A control system and method for adaptively adjusting the intensity of a machine vision light source

Abstract

The present application relates to the field of machine vision technology, and discloses a control system and method for adaptively adjusting the intensity of a machine vision light source. The method comprises: using different multiple first light source intensities to respectively perform video frame acquisition and preprocessing on multiple first objects of different sizes, and obtain target video frame data; establishing a relationship model; when a second object is detected, dynamically adjusting the light source intensity, obtaining the second light source intensity and obtaining real-time video frame data; performing object detection, obtaining object detection data and performing light source control strategy analysis, and obtaining a total light source control strategy; obtaining multiple point light source control nodes and performing control parameter allocation, and obtaining multiple initial point light source control parameters; performing light source intensity control modeling and parameter optimization through multiple initial point light source control parameters, and obtaining multiple target point light source control parameters of the second object. The present application improves the control accuracy of adaptively adjusting the intensity of a machine vision light source.

Description

Technical Field

[0001] The present application relates to the field of machine vision technology, and in particular to a control system and method for adaptively adjusting the intensity of a machine vision light source. Background Art

[0002] Traditional machine vision systems usually use a fixed light intensity, but this method often fails to provide sufficient light for objects of different sizes and materials, resulting in reduced detection accuracy. In addition, too high or too low light intensity will have a negative impact on image quality, increasing image noise or causing object details to be lost.

[0003] In the prior art, there are several major problems with existing machine vision systems. The single light source intensity setting cannot adapt to the diverse detection environments and object characteristics, and it is difficult to maintain detection stability in complex scenarios. Secondly, the traditional light source intensity adjustment method lacks a dynamic adjustment mechanism and cannot be optimized instantly according to the changes in the real-time detection object, resulting in slow system response and poor flexibility. In addition, the manual adjustment of the light source intensity depends on the operator's experience, which is not only inefficient, but also difficult to ensure consistency, increasing the unreliability and operating costs of the system. Summary of the invention

[0004] The present application provides a control system and method for adaptively adjusting the intensity of a machine vision light source, which are used to improve the control accuracy of adaptively adjusting the intensity of a machine vision light source.

[0005] In a first aspect, the present application provides a control method for adaptively adjusting the intensity of a machine vision light source, the control method for adaptively adjusting the intensity of a machine vision light source comprising:

[0006] Using different multiple first light source intensities to respectively perform video frame acquisition and preprocessing on multiple first objects of different sizes to obtain target video frame data;

[0007] Establishing a relationship model between the multiple first light source intensities and the multiple first objects according to the target video frame data;

[0008] When a second object is detected, the light source intensity is dynamically adjusted by using the relationship model and a preset light source intensity threshold to obtain a second light source intensity and acquire real-time video frame data;

[0009] Performing object detection on the real-time video frame data to obtain object detection data, and performing light source control strategy analysis based on the object detection data to obtain a total light source control strategy;

[0010] Acquire multiple point light source control nodes and assign control parameters to the overall light source control strategy to obtain multiple initial point light source control parameters;

[0011] Light source intensity control modeling and parameter optimization are performed using the multiple initial point light source control parameters to obtain multiple target point light source control parameters of the second object.

[0012] In a second aspect, the present application provides a control system for adaptively adjusting the intensity of a machine vision light source, the control system for adaptively adjusting the intensity of a machine vision light source comprising:

[0013] An acquisition module, used to acquire and pre-process video frames of a plurality of first objects of different sizes using a plurality of different first light source intensities, to obtain target video frame data;

[0014] An establishing module, configured to establish a relationship model between the plurality of first light source intensities and the plurality of first objects according to the target video frame data;

[0015] an adjustment module, configured to dynamically adjust the light source intensity by using the relationship model and a preset light source intensity threshold value when a second object is detected, to obtain the second light source intensity and acquire real-time video frame data;

[0016] A detection module, used to perform object detection on the real-time video frame data to obtain object detection data, and perform light source control strategy analysis based on the object detection data to obtain a total light source control strategy;

[0017] An allocation module, used to acquire a plurality of point light source control nodes and allocate control parameters for the light source control master strategy to obtain a plurality of initial point light source control parameters;

[0018] An optimization module is used to perform light source intensity control modeling and parameter optimization through the multiple initial point light source control parameters to obtain multiple target point light source control parameters of the second object.

[0019] In the technical solution provided by the present application, by using different light source intensities to collect video frames of objects of different sizes, multi-step preprocessing is performed. The noise and distortion in the image can be effectively eliminated through the multi-level preprocessing steps. The relationship model between the light source intensity and the object characteristics is established based on the target video frame data, which can accurately reflect the influence of different light source intensities on object detection. Through regression analysis and parameter optimization, the established relationship model has high precision and high reliability, and the light source intensity can be dynamically adjusted to adapt to the changes of different objects and environments. When a new object is detected, the system can adjust the light source intensity in real time through the preset light source intensity threshold and real-time video frame data. The dynamic adjustment mechanism ensures the rapid response capability of the system under different lighting conditions, so that the light source intensity is always kept in the best state. By performing object detection and statistical analysis on real-time video frame data, the system can generate an initial light source control strategy and perform multi-level optimization, and finally form a total light source control strategy. The strategy can be adjusted according to the characteristics of the detected object and real-time monitoring data, ensuring the flexibility and efficiency of light source control, obtaining multiple point light source control nodes, and allocating control parameters to the total light source control strategy. By constructing a control node dependency graph and priority analysis, multi-node collaborative light source control is realized. This distributed control method not only improves the accuracy and stability of light source intensity adjustment, but also enables the system to operate efficiently in complex environments through resource optimization configuration. By transferring dynamic convolutional autoencoders and autoregressive processing, the initial light source control parameters are modeled and optimized, and the spatiotemporal characteristics of the light source intensity variables are extracted. Combined with a shared regression network and dynamic monitoring feedback, the system can perform continuous parameter optimization and online compensation to ensure the accuracy and robustness of the light source control model and adapt to various dynamically changing environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0021] Figure 1 A schematic diagram of an embodiment of a control method for adaptively adjusting the intensity of a machine vision light source in an embodiment of the present application;

