Glass fiber reinforced plastic mold processing method and system based on feedback analysis

Abstract

The invention discloses a glass fiber reinforced plastic mold processing method and system based on feedback analysis. The method comprises the following steps: acquiring video data of N processing steps, extracting key frames to form a data set, extracting contour, texture and color features by applying a Prewitt operator and a color histogram, and constructing a first feature set; learning and reconstructing data features by adopting an auto-encoding network, generating a second feature set, analyzing feature changes in combination with a ResNet model, forming an intermediate change feature set, constructing a full-machining process feature table, comparing a real-time target feature set with the feature table, identifying machining defects and positioning problems, and generating an optimization scheme. Accurate optimization and real-time feedback analysis of the machining steps are achieved, the machining quality and the machining efficiency are improved, and the situation that machining evaluation is conducted depending on manual experience is reduced.

Description

Technical Field

[0001] The present invention relates to the field of mold processing analysis, and more specifically, to a glass fiber reinforced plastic mold processing method and system based on feedback analysis. Background Art

[0002] FRP, or fiber reinforced plastic, is a reinforced plastic with glass fiber or its products as reinforcement material. It has the advantages of high strength, light weight, corrosion resistance, good electrical insulation and thermal properties. It can be used as a substitute for steel in many occasions and is an indispensable material in national economic construction. In the process of FRP mold processing, mold making, FRP production and product quality control are crucial. However, in the existing FRP mold processing and production, mold processing quality analysis is often based on manual work, relying on manual experience, lacking processing analysis based on machine vision, and lacking effective feedback analysis and processing optimization for each processing step, making it difficult to improve processing efficiency and quality. Summary of the invention

[0003] The invention overcomes the defects of the prior art and proposes a glass fiber reinforced plastic mold processing method and system based on feedback analysis.

[0004] The first aspect of the present invention provides a glass fiber reinforced plastic mold processing method based on feedback analysis, comprising: Based on machine vision technology, obtain video data of the preset FRP mold in N processing steps; Extract key frames from the video data to form a processed image data set, identify the FRP mold area on the processed image data set, extract contour, texture and color features through the Prewitt operator and color histogram to form the first feature set; Based on the autoencoder network, data feature learning and data reconstruction are performed on the first feature set to generate a second feature set, the second feature sets of adjacent processing steps are imported into the ResNet model to generate feature graphs, and feature change analysis is performed based on the feature graphs to generate intermediate change feature sets; A full processing flow feature table is constructed based on the first feature set, the second feature set and the corresponding intermediate change feature set, real-time video data of the target FRP mold in N processing steps are collected and key frame extraction and feature extraction are performed to form a target feature set, and feature comparison and processing defect identification are performed between the target feature set and the full processing flow feature table, the existing defects are located in steps, and a processing step optimization plan is generated; According to the processing step optimization plan, the second target FRP mold is processed and produced, and the processing quality is evaluated through the real-time collected second target feature set and the full processing flow feature table. Based on the evaluation, real-time feedback analysis of the processing step optimization plan is carried out.

[0005] In this solution, the video data of the preset FRP mold in N processing steps is obtained based on machine vision technology, specifically: Based on machine vision technology, through a preset high-definition camera device, image data of the entire processing process of the preset fiberglass mold is acquired to obtain corresponding video data; The whole processing includes N processing steps.

[0006] In this solution, the key frames of the video data are extracted to form a processed image data set, the fiberglass mold area is identified on the processed image data set, and the contour, texture and color features are extracted by the Prewitt operator and the color histogram to form a first feature set, which is specifically: Extract key frames from video data to form processed image data; Perform image noise reduction, enhancement, and data standardization preprocessing on the processed image data; The FRP mold area is identified on the processed image data set to obtain the mold area image. The edge of the mold area image is detected by the Prewitt operator, and the contour features and texture features are extracted. Generate a color histogram based on the mold area image, and generate color feature data through statistical information of the color histogram; A first feature set is formed based on the contour feature, texture feature and color feature data.

