines3 reinforcement learning library.

9. The working method of the rearview mirror detection system based on image recognition according to claim 8, characterized in that: S4-1: The image processing unit pre-processes the image collected by the camera module through a Gaussian filter: in, I(x,y) is the pixel value of the original image at coordinate (x,y); G(i,j) is the Gaussian kernel, I′(x,y) is the filtered image; S4-2: The image processing unit extracts features related to the quality of the rearview mirror through CNN to obtain a quality feature vector X = [x1, x2, ..., x n ], where x i represents attributes related to quality; S4-3: The image processing unit is for x i Perform normalization processing; S4-4: The image processing unit inputs the normalized quality feature vector X into the prediction model y=f(X), and the prediction model outputs a quality score.

10. The working method of the rearview mirror detection system based on image recognition according to claim 9, characterized in that: The prediction model f(X) adopts a regression model: S=ω1x 1′ +ω2x 2′ +...+oh n x n′ ω1, ω2...ω n is the weight of each feature.

Images