An unmanned aerial vehicle camera lens production manufacturing online monitoring analysis management method

Abstract

The application discloses a kind of unmanned aerial vehicle camera lens production manufacturing online monitoring analysis management method, and the application relates to unmanned aerial vehicle technical field.The method steps include: combining the pearson correlation and feature mutual information quantity between unmanned aerial vehicle lens feature data and unmanned aerial vehicle lens production environment data, obtain the environmental quality importance index and screen unmanned aerial vehicle lens production environment data, obtain unmanned aerial vehicle lens production environment screening data;Multivariate regression tree model is constructed;Unmanned aerial vehicle lens feature data and the feature weight obtained by analytic hierarchy process are combined, obtain lens evaluation index and classify unmanned aerial vehicle lens feature data by kernel density estimation method, obtain unmanned aerial vehicle lens label;When unmanned aerial vehicle lens label is corresponding unmanned aerial vehicle lens production environment screening data that needs to be rechecked, input multivariate regression tree model, obtain prediction error result;Combined with error prediction result and environmental quality importance index, obtain environmental quality correlation index, carry out traceability early warning.

Description

A method for online monitoring, analysis and management of drone camera lens manufacturing Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) technology, specifically to an online monitoring, analysis, and management method for the production and manufacturing of UAV camera lenses. Background Technology

[0002] With the rapid development of drone technology, drones are being used more and more widely in various fields, especially in photography and surveillance. As one of their core components, drone cameras have extremely high requirements for image quality, stability, and accuracy. Therefore, the manufacturing quality of drone camera lenses directly affects the performance of the entire system. Controlling quality parameters such as image sharpness, distortion, and color deviation is particularly important during the manufacturing process of drone lenses. As market demand for drone products increases, lens manufacturers face higher precision and quality requirements. How to promptly identify and correct quality problems during production has become a key issue in the industry.

[0003] Currently, existing drone lens inspection technologies lack in-depth analysis of the relationship between environmental factors such as temperature and humidity and lens quality, thus failing to provide effective feedback and optimization suggestions for the production process, resulting in inconsistent production quality. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an online monitoring, analysis, and management method for the production and manufacturing of drone camera lenses, thereby resolving the problems mentioned in the background section.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an online monitoring, analysis, and management method for the production and manufacturing of drone camera lenses, comprising the following steps:

[0006] Step S1: Collect lens data from the drone camera to obtain drone lens image data;

[0007] Step S2: Extract features from the UAV lens image data to obtain UAV lens feature data;

[0008] Step S3: Collect production environment data of UAV lenses to obtain UAV lens production environment data. Calculate the Pearson correlation between the UAV lens production environment data and the UAV lens feature data to obtain the environmental quality correlation coefficient. Calculate the feature mutual information between the UAV lens production environment data and the UAV lens feature data to obtain the environmental quality mutual information quantity. Combine the environmental quality correlation coefficient and the environmental quality mutual information quantity to calculate the environmental quality importance index.

[0009] Step S4: Filter the drone lens production environment data using the environmental quality importance index to obtain drone lens production environment screening data. Construct a multivariate regression tree model using the drone lens production environment screening data and drone lens feature data.

[0010] Step S5: Perform hierarchical analysis on the drone lens feature data using the analytic hierarchy process (AHP) to obtain the feature weights of the drone lens feature data. By combining the drone lens feature data and feature weights, calculate the lens evaluation index of the drone lens.

[0011] Step S6: Based on the lens evaluation index, the characteristic data of the UAV lens are classified using the kernel density estimation method to obtain UAV lens labels; when the UAV lens label requires re-inspection, the UAV lens production environment screening data corresponding to the UAV lens label is input into the multivariate regression tree model to obtain the prediction error result; by combining the error prediction result and the environmental quality importance index, the environmental quality correlation index is calculated; if the environmental quality correlation index is greater than a preset threshold, a source tracing and early warning is issued to realize the monitoring of UAV camera lenses.

[0012] Preferably, the step of extracting features from the UAV lens image data to obtain UAV lens feature data includes the following specific steps:

[0013] The texture detail retention of drone lens image data is quantified by calculating the average gradient magnitude of edge regions, thus determining the film's ability to preserve detail. The larger the average gradient magnitude, the clearer the texture details.

[0014]

[0015] TDR stands for Texture Detail Retention. Represents the pixels in the drone lens image data grayscale value, This indicates that the drone lens image data is in The gradient vector at that point, This represents the edge region defined in the calibration diagram. Represents the edge region defined in the calibration diagram. The total number of pixels within;

[0016] The color shift in drone lens image data measures the lens's color reproduction capability; a smaller value indicates a smaller color shift.

[0017]

[0018] Where CD represents the color offset. This represents the measured color value of the l-th color wheel region in the UAV lens image data. This represents the theoretical color value of the l-th color wheel in the calibration diagram. This represents the CIEDE2000 color difference formula, used to calculate the difference between measured and theoretical colors. This indicates the total number of color wheels; the default total number of color wheels is 24.

[0019] The edge distortion of drone lens image data is quantified by the mean Euclidean distance of coordinate offset. The larger the value, the more severe the distortion.

[0020]

[0021] in, For edge distortion degree, This indicates the distortion grid in the drone lens image data. The actual coordinates of the intersection points This represents the theoretical coordinates of the corresponding intersection points in the calibration diagram. Indicates the total number of grid intersections;

[0022] The resolution retention rate of drone lens image data is a key indicator. A resolution retention rate ratio closer to 1 indicates that the lens resolution is closer to the ideal value; a ratio less than 1 indicates a decrease in resolution.

[0023]

[0024] Where RR represents the resolution retention rate of the UAV lens image data. This represents the minimum resolvable linewidth in the drone lens image data. This indicates the minimum theoretical resolution linewidth of the calibration diagram design;

[0025] The grayscale distribution equalization value in drone lens image data. The closer the grayscale distribution equalization value is to 1, the more uniform the grayscale distribution; if it is close to 0, there is obvious brightness unevenness.

[0026]

[0027] Wherein, GU is the gray-level distribution equilibrium value. This represents the average grayscale value of pixels in the grayscale region of the drone lens image data. This represents the standard deviation of pixel grayscale values ​​in the grayscale region of the drone lens image data. This is a constant term, with a range of (0, 1e-3);

[0028] Finally, the drone lens feature data is obtained. Each drone lens feature data point is represented by H, where H = [ , , , , ].