[0022] Figure 2 Schematic diagram of an embodiment of a control system for adaptively adjusting the intensity of a machine vision light source in an embodiment of the present application. DETAILED DESCRIPTION

[0023] The embodiment of the present application provides a control system and method for adaptively adjusting the intensity of a machine vision light source. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0024] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 , an embodiment of the control method for adaptively adjusting the intensity of a machine vision light source in the embodiment of the present application includes:

[0025] Step S101, using different first light source intensities to respectively perform video frame acquisition and preprocessing on a plurality of first objects of different sizes to obtain target video frame data;

[0026] It is understandable that the execution subject of the present application may be a control system for adaptively adjusting the intensity of a machine vision light source, or may be a terminal or a server, which is not specifically limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0027] Specifically, a plurality of first objects of different sizes are respectively collected using different first light source intensities to obtain a plurality of initial video frame data. The first objects in the plurality of initial video frame data are marked to obtain the size, position and light source intensity information of each first object. A high-pass filter is used to process the high-frequency noise of the plurality of initial video frame data to obtain a plurality of denoised video frame data. The high-pass filter can effectively remove the high-frequency noise in the video frame data and improve the clarity and stability of the image. The plurality of denoised video frame data are subjected to median filtering to obtain a plurality of filtered video frame data. Median filtering is a nonlinear filtering technology that can effectively remove the impulse noise in the image and improve the quality of the video frame data. The plurality of filtered video frame data are geometrically corrected to obtain a plurality of corrected video frame data. The geometric distortion caused by shooting angle, lens distortion and the like in the video frame data is corrected. The plurality of corrected video frame data are normalized and video frame integrated to obtain the target video frame data. The normalization processing can unify the data of different video frames into a standard range, eliminate the differences caused by different shooting conditions, and make the subsequent processing more convenient and reliable. Video frame integration combines multiple processed video frame data to form a complete set of target video frame data.

[0028] Step S102, establishing a relationship model between a plurality of first light source intensities and a plurality of first objects according to the target video frame data;

[0029] Specifically, object feature extraction is performed on the target video frame data to obtain object feature data, including information such as size, shape, color, and texture. Cluster analysis is performed on the object feature data to obtain the object category of each first object. Cluster analysis is a data mining technology that can effectively identify different object categories in the video frame data by grouping data points with similar features. Each object category represents a group of objects with similar features. A detection model is trained for the object category of each first object to obtain a trained detection model, so that the model learns to identify and classify objects of different categories. Using a validation data set, the detection accuracy of the trained detection model under multiple first light source intensities is evaluated to obtain detection accuracy data. The validation data set is a set of data that has not been used for training. By using these data to evaluate the model under different light source intensities, the detection accuracy of the model in actual application is obtained. Regression analysis is performed on the detection accuracy data to establish a regression model between multiple first light source intensities and multiple first objects. Regression analysis is a statistical method for determining the relationship between variables. The regression model can reflect the changes in the detection accuracy of the model for different objects under different light source intensities. The parameters of the regression model are optimized to obtain a relationship model between multiple first light source intensities and multiple first objects. Parameter optimization can improve the accuracy and applicability of the model. By adjusting the model parameters, the regression model can better reflect the actual situation.

[0030] Step S103: when a second object is detected, dynamically adjust the light source intensity through the relationship model and a preset light source intensity threshold to obtain the second light source intensity and acquire real-time video frame data;

[0031] Specifically, when the second object is detected, the light source intensity requirement data of the second object is obtained, and the required light source intensity range is determined, including the size of the object, surface characteristics, and ambient lighting conditions. According to the light source intensity requirement data, the corresponding initial light source intensity is calculated using the relational model. The initial light source intensity is calculated by inputting the characteristics of the second object into the relational model, and the specific requirements of the second object are converted into a specific light source intensity value. The initial light source intensity is compared with the preset light source intensity threshold to obtain a comparison result. The preset light source intensity threshold is set according to the design requirements of the system and the actual application environment, and defines the upper and lower limits of the light source intensity. The calculated initial light source intensity is compared with these thresholds to ensure that the final light source intensity is within the range that the system can handle. If the initial light source intensity exceeds the preset threshold range, it is necessary to make corresponding adjustments. The light source intensity is dynamically adjusted according to the comparison result to obtain the second light source intensity. According to the result of the threshold comparison, if the initial light source intensity is within the threshold range, the value is directly used as the second light source intensity; if it is not within the range, it is adjusted according to the comparison result to ensure that the final light source intensity value can both meet the needs of the second object and be within the controllable range of the system. The dynamic adjustment process is based on a feedback mechanism, and the light source intensity can be gradually optimized through multiple iterations. According to the second light source intensity, the second object is monitored in real time by video frames to obtain real-time video frame data.

[0032] Step S104: performing object detection on the real-time video frame data to obtain object detection data, and performing light source control strategy analysis based on the object detection data to obtain a total light source control strategy;

[0033] Specifically, the object detection is performed on the real-time video frame data through the trained detection model to obtain the object detection data. The detection model analyzes the real-time video frame data, identifies and marks the objects therein, and extracts the relevant object feature data. These data include detailed information such as the size, position, shape, and color of the object. The object feature data is statistically analyzed to obtain statistical analysis results, such as size distribution, position distribution, and brightness distribution. The light source control strategy is initialized based on the statistical analysis results to obtain the initial light source control strategy. Based on the statistical analysis results, the initial light source intensity adjustment scheme is determined to ensure that the best light source intensity can be provided in most cases. The initial light source control strategy defines the initial setting value of the light source intensity and the adjustment principle under different detection conditions. The initial light source control strategy is population initialized to generate multiple first light source control strategies, and the fitness value of each first light source control strategy is calculated respectively. Multiple different light source control strategies (population individuals) are generated, and the fitness value of each strategy is evaluated. The fitness value is calculated based on the effect of the light source control strategy in actual application. The higher the fitness value, the better the performance of the strategy in object detection. According to the fitness value, multiple first light source control strategies are divided into populations and optimized to obtain multiple second light source control strategies. Light source control strategies with similar fitness values ​​are divided into a group, and each group of strategies is optimized. The optimization process includes operations such as crossover and mutation to generate new light source control strategies. Through multiple iterations, the light source control strategy is gradually optimized so that it can perform well under various detection conditions. Multiple second light source control strategies are optimized to obtain the overall light source control strategy. Through a comprehensive comparison and analysis of all second light source control strategies, the optimal light source control strategy is selected as the overall light source control strategy. The overall strategy can provide the most suitable light source intensity under different detection environments and object conditions, thereby ensuring the efficient and accurate operation of the machine vision system.