[0007] In this solution, each of the N processing steps corresponds to a processed image data set.

[0008] In this solution, based on the autoencoder network, data feature learning and data reconstruction are performed on the first feature set to generate a second feature set, the second feature sets of adjacent processing steps are imported into the ResNet model to generate feature maps, and feature change analysis is performed based on the feature maps to generate intermediate change feature sets, specifically: Construct an autoencoder network, import the first feature set as raw data into the autoencoder network, map the imported raw data to a low-dimensional hidden representation through the encoder, and perform the mapping process through a multi-layer perceptron to obtain a low-dimensional output; The low-dimensional output is reconstructed through the decoder to generate a high-dimensional output. A preset loss function is introduced to compare the high-dimensional output with the original data input, and the autoencoder network parameters are corrected. The data reconstruction is repeated until the loss function converges to the preset value. The high-dimensional output of the autoencoder network at this time is recorded to obtain the second feature set.

[0009] In this solution, the second feature set of the adjacent processing steps is imported into the ResNet model to generate a feature map, and feature change analysis is performed based on the feature map to generate an intermediate change feature set, specifically: The second feature sets of two adjacent processing steps are respectively imported into the ResNet model for feature learning and forward propagation of the input to the output layer, and feature maps of different levels are generated, and a plurality of first feature maps and a plurality of second feature maps are respectively marked based on the two adjacent processing steps; Based on the same level, the first feature map and the second feature map are respectively added and multiplied, the feature maps are fused, and a transition feature map is generated; Through multi-level feature map operations, multiple transition feature maps are generated. Based on the ResNet model, corresponding feature data is generated for all transition feature maps to form an intermediate change feature set. The feature variation analysis is performed between every two adjacent processing steps to generate multiple variation feature sets.

[0010] In this solution, the full processing flow feature table is constructed according to the first feature set, the second feature set and the corresponding intermediate change feature set, the real-time video data of the target FRP mold in N processing steps are collected and key frame extraction and feature extraction are performed to form a target feature set, the target feature set is compared with the full processing flow feature table for feature comparison and processing defect identification, the existing defects are located in the steps, and the processing step optimization plan is generated, specifically: Based on the processing sequence, the first feature set, the second feature set and the corresponding intermediate change feature set are sorted, and the processing node information of each feature set is associated to form a feature table of the entire processing process; Collect real-time video data of the target FRP mold in N processing steps and perform key frame extraction and multi-dimensional feature extraction, integrate the extracted features to form a target feature set; The target feature set is compared with the feature set in the feature table of the entire processing flow to identify processing defects. The comparison process analyzes the difference of features based on the standard Euclidean distance, and the feature set in the feature table of the entire processing flow is used as the standard feature to identify processing defects and locate the processing steps where the defects are located, generating processing defect information; The comparison process also includes a difference analysis between the target feature set and the intermediate change feature set in the full processing flow feature table; According to the processing defect information, the corresponding processing steps are optimized and the processing step optimization plan is generated.

[0011] In this scheme, according to the processing step optimization scheme, the second target FRP mold is processed and produced, and the processing quality is evaluated through the second target feature set and the full processing flow feature table collected in real time, and the real-time feedback analysis of the processing step optimization scheme is performed based on the evaluation, specifically: According to the processing step optimization plan, the processing steps of the second target FRP mold are optimized; The video data of the second target FRP mold is collected in real time and key frame extraction and image preprocessing are performed to generate a second target feature set; The second target feature set is compared with the feature table of the entire processing flow in real time for processing feature comparison and processing quality evaluation. The evaluation results are used to provide real-time feedback on the processing step optimization plan and real-time adjustment of the processing plan.