[0029] Preferably, the step of obtaining the environmental quality correlation coefficient by calculating the Pearson correlation between UAV lens production environment data and UAV lens feature data includes the following steps:

[0030] For environmental parameters in the drone lens production environment data and lens characteristics in the drone lens characteristic data, calculate their Pearson correlation to obtain the environmental quality correlation coefficient:

[0031]

[0032] in, This represents the environmental quality correlation coefficient between the k-th environmental parameter and the j-th lens feature. For the i-th sample value of the k-th environmental parameter, This represents the average value of the k-th environmental parameter. This represents the i-th sample value of the j-th lens feature. Let represent the average value of the j-th lens feature, n represent the total number of samples, and i represent the index of the i-th sample value.

[0033] Preferably, the step of obtaining the environmental quality mutual information by calculating the feature mutual information of UAV lens production environment data and UAV lens feature data includes the following steps:

[0034] By calculating the mutual information of features between drone lens production environment data and drone lens feature data, the environmental quality mutual information is obtained:

[0035]

[0036] in, This represents the environmental quality mutual information between the k-th environmental parameter and the j-th lens feature. This represents the set of values ​​for the k-th environmental parameter. i is One of the elements, Let represent the set of values ​​for the j-th lens feature. for One of the elements, Indicates environmental parameters Values And lens features Values The probability, Indicates environmental parameters Values The probability, The lens feature value is indicated. The probability of.

[0037] Preferably, the step of calculating the environmental quality importance index by combining the environmental quality correlation coefficient and the environmental quality mutual information includes the following specific steps:

[0038] The environmental quality importance index is calculated by combining the environmental quality correlation coefficient and the environmental quality mutual information.

[0039]

[0040] in, This represents the environmental quality importance index of the k-th environmental parameter and the j-th lens feature. This represents the environmental quality mutual information between the k-th environmental parameter and the j-th lens feature. This represents the environmental quality correlation coefficient between the k-th environmental parameter and the j-th lens feature, and max() represents the maximum value function. express This represents the maximum value of the environmental quality mutual information between the k-th environmental parameter and the j-th lens feature. , For adjustment coefficients, Used to adjust the influence of environmental quality mutual information, ranging from [0,1]. The environmental quality importance index is used to prevent excessive amplification when there is high correlation and high mutual information, and its range is [0, 0.5].

[0041] Preferably, the step of filtering the drone lens production environment data using the environmental quality importance index to obtain drone lens production environment screening data includes the following specific steps:

[0042] The production environment data for the drone lenses is filtered using the environmental quality importance index, and each environmental parameter is calculated first. The importance of overall lens quality for drones:

[0043]

[0044] in, Represents the k-th environmental parameter The importance of the overall lens quality of drones This represents the environmental quality importance index of the k-th environmental parameter and the j-th lens feature. Indicating the first element in the drone lens feature data The index of each feature, where m represents the number of features in the UAV lens feature data;

[0045] Environmental parameters with importance scores higher than a preset threshold in the drone lens production environment data are retained to obtain drone lens production environment screening data.

[0046] Preferably, the step of constructing a multivariate regression tree model by screening data from the drone lens production environment and drone lens feature data includes the following specific steps:

[0047] Using the drone lens production environment screening data and drone lens feature data, a multivariate regression tree model is constructed. For the sample set within the leaf node t of the multivariate regression tree model... The weighted mean square error is calculated as follows:

[0048]

[0049] in: Indicates the current node The weighted mean square error, For the current node Number of samples within, Let be the average value of the j-th lens feature within node t. Let represent the i-th sample value of the j-th lens feature, where i represents the index of the i-th sample value;

[0050] In the recursive segmentation and splitting criterion, all environmental parameters are traversed as candidate splitting points, and based on each candidate splitting point... The sample data of the current node Divide the node into left and right child nodes, and calculate the weighted mean square error after splitting:

[0051]

[0052] in, This represents the change in weighted mean square error before and after the split. Indicates the current leaf node The weighted mean square error, For the current node Number of samples within, Indicates the current node The number of samples in the left child node after the split. For the current node The number of samples in the right child node after the split. This represents the weighted mean square error of the left child node. This represents the weighted mean square error of the right child node;

[0053] Among all candidate split points, the split point that maximizes the weighted mean square error is selected as the split rule for the current node. When the number of samples in a node is lower than the preset minimum sample threshold, the splitting stops and the node is marked as a leaf node, thus forming a multivariate regression tree model.

[0054] Preferably, the step of performing hierarchical analysis on the drone lens feature data using the analytic hierarchy process (AHP) to obtain the feature weights of the drone lens feature data, and then calculating the lens evaluation index of the drone lens by combining the drone lens feature data and the feature weights, includes the following specific steps:

[0055] The feature weights of the UAV lens feature data are obtained by performing hierarchical analysis using the analytic hierarchy process (AHP).

[0056]

[0057] in, This represents the feature weight of the j-th feature in the UAV lens feature data, and m represents the number of features in the UAV lens feature data. Indicating the first element in the drone lens feature data Index of features, This represents the ratio of the relative importance of the j2nd feature to the jth feature in the UAV lens feature data, where j2 is the index of the j2nd feature in the UAV lens feature data;

[0058] By combining the drone lens feature data and feature weights, the lens evaluation index for each drone lens is calculated:

[0059]

[0060] in, Let m be the lens evaluation index, and i be the index of the j-th feature in the UAV lens feature data. This represents the feature weight of the j-th feature in the drone lens feature data. This represents the j-th feature in the drone lens feature data.

[0061] Preferably, the step of classifying the UAV lens feature data based on the lens evaluation index using kernel density estimation to obtain UAV lens labels includes the following specific steps:

[0062] The probability density distribution of the lens evaluation index is calculated using the Gaussian kernel function:

[0063]

[0064] in, This indicates the Gaussian kernel function in the target lens evaluation index value. Kernel density estimation at [location] The larger the value, the higher the target lens evaluation index. The more lenses, the better. Here, is the Gaussian kernel function, h is the bandwidth of the Gaussian kernel function, used to control the smoothing degree, n is the number of samples in the UAV lens feature data, and i represents the index of the i-th sample in the UAV lens feature data. The target lens evaluation index value, Let i be the evaluation index of the i-th lens;

[0065] By calculating the probability density distribution of the lens evaluation index, a probability density curve is plotted with the target lens evaluation index as the horizontal axis and the kernel density estimate of the target evaluation index as the vertical axis. The drone lens feature data corresponding to each lens evaluation index is classified, and the drone lens labels are obtained as follows: excellent, usable, and require re-inspection.