[0034] Step S105, obtaining multiple point light source control nodes and allocating control parameters for the overall light source control strategy to obtain multiple initial point light source control parameters;

[0035] Specifically, multiple point light source control nodes are obtained, and the location information of each point light source control node is obtained. Resource evaluation is performed on multiple point light source control nodes according to the location information, and the range of light source intensity that each point light source control node can provide at its current location is analyzed. Considering factors such as the physical characteristics, energy supply status, and ambient light conditions of each node, the adjustable range of its light source intensity is determined. According to the overall light source control strategy, priority and dependency analysis is performed on multiple point light source control nodes to obtain the priority and dependency of each point light source control node. Priority analysis is to determine which point light source control nodes should give priority to adjusting their light source intensity in different situations to achieve the best detection effect. Dependency analysis is to identify the mutual influence and synergy between each point light source control node to ensure that when adjusting the light source intensity of a node, it will not have a negative impact or interference on other nodes. A control node dependency graph of multiple point light source control nodes is constructed according to the priority and dependency. The control node dependency graph is a graphical representation tool for displaying the priority and dependency between each point light source control node. In the graph, each node is marked, and its relationship with other nodes is represented by arrows or other symbols. According to the control node dependency graph and the light source intensity range, control parameters are assigned to multiple point light source control nodes to obtain multiple initial point light source control parameters. The actual situation of each node and the overall light source control strategy are comprehensively considered to ensure that the parameter setting of each node is in line with its own capability range and can be coordinated with the parameters of other nodes, thereby achieving overall optimization.

[0036] Step S106: performing light source intensity control modeling and parameter optimization using a plurality of initial point light source control parameters to obtain a plurality of target point light source control parameters of the second object.

[0037] Specifically, parameter feature extraction is performed on multiple initial point light source control parameters to obtain parameter feature data. Key data that can describe parameter characteristics, such as light source intensity, light source position, light source angle, etc., are extracted from the initial point light source control parameters. The parameter feature data is decomposed into time and space by passing the dynamic convolution autoencoder to obtain time and space feature data. The dynamic convolution autoencoder is a deep learning model that can effectively capture the time and space characteristics of data. By inputting the parameter feature data into the dynamic convolution autoencoder, the model can learn the changing law of the parameter features in the time and space dimensions and generate time and space feature data. Autoregression processing is performed on the time and space feature data to obtain comprehensive feature data. Autoregression processing is a time series analysis method. By performing autoregression analysis on the time and space feature data, the autocorrelation of the data in the time dimension is revealed to obtain comprehensive feature data. Shared regression analysis is performed on the comprehensive feature data to extract the relationship with the light source intensity variable to obtain regression feature data. Shared regression analysis is a multivariate regression method. By analyzing the relationship between the comprehensive feature data and the light source intensity variable, key regression feature data is extracted. The regression feature data reveals the quantitative relationship between the parameter feature data and the light source intensity. The light source intensity control model is modeled by regressing the characteristic data to obtain the light source intensity control model. The light source intensity control model is a mathematical model that establishes a mapping relationship between parameter characteristic data and light source intensity by analyzing and fitting the regression characteristic data. The light source intensity control model is used to optimize the parameters of multiple initial point light source control parameters to obtain multiple target point light source control parameters of the second object. The light source intensity control model is used to iteratively adjust the initial point light source control parameters to find the optimal parameter settings. Through multiple optimization iterations, the adaptability and effectiveness of the parameters are gradually improved, and finally multiple target point light source control parameters that meet the needs of the second object are obtained.

[0038] In the embodiment of the present application, by using different light source intensities to collect video frames of objects of different sizes, multi-step preprocessing is performed. The noise and distortion in the image can be effectively eliminated through the multi-level preprocessing steps. The relationship model between the light source intensity and the object characteristics is established based on the target video frame data, which can accurately reflect the influence of different light source intensities on object detection. Through regression analysis and parameter optimization, the established relationship model has high precision and high reliability, and the light source intensity can be dynamically adjusted to adapt to the changes of different objects and environments. When a new object is detected, the system can adjust the light source intensity in real time through the preset light source intensity threshold and real-time video frame data. The dynamic adjustment mechanism ensures the rapid response capability of the system under different lighting conditions, so that the light source intensity is always kept in the best state. By performing object detection and statistical analysis on real-time video frame data, the system can generate an initial light source control strategy and perform multi-level optimization, and finally form a total light source control strategy. The strategy can be adjusted according to the characteristics of the detected object and real-time monitoring data, ensuring the flexibility and efficiency of light source control, obtaining multiple point light source control nodes, and allocating control parameters to the total light source control strategy. By constructing a control node dependency graph and priority analysis, multi-node collaborative light source control is realized. This distributed control method not only improves the accuracy and stability of light source intensity adjustment, but also enables the system to operate efficiently in complex environments through resource optimization configuration. By transferring dynamic convolutional autoencoders and autoregressive processing, the initial light source control parameters are modeled and optimized, and the spatiotemporal characteristics of the light source intensity variables are extracted. Combined with a shared regression network and dynamic monitoring feedback, the system can perform continuous parameter optimization and online compensation to ensure the accuracy and robustness of the light source control model and adapt to various dynamically changing environments.

[0039] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0040] (1) using a plurality of first light sources with different intensities to respectively acquire video frames of a plurality of first objects of different sizes, to obtain a plurality of initial video frame data;

[0041] (2) marking the first objects in the plurality of initial video frame data to obtain size, position and light source intensity information of each first object;

[0042] (3) Using a high-pass filter, perform high-frequency noise processing on multiple initial video frame data to obtain multiple denoised video frame data;

[0043] (4) performing median filtering on the multiple denoised video frame data to obtain multiple filtered video frame data;

[0044] (5) performing geometric correction on the plurality of filtered video frame data to obtain a plurality of corrected video frame data;

[0045] (6) Performing normalization processing and video frame integration on multiple corrected video frame data to obtain target video frame data.