[0012] The second aspect of the present invention further provides a glass fiber reinforced plastic mold processing system based on feedback analysis, the system comprising: a memory, a processor, the memory comprising a glass fiber reinforced plastic mold processing program based on feedback analysis, the glass fiber reinforced plastic mold processing program based on feedback analysis being executed by the processor to implement the following steps: Based on machine vision technology, obtain video data of the preset FRP mold in N processing steps; Extract key frames from the video data to form a processed image data set, identify the FRP mold area on the processed image data set, extract contour, texture and color features through the Prewitt operator and color histogram to form the first feature set; Based on the autoencoder network, data feature learning and data reconstruction are performed on the first feature set to generate a second feature set, the second feature sets of adjacent processing steps are imported into the ResNet model to generate feature graphs, and feature change analysis is performed based on the feature graphs to generate intermediate change feature sets; A full processing flow feature table is constructed based on the first feature set, the second feature set and the corresponding intermediate change feature set, real-time video data of the target FRP mold in N processing steps are collected and key frame extraction and feature extraction are performed to form a target feature set, and feature comparison and processing defect identification are performed between the target feature set and the full processing flow feature table, the existing defects are located in steps, and a processing step optimization plan is generated; According to the processing step optimization plan, the second target FRP mold is processed and produced, and the processing quality is evaluated through the real-time collected second target feature set and the full processing flow feature table. Based on the evaluation, real-time feedback analysis of the processing step optimization plan is carried out.

[0013] The third aspect of the present invention also provides a computer-readable storage medium, which includes a FRP mold processing program based on feedback analysis. When the FRP mold processing program based on feedback analysis is executed by a processor, the steps of the FRP mold processing method based on feedback analysis as described in any one of the above items are implemented.

[0014] The present invention discloses a glass fiber reinforced plastic mold processing method and system based on feedback analysis. By acquiring video data of N processing steps, extracting key frames to form a data set, using Prewitt operator and color histogram to extract contour, texture and color features, and constructing a first feature set. Using an autoencoder network to learn and reconstruct data features, a second feature set is generated, and the feature changes are analyzed in combination with a ResNet model to form an intermediate change feature set, and a full processing flow feature table is constructed. The real-time target feature set and the feature table are compared to identify processing defects and locate problem steps, generate optimization solutions, realize accurate optimization of processing steps and real-time feedback analysis, improve processing quality and processing efficiency, and reduce the situation of relying on manual experience for processing evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A flow chart of a glass fiber reinforced plastic mold processing method based on feedback analysis of the present invention is shown; Figure 2 A block diagram of a glass fiber reinforced plastic mold processing system based on feedback analysis of the present invention is shown. DETAILED DESCRIPTION

[0016] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0017] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0018] Figure 1 A flow chart of a glass fiber reinforced plastic mold processing method based on feedback analysis of the present invention is shown.

[0019] like Figure 1 As shown, the first aspect of the present invention provides a glass fiber reinforced plastic mold processing method based on feedback analysis, comprising: S102, based on machine vision technology, obtaining video data of a preset fiberglass mold in N processing steps; S104, extracting key frames from the video data to form a processed image data set, identifying the fiberglass mold area on the processed image data set, extracting contour, texture and color features using a Prewitt operator and a color histogram to form a first feature set; S106, based on the autoencoder network, performing data feature learning and data reconstruction on the first feature set to generate a second feature set, importing the second feature sets of adjacent processing steps into the ResNet model to generate feature graphs, and performing feature change analysis based on the feature graphs to generate an intermediate change feature set; S108, constructing a full processing flow feature table according to the first feature set, the second feature set and the corresponding intermediate change feature set, collecting real-time video data of the target FRP mold in N processing steps and performing key frame extraction and feature extraction to form a target feature set, performing feature comparison and processing defect identification on the target feature set and the full processing flow feature table, locating the steps of the existing defects, and generating a processing step optimization plan; S110, according to the processing step optimization plan, the second target FRP mold is processed and produced, and the processing quality is evaluated through the second target feature set and the full processing flow feature table collected in real time, and real-time feedback analysis of the processing step optimization plan is performed based on the evaluation.