[0066] Preferably, the step of calculating the environmental quality correlation index by combining the error prediction results and the environmental quality importance index includes the following specific steps:

[0067] By combining the error prediction results and the environmental quality importance index, an environmental quality correlation index is calculated. This correlation index is then used to monitor and analyze the lenses of drone cameras.

[0068]

[0069] in, This represents the environmental quality correlation index between the k-th environmental parameter and the j-th lens feature, and the prediction error result. This represents the environmental quality importance index of the k-th environmental parameter and the j-th lens feature. This represents the j-th prediction error result for the lens that needs to be re-examined.

[0070] Beneficial effects

[0071] This invention provides an online monitoring, analysis, and management method for the production and manufacturing of drone camera lenses, involving machine learning and deep learning technologies, which has the following beneficial effects:

[0072] (1) By combining the feature data and feature weights of the UAV lenses, the calculated lens evaluation index can provide a quantitative quality assessment for each lens. This index comprehensively considers various physical properties and optical performance of the lens, enabling the production line to understand the overall quality level of the lenses in real time. Using this evaluation index, different lenses can be accurately classified, providing a scientific basis for subsequent quality control and re-inspection, thereby improving the product qualification rate in the production process and reducing rework and losses caused by quality defects.

[0073] (2) By combining the environmental quality correlation coefficient and the environmental quality mutual information, the calculated environmental quality importance index can comprehensively assess the impact of the production environment on lens quality. Traditional quality monitoring often focuses on the lens characteristics themselves, neglecting the potential impact of changes in the production environment on the final quality. By introducing the environmental quality importance index, it is possible to accurately identify which environmental factors have the greatest impact on lens quality during the production process, thereby helping production managers to adjust production environment parameters in a timely manner, optimize the production process, and improve the consistency and stability of lens quality.

[0074] (3) By combining error prediction results and the environmental quality importance index, the calculated environmental quality correlation index can further reveal the complex relationship between environmental parameters and lens characteristics. This index can quantify the impact of environmental parameter changes on lens quality prediction errors and provide actionable optimization guidance for the production process. When the environmental quality correlation index is higher than the preset threshold, it indicates that certain environmental factors have a significant impact on lens quality, and these environmental factors should be given priority and adjusted during the production process. This method can accurately predict quality fluctuations, identify potential quality problems in advance, effectively reduce production risks, and improve the production efficiency and product quality of UAV lenses. Attached Figure Description

[0075] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0076] Figure 1 is a flowchart of the steps of an online monitoring, analysis and management method for the production and manufacturing of drone camera lenses proposed in this invention;

[0077] Figure 2 is a step-by-step diagram of an online monitoring, analysis and management method for the production and manufacturing of drone camera lenses proposed in this invention. Detailed Implementation

[0078] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0079] Please refer to Figures 1-2. This invention provides a technical solution: an online monitoring, analysis and management method for the production and manufacturing of drone camera lenses.

[0080] Step S1: Collect lens data from the drone camera to obtain drone lens image data.

[0081] At the end of the lens production line, on the conveyor belt before the lenses enter the packaging section after assembly, a high-precision calibration plate is installed. The calibration plate is printed with calibration diagrams including resolution test patterns, distortion correction grids, and grayscale color wheels, used for subsequent quantitative analysis of imaging performance. A multi-band LED light source array, covering visible and infrared bands, is symmetrically arranged on both sides of the calibration plate. The combined output of different bands is controlled by a program to simulate the lighting scenarios of the lenses in actual applications. The light source employs a diffused design to ensure uniform brightness on the calibration plate surface and avoid localized shadows or reflections. A high-speed linear scan camera is vertically mounted directly above the calibration plate, its optical axis strictly perpendicular to the plane of the calibration plate. It is periodically calibrated using a laser calibration tool to eliminate geometric distortion caused by viewing angle tilt. The camera is linked to a high-precision encoder on the conveyor belt drive shaft. When the lens moves at a constant speed along the conveyor belt to the front of the calibration plate, the encoder monitors the lens position in real time and triggers the camera to perform millisecond-level synchronous shooting. The camera's exposure time, gain value, and other parameters are pre-fixed to avoid affecting image consistency due to parameter fluctuations. During shooting, a sealed light shield isolates stray light from the outside, allowing only the calibration light source to illuminate the image, ensuring a clean light path. As each lens passes through the calibration plate, the line-scan camera captures multiple images in a line-by-line scanning manner, covering the imaging effects under different light source wavelengths. The raw images are stored in lossless formats (such as RAW or 16-bit TIFF) and associated with metadata such as the lens's unique number, shooting timestamp, light source mode, and environmental parameters (temperature, humidity), and uploaded to a central database for categorized management. The system verifies image integrity in real time; if blurring, misalignment, or missing calibration patterns are detected, a reshoot mechanism is immediately triggered. Simultaneously, the calibration plate is cleaned and its position calibrated weekly, and the light source spectrum and camera response are calibrated monthly to ensure long-term operational stability. Through this series of standardized processes, all lenses complete imaging data acquisition under completely consistent testing conditions, providing a highly consistent and traceable UAV lens dataset for subsequent quality analysis.

[0082] Step S2: Extract features from the UAV lens image data to obtain UAV lens feature data.

[0083] Feature extraction is performed on the drone lens image data to obtain drone lens feature data, including texture detail retention, color offset, edge distortion, resolution retention, and grayscale distribution balance.

[0084] The texture detail retention of drone lens image data is quantified by calculating the average gradient magnitude of edge regions, thus determining the film's ability to preserve detail. The larger the average gradient magnitude, the clearer the texture details.

[0085]

[0086] TDR stands for Texture Detail Retention. Represents the pixels in the drone lens image data grayscale value, This indicates that the drone lens image data is in The gradient vector at that point, This represents the edge region defined in the calibration diagram. Represents the edge region defined in the calibration diagram. The total number of pixels within.