[0046] Specifically, different first light source intensities are used to respectively capture video frames of multiple first objects of different sizes. Assume that there is an experimental setting in which three different light source intensities are used. to illuminate three objects of different sizes . The camera captures each object under different light intensity and obtains nine sets of initial video frame data. These data contain the image information of each object under different light conditions. The first object in the multiple initial video frame data is marked to obtain the size, position and light intensity information of each first object. In this process, an image processing algorithm is used to identify and mark the object in each frame. For example, an edge detection algorithm is used to identify the outline of the object and then calculate the size of the object. and location ,in Representative At the same time, record the light intensity when shooting ,in Representative Through such marking, a comprehensive data set of size, position and light intensity information of each object under different light intensity is obtained. A high-pass filter is used to process high-frequency noise of multiple initial video frame data to obtain multiple denoised video frame data. The function of the high-pass filter is to remove high-frequency noise in the image, which usually appears as small spots in the image. The image is converted from the spatial domain to the frequency domain using Fourier transform, and then a high-pass filter is applied to filter out the high-frequency part. Assume that the frequency representation of the image is , the high-pass filter can be expressed as , the formula is:

[0047] ;

[0048] in is the frequency representation of the image after passing through the high-pass filter. Perform inverse Fourier transform to obtain a denoised spatial domain image. In this way, high-frequency noise in the image is effectively removed and the main features of the object are retained. Perform median filtering on multiple denoised video frame data to obtain multiple filtered video frame data. Median filtering is a nonlinear filtering technique used to remove impulse noise in an image. The pixel value is replaced by the median of the neighborhood pixel values ​​of a pixel point in the image. For example, for a 3x3 neighborhood, sort its 9 pixel values ​​and replace the central pixel value with the 5th value in the middle. This method can effectively remove sharp noise in the image without blurring the edge of the image. Perform geometric correction on multiple filtered video frame data to obtain multiple corrected video frame data. The purpose of geometric correction is to correct the geometric distortion of the image caused by camera position, angle or lens distortion. Common methods include perspective transformation and lens distortion correction. Perspective transformation can be corrected by calculating the positions of four control points in the image, using the following formula:

[0049] ;

[0050] in, are the coordinates of the original image, are the corrected coordinates, and a, b, c, d, e, f, g, and h are the parameters of the perspective transformation, through which the original image can be mapped to the corrected image. The multiple corrected video frame data are normalized to obtain the target video frame data. The purpose of normalization is to adjust the grayscale value or color value of the image to a uniform range to eliminate the brightness and contrast differences caused by different shooting conditions. The normalized multiple video frame data are integrated to generate a unified target video frame data set. The integration process can be achieved through inter-frame alignment and superposition to ensure that the object position and light source intensity information of each frame are consistent to provide more stable and high-quality video data.

[0051] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0052] (1) Extracting object features from target video frame data to obtain object feature data;

[0053] (2) performing cluster analysis on the object feature data to obtain an object category of each first object;

[0054] (3) training a detection model for each object category of the first object to obtain a trained detection model;

[0055] (4) Using the validation data set, evaluate the detection accuracy of the trained detection model under multiple first light source intensities to obtain detection accuracy data;

[0056] (5) performing regression analysis on the detection accuracy data and generating a regression model between the multiple first light source intensities and the multiple first objects;

[0057] (6) Optimizing the parameters of the regression model to obtain a relationship model between the multiple first light source intensities and the multiple first objects.

[0058] Specifically, first, perform object feature extraction on the target video frame data to obtain object feature data. By analyzing each object in the target video frame, a series of feature information is extracted, such as size, shape, color, texture, position, etc. For example, a camera captures a video frame of products on a production line, and then uses edge detection, color space conversion, shape descriptors and other technologies to extract the features of each product. The object feature data obtained through feature extraction can be represented as a vector ,in Indicates The object feature data is clustered to obtain the object category of each first object. Cluster analysis is an unsupervised learning method used to group similar object feature data. Common clustering algorithms include K-means clustering, hierarchical clustering, DBSCAN, etc. For example, when using the K-means clustering algorithm, the object feature data is clustered. Divide clusters, each cluster represents an object category. By iteratively updating the centroid of the cluster and assigning objects to the nearest cluster, the category label of each object can be finally obtained. ,in Indicates object categories. A detection model is trained for the object category of each first object to obtain a trained detection model. Detection model training is a supervised learning process, in which the model is trained using labeled training data so that it can accurately identify and classify objects. During the training process, object feature data is used as input and category labels are used as output. The model parameters are adjusted by optimizing the loss function to ultimately obtain a trained detection model. For example, a support vector machine or a convolutional neural network is used as a detection model, and the model parameters are optimized using a back propagation algorithm to achieve the best classification performance on the training data. Use a validation data set to evaluate the detection accuracy of the trained detection model under multiple first light source intensities to obtain detection accuracy data. A validation data set is a set of data that has not been used for training. By using this data to evaluate the model under different light source intensities, the detection accuracy of the model can be obtained, thereby verifying the generalization ability of the model, that is, whether high accuracy can be maintained in practical applications. Assume that the validation data set is , where each Contains object features and actual category labels. By calculating the detection accuracy of the model under different light source intensities , get the detection accuracy data ,in Indicates The detection accuracy under the intensity of the light source is determined. The detection accuracy data is subjected to regression analysis and a regression model is generated between the multiple first light source intensities and the multiple first objects. Regression analysis is a statistical method used to determine the relationship between variables. The relationship between the light source intensity and the detection accuracy is modeled using methods such as linear regression or polynomial regression. For example, when a linear regression model is used, the regression equation can be expressed as:

[0059] ;

[0060] in, Indicates Detection accuracy under different light source intensities, Indicates The intensity of the light source, and is the regression coefficient, is the error term. The regression coefficient is estimated by the least square method so that the regression model can accurately describe the relationship between the light source intensity and the detection accuracy. The parameters of the regression model are optimized to obtain a relationship model between the multiple first light source intensities and the multiple first objects. The prediction accuracy and applicability of the model are improved by adjusting the parameters of the regression model so that it can perform well under different objects and light source intensities. For example, the parameters of the regression model are optimized using the gradient descent method or other optimization algorithms to obtain the final relationship model.