[0020] It should be noted that the production of the FRP mold generally includes multiple steps, such as raw material placement, spraying gel coat, spraying glass fiber multiple layers, waiting for molding, curing, applying resin, polishing, etc. Each step corresponds to different mold processing image features, and the features of adjacent processing steps in the county often have a certain correlation. For example, the color features and contour features of the processing steps before and after curing have certain changes and correlations, and the layering processing steps of different glass fibers. In the present invention, by simulating and mining the feature changes of adjacent processing steps and generating corresponding transition features, efficient and rapid processing feature defect identification and step positioning can be achieved in the future, and corresponding step optimization can be performed, such as re-evaluation and replacement of raw materials, regulation of mixture ratio, and regulation of resin usage.

[0021] According to an embodiment of the present invention, the video data of the preset FRP mold in N processing steps is obtained based on machine vision technology, specifically: Based on machine vision technology, through a preset high-definition camera device, image data of the entire processing process of the preset fiberglass mold is acquired to obtain corresponding video data; The whole processing includes N processing steps.

[0022] It should be noted that the preset FRP mold is a selected standardized production mold, which can be screened based on mold test evaluation and has a good comparison effect. It is used to generate the corresponding full processing flow feature table.

[0023] According to an embodiment of the present invention, the key frames of the video data are extracted to form a processed image data set, the fiberglass mold area is identified on the processed image data set, and the contour, texture and color features are extracted by the Prewitt operator and the color histogram to form a first feature set, specifically: Extract key frames from video data to form processed image data; Perform image noise reduction, enhancement, and data standardization preprocessing on the processed image data; The FRP mold area is identified on the processed image data set to obtain the mold area image. The edge of the mold area image is detected by the Prewitt operator, and the contour features and texture features are extracted. Generate a color histogram based on the mold area image, and generate color feature data through statistical information of the color histogram; A first feature set is formed based on the contour feature, texture feature and color feature data.

[0024] It should be noted that the first feature set is a multi-dimensional image feature that describes the mold during the processing, which is used for subsequent processing quality and defect analysis of different processing steps.

[0025] According to an embodiment of the present invention, in the N processing steps, each processing step corresponds to a processed image data set.

[0026] According to an embodiment of the present invention, based on the autoencoder network, data feature learning and data reconstruction are performed on the first feature set to generate a second feature set, the second feature sets of adjacent processing steps are imported into the ResNet model to generate feature graphs, and feature change analysis is performed based on the feature graphs to generate intermediate change feature sets, specifically: Construct an autoencoder network, import the first feature set as raw data into the autoencoder network, map the imported raw data to a low-dimensional hidden representation through the encoder, and perform the mapping process through a multi-layer perceptron to obtain a low-dimensional output; The low-dimensional output is reconstructed through the decoder to generate a high-dimensional output. A preset loss function is introduced to compare the high-dimensional output with the original data input, and the autoencoder network parameters are corrected. The data reconstruction is repeated until the loss function converges to the preset value. The high-dimensional output of the autoencoder network at this time is recorded to obtain the second feature set.

[0027] It should be noted that the autoencoder network includes an encoder, a decoder and a loss function. The preset loss function can be calculated using mean square error.

[0028] Here, the present invention introduces an autoencoder network to reconstruct and reduce the dimensionality of the collected multi-dimensional feature data. The original data is high-dimensional data with a high feature dimension, and the effect of feature fusion analysis may be poor. Therefore, based on data reconstruction, it is possible to effectively extract important feature information in the original data, and subsequently perform feature fusion and change analysis to generate corresponding transition features based on feature changes, thereby realizing feature inspection and defect location throughout the entire processing cycle.