[0087] It should be noted that the calibration map in this article is a standardized pattern tool specifically designed for online monitoring in the production of drone camera lenses. It typically consists of modules such as a high-precision printed or engraved resolution test map, distortion correction grid, grayscale color wheel, and auxiliary alignment markers, covering the visible and infrared bands. These modules provide a quantitative benchmark for quality analysis based on the actual performance of the lens after imaging. The calibration map is fixed at a specific location on the production line, working with multi-band light sources and high-speed linear scan cameras to ensure the acquisition of lens imaging data under constant illumination and a vertical shooting angle. Its physical design utilizes highly stable materials (such as ceramics or optical glass) and maintains long-term accuracy through regular cleaning and calibration (such as light source spectrum verification and position alignment). The calibration map not only serves as a reference standard for quality inspection but also provides a data foundation for subsequent feature extraction (such as edge gradient and color difference calculation), ultimately achieving objective evaluation and closed-loop optimization of quality parameters during lens production.

[0088] It should be noted that the edge region defined in the calibration diagram consists of sharp lines or geometric patterns alternating between black and white. The line width decreases in a logarithmic gradient, and the minimum line width corresponds to the theoretical resolution limit of the lens.

[0089] The color shift in drone lens image data measures the lens's color reproduction capability; a smaller value indicates a smaller color shift.

[0090]

[0091] Where CD represents the color offset. This represents the measured color value of the l-th color wheel region in the UAV lens image data. This represents the theoretical color value of the l-th color wheel in the calibration diagram. This represents the CIEDE2000 color difference formula, used to calculate the difference between measured and theoretical colors. This indicates the total number of color wheels, which defaults to 24.

[0092] It should be noted that the color wheel is a key module in the calibration diagram used to test the color reproduction capability of a lens.

[0093] It includes a 24-color standard color chart, covering primary colors such as red, green, and blue, as well as intermediate tones. The color blocks are evenly distributed according to the Lab color space, and each color block is labeled with a theoretical color value (such as the sRGB standard) for comparison with the measured values ​​in lens imaging.

[0094] The edge distortion of drone lens image data is quantified by the mean Euclidean distance of coordinate offset. The larger the value, the more severe the distortion.

[0095]

[0096] in, For edge distortion degree, This indicates the distortion grid in the drone lens image data. The actual coordinates of the intersection points This represents the theoretical coordinates of the corresponding intersection points in the calibration diagram. This represents the total number of grid intersections.

[0097] It should be noted that the distortion grid is a geometric pattern used in the calibration diagram to quantify optical distortion. Grid types: linear grid: a uniform grid composed of horizontal and vertical straight lines, with a spacing of 5% to 10% of the lens field of view (e.g., one line every 10 mm); concentric circle grid: concentric circles and radial lines distributed radially with the center of the calibration diagram as the origin, used to detect radial distortion (e.g., barrel distortion).

[0098] The resolution retention rate of drone lens image data is a key indicator. A resolution retention rate ratio closer to 1 indicates that the lens resolution is closer to the ideal value; a ratio less than 1 indicates a decrease in resolution.

[0099]

[0100] Where RR represents the resolution retention rate of the UAV lens image data. This represents the minimum resolvable linewidth in the drone lens image data. This indicates the minimum theoretical linewidth required for calibration diagram design.

[0101] The grayscale distribution equalization value in drone lens image data. The closer the grayscale distribution equalization value is to 1, the more uniform the grayscale distribution; if it is close to 0, there is obvious brightness unevenness.

[0102]

[0103] Wherein, GU is the gray-level distribution equilibrium value. This represents the average grayscale value of pixels in the grayscale region of the drone lens image data. This represents the standard deviation of pixel grayscale values ​​in the grayscale region of the drone lens image data. This is a constant term, with a range of (0, 1e-3).

[0104] Finally, the drone lens feature data is obtained. Each drone lens feature data point is represented by H, where H = [ , , , , ].

[0105] Step S3: Collect production environment data of UAV lenses to obtain UAV lens production environment data. Calculate the Pearson correlation between the UAV lens production environment data and the UAV lens feature data to obtain the environmental quality correlation coefficient. Calculate the feature mutual information between the UAV lens production environment data and the UAV lens feature data to obtain the environmental quality mutual information quantity. Combine the environmental quality correlation coefficient and the environmental quality mutual information quantity to calculate the environmental quality importance index.

[0106] Collect environmental data on drone lens production, covering basic environmental indicators such as temperature, humidity, and air pressure in the production workshop. Examples include real-time temperature during the molding process, cooling rate during injection molding, and cleanliness levels (e.g., dust particle concentration) in the coating workshop. These parameters directly affect the physical properties and processing precision of the lens materials. Record key process parameters of the production equipment, such as mold cavity pressure and mold temperature in the molding equipment, film thickness, coating material purity, and rotation offset angle in the coating equipment, as well as positioning accuracy and welding temperature during lens assembly. These data are directly related to the lens's geometry and optical performance.

[0107] By using a unique lens number, the production environment data of the drone lens is linked with the characteristic data of the drone lens, ensuring that the data is aligned according to production batch and time sequence.

[0108] For environmental parameters in the drone lens production environment data and lens characteristics in the drone lens characteristic data, calculate their Pearson correlation to obtain the environmental quality correlation coefficient:

[0109]

[0110] in, This represents the environmental quality correlation coefficient between the k-th environmental parameter and the j-th lens feature. For the i-th sample value of the k-th environmental parameter, This represents the average value of the k-th environmental parameter. This represents the i-th sample value of the j-th lens feature. Let represent the average value of the j-th lens feature, n represent the total number of samples, and i represent the index of the i-th sample value.

[0111] It should be noted that the environmental quality correlation coefficient is used to quantify the degree of linear correlation between production environment parameters (such as temperature, humidity, cleanliness level, etc.) and drone lens quality characteristics (such as texture detail retention, color shift, etc.).

[0112] By calculating the mutual information of features between drone lens production environment data and drone lens feature data, the environmental quality mutual information is obtained:

[0113]

[0114] in, This represents the environmental quality mutual information between the k-th environmental parameter and the j-th lens feature. This represents the set of values ​​for the k-th environmental parameter. i is One of the elements, Let represent the set of values ​​for the j-th lens feature. for One of the elements, Indicates environmental parameters Values And lens features Values The probability, Indicates environmental parameters Values The probability, The lens feature value is indicated. The probability of.

[0115] It should be noted that environmental quality mutual information is used to measure the statistical dependence between production environment parameters (such as temperature, humidity, and cleanliness level) and drone lens quality characteristics (such as texture detail retention and color shift), including linear and nonlinear correlations.