[0061] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0062] (1) when a second object is detected, obtaining light source intensity requirement data of the second object;

[0063] (2) Based on the light source intensity requirement data, the corresponding initial light source intensity is calculated using the relational model;

[0064] (3) Using a preset light source intensity threshold, perform a threshold comparison on the initial light source intensity to obtain a comparison result;

[0065] (4) Dynamically adjusting the intensity of the light source based on the comparison result to obtain the second light source intensity;

[0066] (5) Performing real-time video frame monitoring of the second object according to the intensity of the second light source to obtain real-time video frame data.

[0067] Specifically, when the second object is detected, the light intensity requirement data of the second object is obtained. The second object is identified and its relevant features, such as size, surface reflectivity, material, etc., are extracted. These features are used to determine the light intensity required for the object. For example, suppose a new part is detected in an automated inspection system. The surface characteristics and size of the part are analyzed by a camera and an image processing algorithm to determine the light intensity requirement data required. The light intensity requirement data can be represented as a vector ,in Representative The light source intensity requirement value corresponding to each feature. According to the light source intensity requirement data, the corresponding initial light source intensity is calculated using the relational model. The relational model was established in the early stage through a large amount of experimental data and regression analysis to describe the relationship between object features and light source intensity. Assume that the relational model is in the form of ,in Indicates the intensity of the light source, Indicates the light source intensity requirement data, by passing the light source intensity requirement data of the second object Input into the relational model to calculate the initial light source intensity For example, if the relationship model is a linear regression model, the calculation formula can be expressed as:

[0068] ;

[0069] in, and are model parameters, is the light source intensity demand data. The initial light source intensity is compared with the preset light source intensity threshold to obtain the comparison result. The light source intensity threshold is one or more boundary values ​​preset by the system to limit the reasonable range of the light source intensity. The light source intensity is dynamically adjusted according to the comparison result to obtain the second light source intensity. If the comparison result is "appropriate", the initial light source intensity is directly used as the second light source intensity. If the comparison result is "too low" or "too high", the initial light source intensity needs to be adjusted. The adjustment process can increase or decrease the light source intensity so that it falls within the preset threshold range. Through the dynamic adjustment mechanism, it is ensured that the final determined light source intensity is within a reasonable range to meet the light source demand of the second object. According to the second light source intensity, the second object is monitored in real time by video frames to obtain real-time video frame data. The adjusted light source intensity is applied to the actual detection process, and the real-time video frame data of the second object is captured by the camera. The real-time video frame data is not only used for the detection and analysis of the current object, but also can be used as feedback data to further optimize the light source intensity control strategy.

[0070] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0071] (1) Perform object detection on real-time video frame data using the trained detection model to obtain object detection data;

[0072] (2) Perform statistical analysis on the object feature data to obtain statistical analysis results;

[0073] (3) Initialize the light source control strategy through statistical analysis results to obtain the initial light source control strategy;

[0074] (4) Initializing the population of the initial light source control strategy, generating multiple first light source control strategies, and calculating the fitness value of each first light source control strategy respectively;

[0075] (5) According to the fitness value, the plurality of first light source control strategies are divided into populations and the strategies are optimized to obtain a plurality of second light source control strategies;

[0076] (6) Optimize and solve multiple second light source control strategies to obtain the overall light source control strategy.

[0077] Specifically, the object detection is performed on the real-time video frame data through the trained detection model to obtain the object detection data. The detection model identifies and locates the object in the video frame by analyzing the real-time video frame, and extracts its relevant features, such as size, position, shape, etc. The detection model can be a deep learning model, such as a convolutional neural network, which is trained by a large amount of labeled data. In real-time applications, the model automatically analyzes each frame of video and outputs the detection result, including the position, category and relevant feature data of each object. The object feature data is statistically analyzed to obtain the statistical analysis result. By processing and analyzing the detected object feature data, useful information is extracted. For example, the average size, position distribution, shape characteristics, etc. of the object are calculated. The data is processed using a variety of statistical methods, such as mean, variance, standard deviation, etc. The light source control strategy is initialized through the statistical analysis results to obtain the initial light source control strategy. Based on the statistical analysis results, the initial light source intensity setting and control strategy are determined. For example, if the statistical analysis results show that the size and position of the object vary greatly, these factors need to be considered in the light source control strategy to ensure that the light source intensity can be adaptively adjusted at different positions and sizes. The initial light source control strategy can be expressed as a set of parameters, including light source intensity, position, angle, etc. These parameters will serve as the basis for population initialization. The population is initialized for the initial light source control strategy, multiple first light source control strategies are generated, and the fitness value of each first light source control strategy is calculated respectively. By generating multiple different light source control strategies (i.e., population individuals), the optimal light source control scheme is explored. Each light source control strategy can be expressed as a vector ,in Indicates The fitness value is calculated based on the effect of each light source control strategy in actual application. The strategy with higher fitness value indicates better performance in object detection. The fitness value can be calculated by formula, for example:

[0078] ;

[0079] in, Indicates the light source control strategy The smaller the error in object detection, the higher the fitness value. According to the fitness value, multiple first light source control strategies are divided into populations and strategies are optimized to obtain multiple second light source control strategies. Light source control strategies with similar fitness values ​​are divided into a group, and each group of strategies is optimized. The optimization process includes operations such as crossover and mutation to generate new light source control strategies, thereby improving the overall fitness in the next generation population. For example, through the crossover operation, two light source control strategies are combined. and Exchange some parameters to generate a new light source control strategy

[0080] ;

[0081] in, Indicates The first light source control strategy parameters. Multiple second light source control strategies are optimized to obtain the overall light source control strategy. By comprehensively comparing and analyzing all second light source control strategies, the optimal light source control strategy is selected as the overall light source control strategy. The selection of the optimal strategy can be based on the maximization of the fitness value, that is, the light source control strategy with the highest fitness value is selected as the final overall light source control strategy. The overall light source control strategy can provide the most suitable light source intensity under different detection environments and object conditions, thereby ensuring the efficient and accurate operation of the machine vision system.