[0029] According to an embodiment of the present invention, the second feature set of adjacent processing steps is imported into the ResNet model to generate a feature map, and feature change analysis is performed based on the feature map to generate an intermediate change feature set, specifically: The second feature sets of two adjacent processing steps are respectively imported into the ResNet model for feature learning and forward propagation of the input to the output layer, and feature maps of different levels are generated, and a plurality of first feature maps and a plurality of second feature maps are respectively marked based on the two adjacent processing steps; Based on the same level, the first feature map and the second feature map are respectively added and multiplied, the feature maps are fused, and a transition feature map is generated; Through multi-level feature map operations, multiple transition feature maps are generated. Based on the ResNet model, corresponding feature data is generated for all transition feature maps to form an intermediate change feature set. The feature variation analysis is performed between every two adjacent processing steps to generate multiple variation feature sets.

[0030] It should be noted that the feature maps at different levels correspond to feature map analysis of different feature dimensions, including contour, texture and color dimensions. Assume that two adjacent processing steps are S1 and S2, respectively. Step S1 corresponds to multiple first feature maps, and step S2 corresponds to multiple second feature maps. Through the addition and multiplication operations of feature maps, the processing feature information of different processing flows can be effectively fused and analyzed, and feature maps of intermediate transitions can be generated, and finally a change feature set can be formed. Through the change feature set, the multi-dimensional feature changes and transition change information in the mold processing process can be effectively mined. By mining the transition change feature information, accurate defect analysis and quality assessment of different processing steps can be achieved in the subsequent process, and precise positioning of the processing optimization steps can be achieved, and the corresponding processing parameters can be adjusted to effectively optimize the mold processing process.

[0031] According to an embodiment of the present invention, the full processing flow feature table is constructed according to the first feature set, the second feature set and the corresponding intermediate change feature set, real-time video data of the target FRP mold in N processing steps are collected and key frame extraction and feature extraction are performed to form a target feature set, and the target feature set is compared with the full processing flow feature table for feature comparison and processing defect identification, the existing defects are located in steps, and a processing step optimization plan is generated, specifically: Based on the processing sequence, the first feature set, the second feature set and the corresponding intermediate change feature set are sorted, and the processing node information of each feature set is associated to form a feature table of the entire processing process; Collect real-time video data of the target FRP mold in N processing steps and perform key frame extraction and multi-dimensional feature extraction, integrate the extracted features to form a target feature set; The target feature set is compared with the feature set in the feature table of the entire processing flow to identify processing defects. The comparison process analyzes the difference of features based on the standard Euclidean distance, and the feature set in the feature table of the entire processing flow is used as the standard feature to identify processing defects and locate the processing steps where the defects are located, generating processing defect information; The comparison process also includes a difference analysis between the target feature set and the intermediate change feature set in the full processing flow feature table; According to the processing defect information, the corresponding processing steps are optimized and the processing step optimization plan is generated.

[0032] It should be noted that the processing node information is used to store the processing steps and process relationships to which the corresponding feature set belongs. Multi-dimensional feature extraction includes contour, texture and color dimensions. The corresponding processing defect information includes defect category, defect corresponding processing step and other information.

[0033] During the comparison process, a feature table of the entire processing flow is introduced. The table includes a set of change features for multiple steps. During the comparison, the current processing step features (target feature set) can be compared with the change features in the table, the feature set of each step, and the feature map to analyze whether there are defects in the current processing step features. Based on the comparison of the change features, it is possible to locate in which step the defect occurs or between which two steps the defect belongs, and generate the corresponding optimization plan after locating the step.

[0034] It is worth mentioning here that in the production process of FRP molds, there are many processing defects, such as cracks, delamination, bubbles, scratches, uneven color and deformation, etc. Different types of defects generally involve one or more processing steps. Therefore, it is necessary to conduct a refined evaluation of processing defects and quality and locate the problem steps. And in some defects, it is difficult to identify them by relying on the characteristics of a single step. For example, in the two adjacent processing steps S1 and S2, there are bubbles in the mold, and they are only identified in S2 or subsequent steps, but they actually exist in the S1 step. At this time, based on the feature table of the entire processing process, the corresponding transition feature information can be effectively obtained from the adjacent steps, and the defect can be identified with high accuracy based on the transition features, so as to effectively locate the defect between the S1 and S2 steps, and then optimize and feedback the associated processing steps, realize real-time processing optimization and real-time feedback analysis of the entire processing process, improve processing quality and processing efficiency, and reduce the situation of relying on manual experience for processing evaluation. At the same time, during the processing process, the changes of minor quality problems can be effectively mined, and fast and efficient processing defect analysis and comparison can be achieved.