[0116] The environmental quality importance index is calculated by combining the environmental quality correlation coefficient and the environmental quality mutual information.

[0117]

[0118] in, This represents the environmental quality importance index of the k-th environmental parameter and the j-th lens feature. This represents the environmental quality mutual information between the k-th environmental parameter and the j-th lens feature. This represents the environmental quality correlation coefficient between the k-th environmental parameter and the j-th lens feature, and max() represents the maximum value function. express This represents the maximum value of the environmental quality mutual information between the k-th environmental parameter and the j-th lens feature. , For adjustment coefficients, Used to adjust the influence of environmental quality mutual information, ranging from [0,1]. The environmental quality importance index is used to prevent excessive amplification when there is high correlation and high mutual information, and its range is [0, 0.5].

[0119] It should be noted that the environmental quality importance index is calculated by combining the environmental quality correlation coefficient and the environmental quality mutual information. The Pearson correlation coefficient excels at quantifying the linear relationship between environmental parameters and quality characteristics (e.g., a linear increase in distortion due to rising temperature), but may overlook complex nonlinear relationships (e.g., the threshold effect of the combined effect of humidity and temperature on transmittance). Mutual information, on the other hand, can comprehensively capture any form of statistical dependence, including nonlinear and non-monotonic relationships. The combination of the two ensures that the analysis covers both intuitive linear trends and does not overlook hidden nonlinear effects. For example, if an environmental parameter has a low Pearson coefficient (weak linear correlation) but a high mutual information (nonlinear dependence), the importance index can still identify its potential impact and avoid misjudgment.

[0120] Step S4: Filter the drone lens production environment data using the environmental quality importance index to obtain drone lens production environment screening data. Construct a multivariate regression tree model using the drone lens production environment screening data and drone lens feature data.

[0121] The production environment data for the drone lenses is filtered using the environmental quality importance index, and each environmental parameter is calculated first. The importance of overall lens quality for drones:

[0122]

[0123] in, Represents the k-th environmental parameter The importance of the overall lens quality of drones This represents the environmental quality importance index of the k-th environmental parameter and the j-th lens feature. Indicating the first element in the drone lens feature data The index of each feature, where m represents the number of features in the UAV lens feature data.

[0124] Environmental parameters with importance scores higher than a preset threshold in the drone lens production environment data are retained to obtain drone lens production environment screening data.

[0125] Screening data on the production environment of drone lenses With drone lens feature data (denoted as The training set is constructed by aligning the lens's unique serial number and timestamp. Among them, data on the screening of drone lens production environments. UAV lens feature data (as independent variable) The dependent variable is . The goal of a multivariate regression tree model is to predict multidimensional quality characteristics through environmental parameters and establish a nonlinear mapping relationship between environment and quality.

[0126] Using the drone lens production environment screening data and drone lens feature data, a multivariate regression tree model is constructed. For the sample set within the leaf node t of the multivariate regression tree model... The weighted mean square error is calculated as follows:

[0127]

[0128] in: Indicates the current node The weighted mean square error, For the current node Number of samples within, Let be the average value of the j-th lens feature within node t. Let represent the i-th sample value of the j-th lens feature, where i represents the index of the i-th sample value.

[0129] In the recursive segmentation and splitting criteria, the first step is to iterate through all environmental parameters and their possible candidate splitting points, and then select the splitting method that minimizes prediction error through system evaluation. The specific process is as follows: For continuous environmental parameters, candidate splitting points are generated based on quantiles, for example, dividing the parameter values ​​into multiple intervals; for discrete parameters (such as different categories of cleanliness levels), all possible values ​​are enumerated as splitting points. This step determines the possible segmentation methods for each parameter. Subsequently, for each candidate splitting point... The sample data of the current node Divide the node into left and right child nodes, and calculate the weighted mean square error after splitting:

[0130]

[0131] in, This represents the change in weighted mean square error before and after the split. Indicates the current leaf node The weighted mean square error, For the current node Number of samples within, Indicates the current node The number of samples in the left child node after the split. For the current node The number of samples in the right child node after the split. This represents the weighted mean square error of the left child node. This represents the weighted mean square error of the right child node.

[0132] The improvement in model prediction accuracy by splitting is quantified by comparing the weighted mean squared error before and after the split. Among all candidate split points, the split point that maximizes the weighted mean squared error is selected as the splitting rule for the current node. Finally, recursive termination conditions are set to control model complexity: when the number of samples for a node falls below a preset minimum sample threshold (e.g., less than 50 samples), or the tree depth reaches a preset maximum level, further splitting stops, and the node is marked as a leaf node. These conditions effectively prevent overfitting and avoid reducing generalization ability due to excessive tree structure complexity, ultimately forming a multivariate regression tree.

[0133] Step S5: Perform hierarchical analysis on the drone lens feature data using the analytic hierarchy process (AHP) to obtain the feature weights of the drone lens feature data. By combining the drone lens feature data and feature weights, calculate the lens evaluation index of the drone lens.

[0134] Using the lens evaluation index as the target layer and each feature in the UAV lens feature data as the criterion layer, the UAV lens feature data and feature weights are used to calculate the lens evaluation index.

[0135] Construct the judgment matrix A for the target layer and the criterion layer. The judgment matrix A is an n*n matrix and satisfies (i,j=1,2,...,n), >0, when j2=j, that is, to determine the right diagonal of matrix A. =1, The ratio of the relative importance of the j2th feature to the jth feature in the UAV lens feature data is qualitatively determined by the decision-maker based on Table 1.

[0136]

[0137] Table 1

[0138] It should be noted that the features are indicators in the drone lens feature data, such as texture detail retention, color shift, edge distortion, resolution retention, and grayscale distribution uniformity.

[0139] It should be noted that the ratio of the relative importance of the i-th feature to the j-th feature means, under a specific decision-making objective, judging the difference in importance between the two features based on expert experience or actual scenario requirements. For example, when determining the lens evaluation index, texture detail retention is considered more important than color shift; in this case, the ratio is 3, meaning that texture detail retention is three times more important than color shift.

[0140] Calculate the largest eigenvalue of judgment matrix A :

[0141]

[0142] in, Let A represent the largest eigenvalue of the judgment matrix A, j be the index of the j-th feature in the UAV lens feature data, m represent the number of features in the UAV lens feature data, and A be the judgment matrix of the UAV lens feature data. This represents the feature weight of the j-th feature in the drone lens feature data.