[0082] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0083] (1) Obtain multiple point light source control nodes and obtain the location information of each point light source control node;

[0084] (2) Perform resource evaluation on multiple point light source control nodes based on location information to obtain the light source intensity range of each point light source control node;

[0085] (3) According to the overall light source control strategy, the priority and dependency relationship of multiple point light source control nodes are analyzed to obtain the priority and dependency relationship of each point light source control node;

[0086] (4) Construct a control node dependency graph of multiple point light source control nodes based on priorities and dependencies;

[0087] (5) According to the control node dependency graph and the light source intensity range, control parameters are assigned to multiple point light source control nodes respectively to obtain multiple initial point light source control parameters.

[0088] Specifically, multiple point light source control nodes are obtained, and the position information of each point light source control node is obtained. In a machine vision system, multiple point light source control nodes are usually distributed in different positions to ensure the uniformity and flexibility of lighting. The position coordinates of each point light source control node are obtained through sensors or other positioning technologies. For example, on a two-dimensional plane, the position of each point light source control node can be expressed as ,in Representative Light source control nodes. Perform resource evaluation on multiple point light source control nodes based on location information to obtain the light source intensity range of each point light source control node. Determine the light source intensity range that each point light source control node can provide at its current location. This involves considering factors such as the physical characteristics of each node, energy supply status, and ambient light conditions, so as to determine the adjustable range of its light source intensity. For example, if the first The minimum and maximum light intensity of each point light source control node are and , then the light source intensity range can be expressed as . According to the overall light source control strategy, the priority and dependency relationship of multiple point light source control nodes are analyzed to obtain the priority and dependency relationship of each point light source control node, and determine which point light source control nodes should give priority to adjusting their light source intensity in different situations to achieve the best detection effect. Dependency analysis is to identify the mutual influence and synergy between each point light source control node to ensure that when adjusting the light source intensity of a node, it will not have a negative impact or interference on other nodes. Through priority and dependency analysis, the role and status of each node in the entire light source control strategy are determined. A control node dependency graph of multiple point light source control nodes is constructed based on the priority and dependency relationship. The control node dependency graph is a graphical representation tool for showing the priority and dependency relationship between each point light source control node. In the graph, each node is marked and its relationship with other nodes is represented by arrows or other symbols. For example, if the priority of node A is higher than that of node B, and the light source intensity adjustment of A will affect B, then in the dependency graph, the arrow points from A to B. According to the control node dependency graph and the light source intensity range, control parameters are assigned to multiple point light source control nodes respectively to obtain multiple initial point light source control parameters. Based on the aforementioned priorities, dependencies, and the light intensity range of each node, specific light intensity parameters are assigned to each point light control node. Taking into account the actual conditions of each node and the overall light control strategy, ensure that the parameter settings of each node are in line with its own capability range and can be coordinated with the parameters of other nodes, so as to achieve overall optimization. Assume that there is a total light control strategy ,Include Point light source control nodes, the light intensity of each node is By controlling the node dependency graph, the control parameter allocation formula for each node is defined:

[0089] ;

[0090] in, and It is The light intensity range of each node, is its priority, is its dependency. For example, in a visual inspection system of an automated production line, there are multiple point light source control nodes distributed at different locations of the production line. The position information of each node is obtained through sensors, such as , Etc. According to the physical characteristics of each node and the ambient lighting conditions, evaluate the light intensity range of each node, for example, the light intensity range of node 1 is [50, 200], and the light intensity range of node 2 is [30, 150]. According to the overall light control strategy, analyze the priority and dependency of each node, for example, the priority of node 1 is higher than that of node 2, and the adjustment of node 1 will affect node 2. Construct a control node dependency graph, in which an arrow is used to indicate that node 1 points to node 2. According to the dependency graph and the light intensity range, assign the control parameters of each node to ensure that the light intensity of node 1 is set to 150 and the light intensity of node 2 is set to 100.

[0091] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0092] (1) Extract parameter features of multiple initial point light source control parameters to obtain parameter feature data;

[0093] (2) Decomposing the parameter feature data in time and space by passing the dynamic convolutional autoencoder to obtain the time and space feature data;

[0094] (3) Perform autoregressive processing on the spatiotemporal feature data to obtain comprehensive feature data;

[0095] (4) Performing shared regression analysis on the comprehensive feature data, extracting the relationship with the light source intensity variable, and obtaining regression feature data;

[0096] (5) Modeling the light source intensity control by regressing the characteristic data to obtain a light source intensity control model;

[0097] (6) Optimizing the control parameters of multiple initial point light sources through the light source intensity control model to obtain the control parameters of multiple target point light sources of the second object.

[0098] Specifically, parameter feature extraction is performed on multiple initial point light source control parameters to obtain parameter feature data. The initial point light source control parameters refer to the light source intensity, position, angle and other parameters used by the system in the initial setting stage. The feature data of these parameters include the specific setting values ​​of each point light source, such as the light source intensity. , location coordinates ,angle By extracting the features of these parameters, a set of feature data vectors describing the light source control state is obtained. ,in Representative The parameter feature values ​​are passed to the dynamic convolutional autoencoder to perform spatiotemporal feature decomposition on the parameter feature data to obtain spatiotemporal feature data. The dynamic convolutional autoencoder is a deep learning model that can effectively capture the spatiotemporal features of data. The model can learn the changing rules of parameter features in time and space dimensions and generate spatiotemporal feature data. Assume that the parameter feature data At the moment The expression is , after being processed by the model, the spatiotemporal feature data is obtained Autoregression processing is performed on the spatiotemporal feature data to obtain comprehensive feature data. Autoregression processing is a time series analysis method that can reveal the autocorrelation of data in the time dimension by performing autoregression analysis on spatiotemporal feature data, thereby obtaining comprehensive feature data. Assume that the spatiotemporal feature data The autoregressive model is:

[0099] ;