[0035] According to an embodiment of the present invention, according to the processing step optimization scheme, the second target FRP mold is processed and produced, and the processing quality is evaluated through the second target feature set and the full processing flow feature table collected in real time, and the real-time feedback analysis of the processing step optimization scheme is performed based on the evaluation, specifically: According to the processing step optimization plan, the processing steps of the second target FRP mold are optimized; The video data of the second target FRP mold is collected in real time and key frame extraction and image preprocessing are performed to generate a second target feature set; The second target feature set is compared with the feature table of the entire processing flow in real time for processing feature comparison and processing quality evaluation. The evaluation results are used to provide real-time feedback on the processing step optimization plan and real-time adjustment of the processing plan.

[0036] It should be noted that the corresponding production optimization plan of the first target FRP mold can be applied to the subsequent processing of the second target FRP mold.

[0037] According to an embodiment of the present invention, it also includes: During the use of the target FRP mold, the multi-dimensional image feature sets before and after use are collected in real time through machine vision technology and marked as the first mold feature set and the second mold feature set; Performing a difference analysis of contour and texture features according to the first mold feature set and the second mold feature set, and performing deformation defect evaluation by feature difference calculation to generate a first deformation index; Obtaining the first feature set and the second feature set of the final processing step according to the feature table of the entire processing flow, and integrating them into a comparison feature set; Performing difference analysis of contour and texture features on the second mold feature set and the comparison feature set, and performing deformation defect evaluation by feature difference calculation to generate a second deformation index; Based on the weighted average of the first deformation index and the second deformation index, a real-time deformation evaluation value is obtained, and a use warning of the target fiberglass reinforced plastic mold is issued through the real-time deformation evaluation value.

[0038] It should be noted that in the weighted average, the weights of the first deformation index and the second deformation index are set to Q1 and Q2 respectively, and Q1>Q2. The multi-dimensional image feature set mainly includes contour and texture feature dimensions for deformation analysis. Here, the difference analysis can calculate the feature difference based on the standard Euclidean distance, reflecting the contour difference and mold deformation.

[0039] The first mold feature set and the second mold feature set correspond to multi-dimensional image feature sets before and after use, respectively.

[0040] It can be understood here that the FRP mold is easy to deform and has poor thermal conductivity. In actual use, it is easy to produce micro deformation. In the existing technology, it is difficult to effectively analyze the impact of deformation on mold use and provide real-time warning, and it generally relies on manual observation. Therefore, the present invention analyzes the feature difference before and after use, and uses the feature set of the feature table of the entire processing flow to evaluate the deformation during use, comprehensively analyzes the deformation situation, and realizes early warning of mold use. Through the real-time deformation evaluation value, effective warning of the impact of mold use can be provided.

[0041] Figure 2 A block diagram of a glass fiber reinforced plastic mold processing system based on feedback analysis of the present invention is shown.

[0042] The second aspect of the present invention further provides a glass fiber reinforced plastic mold processing system 2 based on feedback analysis, the system comprising: a memory 21, a processor 22, the memory 21 comprising a glass fiber reinforced plastic mold processing program based on feedback analysis, the glass fiber reinforced plastic mold processing program based on feedback analysis being executed by the processor 22 to implement the following steps: Based on machine vision technology, obtain video data of the preset FRP mold in N processing steps; Extract key frames from the video data to form a processed image data set, identify the FRP mold area on the processed image data set, extract contour, texture and color features through the Prewitt operator and color histogram to form the first feature set; Based on the autoencoder network, data feature learning and data reconstruction are performed on the first feature set to generate a second feature set, the second feature sets of adjacent processing steps are imported into the ResNet model to generate feature graphs, and feature change analysis is performed based on the feature graphs to generate intermediate change feature sets; A full processing flow feature table is constructed based on the first feature set, the second feature set and the corresponding intermediate change feature set, real-time video data of the target FRP mold in N processing steps are collected and key frame extraction and feature extraction are performed to form a target feature set, and feature comparison and processing defect identification are performed between the target feature set and the full processing flow feature table, the existing defects are located in steps, and a processing step optimization plan is generated; According to the processing step optimization plan, the second target FRP mold is processed and produced, and the processing quality is evaluated through the real-time collected second target feature set and the full processing flow feature table. Based on the evaluation, real-time feedback analysis of the processing step optimization plan is carried out.