[0143] And calculate the consistency index : Calculate the consistency ratio : ,in, For consistency ratio, As a consistency indicator, CR is a random consistency index. When CR < 0.1, the consistency of the judgment matrix A of the UAV lens feature data is considered acceptable.

[0144] By calculating and normalizing the eigenvectors of the judgment matrix A, the feature weights in the UAV lens feature data are obtained:

[0145]

[0146] in, This represents the feature weight of the j-th feature in the UAV lens feature data, and m represents the number of features in the UAV lens feature data. Indicating the first element in the drone lens feature data Index of features, This represents the ratio of the relative importance of the j2nd feature to the jth feature in the UAV lens feature data, where j2 is the index of the j2nd feature in the UAV lens feature data.

[0147] By combining the drone lens feature data and feature weights, the lens evaluation index for each drone lens is calculated:

[0148]

[0149] in, Let m be the lens evaluation index, and i be the index of the j-th feature in the UAV lens feature data. This represents the feature weight of the j-th feature in the drone lens feature data. This represents the j-th feature in the drone lens feature data.

[0150] Step S6: Based on the lens evaluation index, the characteristic data of the UAV lens are classified using the kernel density estimation method to obtain UAV lens labels; when the UAV lens label requires re-inspection, the UAV lens production environment screening data corresponding to the UAV lens label is input into the multivariate regression tree model to obtain the prediction error result; by combining the error prediction result and the environmental quality importance index, the environmental quality correlation index is calculated; if the environmental quality correlation index is greater than a preset threshold, a source tracing and early warning is issued to realize the monitoring of UAV camera lenses.

[0151] Based on the lens evaluation index, after standardizing the lens evaluation index, the drone lens feature data is classified by kernel density estimation to obtain drone lens labels.

[0152] The probability density distribution of the lens evaluation index is calculated using the Gaussian kernel function:

[0153]

[0154] in, This indicates the Gaussian kernel function in the target lens evaluation index value. Kernel density estimation at [location] The larger the value, the higher the target lens evaluation index. The more lenses, the better. Here, is the Gaussian kernel function, and h is the bandwidth of the Gaussian kernel function, used to control the smoothing degree. It is determined by the number of samples in the UAV lens feature data; the larger the number of samples, the smaller the bandwidth. n is the number of samples in the UAV lens feature data, and i represents the index of the i-th sample in the UAV lens feature data. The target lens evaluation index value, Let be the evaluation index of the i-th lens.

[0155] By calculating the probability density distribution of the lens evaluation index, a probability density curve is plotted with the target lens evaluation index as the horizontal axis and the kernel density estimate of the target evaluation index as the vertical axis. The drone lens feature data corresponding to each lens evaluation index is classified, and the drone lens labels are obtained as follows: excellent, usable, and require re-inspection.

[0156] It should be noted that when classifying the drone lens feature data, if the probability density curve distribution is multi-peaked (such as bi-peak or tri-peak), the valley bottom is used as the threshold. For example, in the case of a bi-peaked distribution, the valley bottom between the two peaks is set as the threshold, and the data is divided into three categories: low-peak region requiring re-inspection, medium-peak region usable, and high-quality region.

[0157] When the drone lens label requires re-inspection, the drone lens production environment screening data corresponding to the drone lens label is input into the multivariate regression tree model to obtain the prediction error result:

[0158]

[0159] in, This represents the j-th prediction error result for the lens that needs to be re-examined. Indicates the first The weights of the quality features represent the features of the j-th lens. This represents the j-th feature predicted by the multivariate regression tree model.

[0160] By combining the error prediction results and the environmental quality importance index, an environmental quality correlation index is calculated. This correlation index is then used to monitor and analyze the lenses of drone cameras.

[0161]

[0162] in, This represents the environmental quality correlation index between the k-th environmental parameter and the j-th lens feature, and the prediction error result. This represents the environmental quality importance index of the k-th environmental parameter and the j-th lens feature. This represents the j-th prediction error result for the lens that needs to be re-examined.

[0163] It should be noted that the environmental quality correlation index is calculated by combining the error prediction results with the environmental quality importance index. The error prediction results reflect the deviation between the actual quality characteristics of the lenses and the model predictions, directly characterizing any abnormal fluctuations that may exist during the production process. For example, if the edge distortion (ED) of a lens is significantly higher than the model prediction, it indicates that there may be uncaptured interference factors in actual production. The environmental quality importance index quantifies the influence of different environmental parameters (such as temperature and humidity) on quality characteristics (such as ED and CD) through Pearson correlation coefficient and mutual information, covering both linear correlations and capturing nonlinear dependencies. By linking quality deviations (errors) with potential environmental influencing factors (importance), it is clear which fluctuations in environmental parameters are the key causes of current quality problems. For example, if the color shift (CD) prediction error of a batch of lenses is high and the importance index of the cleanliness level is significant, then the cleanliness control of the coating workshop should be prioritized. Prioritization: The correlation index provides a basis for decision-making regarding production parameter adjustments. Environmental parameters with high correlation indices require immediate intervention, while parameters with low correlation indices can be temporarily ignored, thereby optimizing resource allocation.

[0164] The environmental quality correlation index is used to monitor and analyze the lenses of drone cameras. If the environmental quality correlation index is greater than a preset threshold, it indicates that the k-th environmental parameter has the greatest impact on the j-th lens feature, leading to a large difference between the actual result and the prediction error, requiring adjustment of the k-th environmental parameter. For example, when k represents cleanliness and j represents grayscale distribution uniformity... When the value is greater than the preset threshold (=0.9), it indicates that the grayscale distribution of the lens is abnormal. The associated index is displayed as cleanliness. Then, check the cleanliness sensor data to find the dust particle concentration and trace whether the filter in the coating workshop has been replaced in time.

[0165] This paper proposes an online monitoring, analysis, and management method for the production of drone camera lenses. By collecting and analyzing characteristic data of drone lenses and environmental data, and combining various mathematical models and algorithms, real-time monitoring and accurate evaluation of drone lens quality can be achieved. The innovation of this paper lies in the introduction of multiple indicators, such as lens evaluation index, environmental quality importance index, and environmental quality correlation index, to systematically analyze quality issues in the lens production process. Furthermore, a data-driven optimization path is proposed, providing a new quality monitoring and optimization method for the drone lens manufacturing industry.