[0100] in, is the autoregressive coefficient, is the order of the autoregressive model, is a white noise error term. Through the autoregressive model, the spatiotemporal feature data is converted into comprehensive feature data . Shared regression analysis is performed on the comprehensive feature data to extract the relationship with the light source intensity variable and obtain regression feature data. Shared regression analysis is a multivariate regression method that extracts key regression feature data by analyzing the relationship between the comprehensive feature data and the light source intensity variable. Assume that the light source intensity variable is The comprehensive feature data is , then the shared regression model can be expressed as:

[0101] ;

[0102] in, and is the regression coefficient, It is the first comprehensive feature data eigenvalues, is the error term. The regression coefficient can be estimated by the least squares method, so that the regression model can accurately describe the relationship between the comprehensive characteristic data and the light source intensity. The light source intensity control model is modeled by the regression characteristic data to obtain the light source intensity control model. The light source intensity control model is a mathematical model that establishes a mapping relationship between parameter characteristic data and light source intensity by analyzing and fitting the regression characteristic data. This model can be used to predict the required light source intensity under different conditions and provide guidance for the dynamic adjustment of the light source intensity. Assume that the light source intensity control model is ,in is the parameter feature data. By training the light source intensity control model, the predicted value of the light source intensity under different parameter settings is obtained. The light source intensity control model is used to optimize the parameters of multiple initial point light source control parameters to obtain multiple target point light source control parameters of the second object. Using the light source intensity control model, the initial point light source control parameters are iteratively adjusted to find the optimal parameter settings. Assume that the initial point light source control parameters are , the goal is to find the optimal parameters , so that the light source intensity reaches the expected target value. Parameter optimization can be achieved through gradient descent method or other optimization algorithms, and the optimization process can be expressed as:

[0103] ;

[0104] in, is the target light intensity, is the predicted value of the light source intensity control model.

[0105] The above describes the control method for adaptively adjusting the intensity of the machine vision light source in the embodiment of the present application. The following describes the control system for adaptively adjusting the intensity of the machine vision light source in the embodiment of the present application. Figure 2 In the embodiment of the present application, an embodiment of a control system for adaptively adjusting the intensity of a machine vision light source includes:

[0106] The acquisition module 201 is used to acquire and pre-process video frames of a plurality of first objects of different sizes using a plurality of different first light source intensities to obtain target video frame data;

[0107] An establishing module 202, configured to establish a relationship model between a plurality of first light source intensities and a plurality of first objects according to target video frame data;

[0108] An adjustment module 203 is used to dynamically adjust the light source intensity through the relationship model and a preset light source intensity threshold when a second object is detected, obtain the second light source intensity and acquire real-time video frame data;

[0109] The detection module 204 is used to perform object detection on the real-time video frame data to obtain object detection data, and to perform light source control strategy analysis based on the object detection data to obtain a total light source control strategy;

[0110] An allocation module 205 is used to acquire multiple point light source control nodes and allocate control parameters for the overall light source control strategy to obtain multiple initial point light source control parameters;

[0111] The optimization module 206 is used to perform light source intensity control modeling and parameter optimization through a plurality of initial point light source control parameters to obtain a plurality of target point light source control parameters of the second object.

[0112] Through the cooperation of the above components, video frames of objects of different sizes are collected by using different light source intensities, and multi-step preprocessing is performed. The noise and distortion in the image can be effectively eliminated through multi-level preprocessing steps. The relationship model between light source intensity and object characteristics is established based on the target video frame data, which can accurately reflect the influence of different light source intensities on object detection. Through regression analysis and parameter optimization, the established relationship model has high precision and high reliability, and can dynamically adjust the light source intensity to adapt to the changes of different objects and environments. When a new object is detected, the system can adjust the light source intensity instantly through the preset light source intensity threshold and real-time video frame data. The dynamic adjustment mechanism ensures the rapid response capability of the system under different lighting conditions, so that the light source intensity is always kept in the best state. Through object detection and statistical analysis of real-time video frame data, the system can generate the initial light source control strategy and perform multi-level optimization, and finally form the overall light source control strategy. The strategy can be adjusted according to the characteristics of the detected object and real-time monitoring data, ensuring the flexibility and efficiency of light source control, obtaining multiple point light source control nodes, and allocating control parameters for the overall light source control strategy. By constructing a control node dependency graph and priority analysis, multi-node collaborative light source control is realized. This distributed control method not only improves the accuracy and stability of light source intensity adjustment, but also enables the system to operate efficiently in complex environments through resource optimization configuration. By transferring dynamic convolutional autoencoders and autoregressive processing, the initial light source control parameters are modeled and optimized, and the spatiotemporal characteristics of the light source intensity variables are extracted. Combined with a shared regression network and dynamic monitoring feedback, the system can perform continuous parameter optimization and online compensation to ensure the accuracy and robustness of the light source control model and adapt to various dynamically changing environments.

[0113] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0114] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0115] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A control method for adaptively adjusting the intensity of a machine vision light source, characterized in that: The control method for adaptively adjusting the intensity of a machine vision light source comprises: Using different multiple first light source intensities to respectively perform video frame acquisition and preprocessing on multiple first objects of different sizes to obtain target video frame data; Establishing a relationship model between the multiple first light source intensities and the multiple first objects according to the target video frame data; specifically comprising: performing object feature extraction on the target video frame data to obtain object feature data; performing cluster analysis on the object feature data to obtain the object category of each first object; performing detection model training on the object category of each first object to obtain a trained detection model; using a validation data set to evaluate the detection accuracy of the trained detection model under multiple first light source intensities to obtain detection accuracy data; performing regression analysis on the detection accuracy data and a regression model between the multiple first light source intensities and the multiple first objects; performing parameter optimization on the regression model to obtain a relationship model between the multiple first light source intensities and the multiple first objects; When a second object is detected, the light source intensity is dynamically adjusted by the relationship model and the preset light source intensity threshold, the second light source intensity is obtained and real-time video frame data is acquired; specifically comprising: when a second object is detected, the light source intensity requirement data of the second object is acquired; according to the light source intensity requirement data, the corresponding initial light source intensity is calculated using the relationship model; according to the preset light source intensity threshold, the initial light source intensity is threshold-compared to obtain a comparison result; according to the comparison result, the light source intensity is dynamically adjusted to obtain the second light source intensity; according to the second light source intensity, the video frame of the second object is monitored in real time to obtain real-time video frame data; Performing object detection on the real-time video frame data to obtain object detection data, and performing light source control strategy analysis based on the object detection data to obtain a total light source control strategy; specifically comprising: performing object detection on the real-time video frame data through the trained detection model to obtain object detection data; performing statistical analysis on the object feature data to obtain statistical analysis results; initializing the light source control strategy based on the statistical analysis results to obtain an initial light source control strategy; performing population initialization on the initial light source control strategy to generate multiple first light source control strategies, and calculating the fitness value of each first light source control strategy respectively; performing population division and strategy optimization on the multiple first light source control strategies based on the fitness values ​​to obtain multiple second light source control strategies; performing optimization solutions on the multiple second light source control strategies to obtain a total light source control strategy; Acquire multiple point light source control nodes and assign control parameters to the overall light source control strategy to obtain multiple initial point light source control parameters; Light source intensity control modeling and parameter optimization are performed using the multiple initial point light source control parameters to obtain multiple target point light source control parameters of the second object.