[0043] It should be noted that the production of the FRP mold generally includes multiple steps, such as raw material placement, spraying gel coat, spraying glass fiber multiple layers, waiting for molding, curing, applying resin, polishing, etc. Each step corresponds to different mold processing image features, and the features of adjacent processing steps in the county often have a certain correlation. For example, the color features and contour features of the processing steps before and after curing have certain changes and correlations, and the layering processing steps of different glass fibers. In the present invention, by simulating and mining the feature changes of adjacent processing steps and generating corresponding transition features, efficient and rapid processing feature defect identification and step positioning can be achieved in the future, and corresponding step optimization can be performed, such as re-evaluation and replacement of raw materials, regulation of mixture ratio, and regulation of resin usage.

[0044] According to an embodiment of the present invention, the video data of the preset FRP mold in N processing steps is obtained based on machine vision technology, specifically: Based on machine vision technology, through a preset high-definition camera device, image data of the entire processing process of the preset fiberglass mold is acquired to obtain corresponding video data; The whole processing includes N processing steps.

[0045] It should be noted that the preset FRP mold is a selected standardized production mold, which can be screened based on mold test evaluation and has a good comparison effect. It is used to generate the corresponding full processing flow feature table.

[0046] According to an embodiment of the present invention, the key frames of the video data are extracted to form a processed image data set, the fiberglass mold area is identified on the processed image data set, and the contour, texture and color features are extracted by the Prewitt operator and the color histogram to form a first feature set, specifically: Extract key frames from video data to form processed image data; Perform image noise reduction, enhancement, and data standardization preprocessing on the processed image data; The FRP mold area is identified on the processed image data set to obtain the mold area image. The edge of the mold area image is detected by the Prewitt operator, and the contour features and texture features are extracted. Generate a color histogram based on the mold area image, and generate color feature data through statistical information of the color histogram; A first feature set is formed based on the contour feature, texture feature and color feature data.

[0047] It should be noted that the first feature set is a multi-dimensional image feature that describes the mold during the processing, which is used for subsequent processing quality and defect analysis of different processing steps.

[0048] According to an embodiment of the present invention, in the N processing steps, each processing step corresponds to a processed image data set.

[0049] According to an embodiment of the present invention, based on the autoencoder network, data feature learning and data reconstruction are performed on the first feature set to generate a second feature set, the second feature sets of adjacent processing steps are imported into the ResNet model to generate feature graphs, and feature change analysis is performed based on the feature graphs to generate intermediate change feature sets, specifically: Construct an autoencoder network, import the first feature set as raw data into the autoencoder network, map the imported raw data to a low-dimensional hidden representation through the encoder, and perform the mapping process through a multi-layer perceptron to obtain a low-dimensional output; The low-dimensional output is reconstructed through the decoder to generate a high-dimensional output. A preset loss function is introduced to compare the high-dimensional output with the original data input, and the autoencoder network parameters are corrected. The data reconstruction is repeated until the loss function converges to the preset value. The high-dimensional output of the autoencoder network at this time is recorded to obtain the second feature set.

[0050] It should be noted that the autoencoder network includes an encoder, a decoder and a loss function. The preset loss function can be calculated using mean square error.