[0166] By combining the feature data and feature weights of the drone lenses, a lens evaluation index is calculated to provide a quantitative quality assessment for each lens. This index comprehensively considers various physical properties and optical performance of the lens, enabling the production line to understand the overall quality level of the lenses in real time. Using this evaluation index, different lenses can be accurately classified, providing a scientific basis for subsequent quality control and re-inspection, thereby improving the product qualification rate in the production process and reducing rework and losses caused by quality defects.

[0167] By combining environmental quality correlation coefficients and environmental quality mutual information, the calculated environmental quality importance index can comprehensively assess the impact of the production environment on lens quality. Traditional quality monitoring often focuses on the lens characteristics themselves, neglecting the potential impact of changes in the production environment on the final quality. By introducing the environmental quality importance index, it is possible to accurately identify which environmental factors have the greatest impact on lens quality during the production process, thereby helping production managers to adjust production environment parameters in a timely manner, optimize the production process, and improve the consistency and stability of lens quality.

[0168] By combining error prediction results with an environmental quality importance index, the calculated environmental quality correlation index further reveals the complex relationship between environmental parameters and lens characteristics. This index quantifies the impact of environmental parameter changes on lens quality prediction errors and provides actionable optimization guidance for the production process. When the environmental quality correlation index exceeds a preset threshold, it indicates that certain environmental factors have a significant impact on lens quality, and these factors should be prioritized and adjusted during production. This method can accurately predict quality fluctuations, identify potential quality problems in advance, effectively reduce production risks, and improve the production efficiency and product quality of UAV lenses.

[0169] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, the phrase "comprising an element defined as..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0170] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for online monitoring, analysis, and management of drone camera lens manufacturing, characterized in that: The process includes the following steps: Step S1: Collect lens data from the drone camera to obtain drone lens image data; Step S2: Extract features from the drone lens image data to obtain drone lens feature data; Step S3: Collect drone lens production environment data to obtain drone lens production environment data; calculate the Pearson correlation between the drone lens production environment data and the drone lens feature data to obtain the environmental quality correlation coefficient; calculate the mutual information of the features between the drone lens production environment data and the drone lens feature data to obtain the environmental quality mutual information; and calculate the environmental quality importance index by combining the environmental quality correlation coefficient and the environmental quality mutual information, specifically including the following steps: Calculate the environmental quality importance index by combining the environmental quality correlation coefficient and the environmental quality mutual information. ;in, This represents the environmental quality importance index of the k-th environmental parameter and the j-th lens feature. This represents the environmental quality mutual information between the k-th environmental parameter and the j-th lens feature. This represents the environmental quality correlation coefficient between the k-th environmental parameter and the j-th lens feature, and max() represents the maximum value function. This represents the maximum value of the environmental quality mutual information between the k-th environmental parameter and the j-th lens feature. ， For adjustment coefficients, Used to adjust the influence of environmental quality mutual information, ranging from [0,1]. To prevent excessive amplification of the environmental quality importance index when there is high correlation and high mutual information, the range is [0, 0.5]. Step S4: The production environment data of the UAV lens is filtered using the environmental quality importance index to obtain the UAV lens production environment screening data. A multivariate regression tree model is constructed using the UAV lens production environment screening data and the UAV lens feature data. Step S5: The UAV lens feature data is subjected to hierarchical analysis using the analytic hierarchy process (AHP) to obtain the feature weights of the UAV lens feature data. By combining the UAV lens feature data and feature weights, the lens evaluation index of the UAV lens is calculated. Step S6: Based on the lens evaluation index, the UAV lens feature data is classified using the kernel density estimation method to obtain... Drone lens label; when the drone lens label requires re-inspection, the drone lens production environment screening data corresponding to the drone lens label is input into a multivariate regression tree model to obtain error prediction results; by combining the error prediction results and the environmental quality importance index, an environmental quality correlation index is calculated; if the environmental quality correlation index is greater than a preset threshold, a source tracing and early warning is issued to monitor the drone camera lens; the calculation of the environmental quality correlation index by combining the error prediction results and the environmental quality importance index includes the following specific steps: calculating the environmental quality correlation index by combining the error prediction results and the environmental quality importance index, and using the environmental quality correlation index to achieve monitoring and analysis of the drone camera lens. ;in, This represents the environmental quality correlation index between the k-th environmental parameter and the j-th lens feature. This represents the environmental quality importance index of the k-th environmental parameter and the j-th lens feature. This is the j-th error prediction result for the lens that needs to be re-examined.

2. The method for online monitoring, analysis and management of UAV camera lens manufacturing according to claim 1, characterized in that: The process of extracting features from drone lens image data to obtain drone lens feature data includes the following specific steps: The texture detail retention of the drone lens image data is quantified by calculating the average gradient magnitude of the edge regions to quantify the film's ability to retain details; the larger the average gradient magnitude, the clearer the texture details. Among them, TDR stands for Texture Detail Retention. Represents the pixels in the drone lens image data grayscale value, Indicates drone lens image data in The gradient vector at that point, This represents the edge region defined in the calibration diagram. Represents the edge region defined in the calibration diagram. The total number of pixels within the image; the color shift in the drone lens image data, which measures the lens's color reproduction capability; the smaller the value, the smaller the color shift. Where CD represents the color offset. This represents the measured color value of the l-th color wheel region in the UAV lens image data. This represents the theoretical color value of the l-th color wheel in the calibration diagram. This represents the CIEDE2000 color difference formula, used to calculate the difference between measured and theoretical colors. This indicates the total number of color wheels, which defaults to 24. The edge distortion degree of the drone lens image data is quantified by the mean of the Euclidean distance of the coordinate offset. The larger the optical distortion value, the more severe the distortion. ;in, For edge distortion degree, This indicates the distortion grid in the drone lens image data. The actual coordinates of the intersection points This represents the theoretical coordinates of the corresponding intersection points in the calibration diagram. This represents the total number of grid intersections; the resolution retention rate of the UAV lens image data. The closer the resolution retention rate ratio is to 1, the closer the lens resolution is to the ideal value; if it is less than 1, the resolution decreases. Where RR represents the resolution retention rate of the UAV lens image data. This represents the minimum resolvable linewidth in the drone lens image data. This represents the minimum theoretical linewidth to be resolved in the calibration map design; the grayscale distribution uniformity value of the UAV lens image data, the closer the grayscale distribution uniformity value is to 1, the more uniform the grayscale distribution; if it is close to 0, there is obvious brightness unevenness. Where GU is the gray-level distribution equilibrium value, This represents the average grayscale value of pixels in the grayscale region of the drone lens image data. This represents the standard deviation of pixel grayscale values ​​in the grayscale region of the drone lens image data. The constant term is in the range (0, 1e-3); finally, the UAV lens feature data is obtained, where each UAV lens feature data is H, and H = [ , , , , ]。 3. The method for online monitoring, analysis and management of UAV camera lens manufacturing according to claim 2, characterized in that: The environmental quality correlation coefficient is obtained by calculating the Pearson correlation between drone lens production environment data and drone lens characteristic data. The process includes the following steps: Calculating the Pearson correlation between environmental parameters in the drone lens production environment data and lens characteristics in the drone lens feature data to obtain the environmental quality correlation coefficient. ;in, This represents the environmental quality correlation coefficient between the k-th environmental parameter and the j-th lens feature. For the i-th sample value of the k-th environmental parameter, This represents the average value of the k-th environmental parameter. This represents the i-th sample value of the j-th lens feature. Let represent the average value of the j-th lens feature, n represent the total number of samples, and i represent the index of the i-th sample value.