2. The control method for adaptively adjusting the intensity of a machine vision light source according to claim 1, characterized in that: The method of using different first light source intensities to respectively perform video frame acquisition and preprocessing on a plurality of first objects of different sizes to obtain target video frame data includes: Using different first light source intensities to respectively collect video frames of a plurality of first objects of different sizes to obtain a plurality of initial video frame data; Marking the first objects in the plurality of initial video frame data to obtain size, position and light source intensity information of each first object; Using a high-pass filter, performing high-frequency noise processing on the plurality of initial video frame data to obtain a plurality of denoised video frame data; Performing median filtering on the plurality of denoised video frame data to obtain a plurality of filtered video frame data; Performing geometric correction on the plurality of filtered video frame data to obtain a plurality of corrected video frame data; The plurality of corrected video frame data are normalized and video frame integrated to obtain target video frame data.

3. The control method for adaptively adjusting the intensity of a machine vision light source according to claim 1, characterized in that: The step of acquiring a plurality of point light source control nodes and allocating control parameters to the overall light source control strategy to obtain a plurality of initial point light source control parameters includes: Obtain multiple point light source control nodes, and obtain the position information of each point light source control node; Perform resource evaluation on the multiple point light source control nodes according to the location information to obtain a light source intensity range of each point light source control node; According to the overall light source control strategy, priority and dependency analysis is performed on the multiple point light source control nodes to obtain the priority and dependency of each point light source control node; Constructing a control node dependency graph of the plurality of point light source control nodes according to the priorities and the dependency relationships; According to the control node dependency graph and the light source intensity range, control parameters are respectively allocated to the multiple point light source control nodes to obtain multiple initial point light source control parameters.

4. The control method for adaptively adjusting the intensity of a machine vision light source according to claim 1, characterized in that: The light source intensity control modeling and parameter optimization are performed by using the multiple initial point light source control parameters to obtain multiple target point light source control parameters of the second object, including: Extracting parameter features of the multiple initial point light source control parameters to obtain parameter feature data; Decomposing the parameter feature data in time and space by passing a dynamic convolutional autoencoder to obtain time and space feature data; Performing autoregressive processing on the spatiotemporal feature data to obtain comprehensive feature data; Performing a shared regression analysis on the comprehensive feature data, extracting the relationship with the light source intensity variable, and obtaining regression feature data; Performing light source intensity control modeling through the regression feature data to obtain a light source intensity control model; The multiple initial point light source control parameters are optimized using the light source intensity control model to obtain multiple target point light source control parameters of the second object.

5. A control system for adaptively adjusting the intensity of a machine vision light source, characterized in that: The control system for adaptively adjusting the intensity of the machine vision light source includes: An acquisition module, used to acquire and pre-process video frames of a plurality of first objects of different sizes using a plurality of different first light source intensities, to obtain target video frame data; An establishment module is used to establish a relationship model between the multiple first light source intensities and the multiple first objects according to the target video frame data; specifically comprising: performing object feature extraction on the target video frame data to obtain object feature data; performing cluster analysis on the object feature data to obtain the object category of each first object; performing detection model training on the object category of each first object to obtain a trained detection model; using a validation data set to evaluate the detection accuracy of the trained detection model under multiple first light source intensities to obtain detection accuracy data; performing regression analysis on the detection accuracy data and a regression model between the multiple first light source intensities and the multiple first objects; performing parameter optimization on the regression model to obtain a relationship model between the multiple first light source intensities and the multiple first objects; The adjustment module is used to dynamically adjust the light source intensity through the relationship model and the preset light source intensity threshold when the second object is detected, obtain the second light source intensity and obtain real-time video frame data; specifically includes: when the second object is detected, obtain the light source intensity requirement data of the second object; according to the light source intensity requirement data, use the relationship model to calculate the corresponding initial light source intensity; according to the preset light source intensity threshold, perform threshold comparison on the initial light source intensity to obtain a comparison result; dynamically adjust the light source intensity according to the comparison result to obtain the second light source intensity; according to the second light source intensity, perform real-time video frame monitoring on the second object to obtain real-time video frame data; A detection module is used to perform object detection on the real-time video frame data to obtain object detection data, and to perform light source control strategy analysis based on the object detection data to obtain a total light source control strategy; specifically comprising: performing object detection on the real-time video frame data through the trained detection model to obtain object detection data; performing statistical analysis on the object feature data to obtain statistical analysis results; initializing the light source control strategy based on the statistical analysis results to obtain an initial light source control strategy; performing population initialization on the initial light source control strategy to generate multiple first light source control strategies, and calculating the fitness value of each first light source control strategy respectively; performing population division and strategy optimization on the multiple first light source control strategies based on the fitness values ​​to obtain multiple second light source control strategies; performing optimization solutions on the multiple second light source control strategies to obtain a total light source control strategy; An allocation module, used to acquire a plurality of point light source control nodes and allocate control parameters for the light source control master strategy to obtain a plurality of initial point light source control parameters; An optimization module is used to perform light source intensity control modeling and parameter optimization through the multiple initial point light source control parameters to obtain multiple target point light source control parameters of the second object.

Images