4. The method for online monitoring, analysis and management of UAV camera lens manufacturing according to claim 3, characterized in that: The method of obtaining environmental quality mutual information by calculating the mutual information of features between UAV lens production environment data and UAV lens feature data includes the following steps: Obtaining environmental quality mutual information by calculating the mutual information of features between UAV lens production environment data and UAV lens feature data: ;in, This represents the environmental quality mutual information between the k-th environmental parameter and the j-th lens feature. This represents the set of values ​​for the k-th environmental parameter. i is One of the elements, Let represent the set of values ​​for the j-th lens feature. for One of the elements, Indicates environmental parameters Values And lens features Values The probability, Indicates environmental parameters Values The probability, The lens feature value is indicated. The probability of.

5. The method for online monitoring, analysis and management of UAV camera lens manufacturing according to claim 4, characterized in that: The process of filtering the drone lens production environment data using the environmental quality importance index to obtain drone lens production environment screening data includes the following specific steps: First, the drone lens production environment data is filtered using the environmental quality importance index, and then each environmental parameter is calculated. The importance of overall lens quality for drones: ;in, Represents the k-th environmental parameter The importance of the overall lens quality of drones This represents the environmental quality importance index of the k-th environmental parameter and the j-th lens feature. Indicating the first element in the drone lens feature data The index of each feature is given, where m represents the number of features in the drone lens feature data. Environmental parameters with importance scores higher than a preset threshold are retained in the drone lens production environment data to obtain drone lens production environment screening data.

6. The method for online monitoring, analysis and management of UAV camera lens manufacturing according to claim 5, characterized in that: The step of constructing a multivariate regression tree model using the drone lens production environment screening data and drone lens feature data includes the following specific steps: Constructing a multivariate regression tree model using the drone lens production environment screening data and drone lens feature data; for the sample set within the leaf node t in the multivariate regression tree model... The weighted mean square error is calculated as follows: ;in: Indicates the current node The weighted mean square error, For the current node Number of samples within, Let be the average value of the j-th lens feature within node t. This represents the i-th sample value of the j-th lens feature, where i represents the index of the i-th sample value. In the recursive segmentation and splitting criterion, all environmental parameters are traversed as candidate splitting points, and based on each candidate splitting point... The sample data of the current node Divide the node into left and right child nodes, and calculate the weighted mean square error after splitting: ;in, This represents the change in weighted mean square error before and after the split. Indicates the current leaf node The weighted mean square error, For the current node Number of samples within, Indicates the current node The number of samples in the left child node after the split. For the current node The number of samples in the right child node after the split. This represents the weighted mean square error of the left child node. This represents the weighted mean square error of the right child node. Among all candidate split points, the split point that maximizes the weighted mean square error is selected as the split rule for the current node. When the number of samples in a node is lower than the preset minimum sample threshold, the splitting stops and the node is marked as a leaf node, thus forming a multivariate regression tree model.

7. The method for online monitoring, analysis and management of UAV camera lens manufacturing according to claim 6, characterized in that: The step of performing hierarchical analysis on the drone lens feature data using the analytic hierarchy process (AHP) to obtain the feature weights of the drone lens feature data, and then calculating the lens evaluation index of the drone lens by combining the drone lens feature data and feature weights, includes the following specific steps: performing hierarchical analysis on the drone lens feature data using the AHP to obtain the feature weights of the drone lens feature data: ;in, This represents the feature weight of the j-th feature in the UAV lens feature data, and m represents the number of features in the UAV lens feature data. Indicating the first element in the drone lens feature data Index of features, This represents the ratio of the relative importance of the j2nd feature to the jth feature in the drone lens feature data, where j2 is the index of the j2nd feature in the drone lens feature data; by combining the drone lens feature data and feature weights, the lens evaluation index for each drone lens is calculated: ;in, Let m be the lens evaluation index, m be the number of features in the UAV lens feature data, and j be the index of the j-th feature in the UAV lens feature data. This represents the feature weight of the j-th feature in the drone lens feature data. This represents the j-th feature in the drone lens feature data.

8. The method for online monitoring, analysis and management of UAV camera lens manufacturing according to claim 7, characterized in that: The process of classifying the drone lens feature data based on the lens evaluation index using kernel density estimation to obtain drone lens labels includes the following specific steps: Calculating the probability density distribution of the lens evaluation index using a Gaussian kernel function: ;in, This indicates the Gaussian kernel function in the target lens evaluation index value. Kernel density estimation at [location] The larger the value, the higher the target lens evaluation index. The more lenses, Here, is the Gaussian kernel function, and h is the bandwidth of the Gaussian kernel function, used to control the smoothing degree. It is determined by the number of samples in the UAV lens feature data; the larger the number of samples, the smaller the bandwidth. n is the total number of samples, and i represents the index of the i-th sample in the UAV lens feature data. The target lens evaluation index value, Let be the evaluation index of the i-th lens. By calculating the probability density distribution of the lens evaluation index, a probability density curve is plotted with the target lens evaluation index as the horizontal axis and the kernel density estimate of the target evaluation index as the vertical axis. The drone lens feature data corresponding to each lens evaluation index is classified to obtain drone lens labels including: high quality, usable, and requiring re-inspection.

Images