[0051] Here, the present invention introduces an autoencoder network to reconstruct and reduce the dimensionality of the collected multi-dimensional feature data. The original data is high-dimensional data with a high feature dimension, and the effect of feature fusion analysis may be poor. Therefore, based on data reconstruction, it is possible to effectively extract important feature information in the original data, and subsequently perform feature fusion and change analysis to generate corresponding transition features based on feature changes, thereby realizing feature inspection and defect location throughout the entire processing cycle.

[0052] According to an embodiment of the present invention, the second feature set of adjacent processing steps is imported into the ResNet model to generate a feature map, and feature change analysis is performed based on the feature map to generate an intermediate change feature set, specifically: The second feature sets of two adjacent processing steps are respectively imported into the ResNet model for feature learning and forward propagation of the input to the output layer, and feature maps of different levels are generated, and a plurality of first feature maps and a plurality of second feature maps are respectively marked based on the two adjacent processing steps; Based on the same level, the first feature map and the second feature map are respectively added and multiplied, the feature maps are fused, and a transition feature map is generated; Through multi-level feature map operations, multiple transition feature maps are generated. Based on the ResNet model, corresponding feature data is generated for all transition feature maps to form an intermediate change feature set. The feature variation analysis is performed between every two adjacent processing steps to generate multiple variation feature sets.

[0053] It should be noted that the feature maps at different levels correspond to feature map analysis of different feature dimensions, including contour, texture and color dimensions. Assume that two adjacent processing steps are S1 and S2, respectively. Step S1 corresponds to multiple first feature maps, and step S2 corresponds to multiple second feature maps. Through the addition and multiplication operations of feature maps, the processing feature information of different processing flows can be effectively fused and analyzed, and feature maps of intermediate transitions can be generated, and finally a change feature set can be formed. Through the change feature set, the multi-dimensional feature changes and transition change information in the mold processing process can be effectively mined. By mining the transition change feature information, accurate defect analysis and quality assessment of different processing steps can be achieved in the subsequent process, and precise positioning of the processing optimization steps can be achieved, and the corresponding processing parameters can be adjusted to effectively optimize the mold processing process.

[0054] According to an embodiment of the present invention, the full processing flow feature table is constructed according to the first feature set, the second feature set and the corresponding intermediate change feature set, real-time video data of the target FRP mold in N processing steps are collected and key frame extraction and feature extraction are performed to form a target feature set, and the target feature set is compared with the full processing flow feature table for feature comparison and processing defect identification, the existing defects are located in steps, and a processing step optimization plan is generated, specifically: Based on the processing sequence, the first feature set, the second feature set and the corresponding intermediate change feature set are sorted, and the processing node information of each feature set is associated to form a feature table of the entire processing process; Collect real-time video data of the target FRP mold in N processing steps and perform key frame extraction and multi-dimensional feature extraction, integrate the extracted features to form a target feature set; The target feature set is compared with the feature set in the feature table of the entire processing flow to identify processing defects. The comparison process analyzes the difference of features based on the standard Euclidean distance, and the feature set in the feature table of the entire processing flow is used as the standard feature to identify processing defects and locate the processing steps where the defects are located, generating processing defect information; The comparison process also includes a difference analysis between the target feature set and the intermediate change feature set in the full processing flow feature table; According to the processing defect information, the corresponding processing steps are optimized and the processing step optimization plan is generated.

[0055] It should be noted that the processing node information is used to store the processing steps and process relationships to which the corresponding feature set belongs. Multi-dimensional feature extraction includes contour, texture and color dimensions. The corresponding processing defect information includes defect category, defect corresponding processing step and other information.

[0056] During the comparison process, a feature table of the entire processing flow is introduced. The table includes a set of change features for multiple steps. During the comparison, the current processing step features (target feature set) can be compared with the change features in the table, the feature set of each step, and the feature map to analyze whether there are defects in the current processing step features. Based on the comparison of the change features, it is possible to locate in which step the defect occurs or between which two steps the defect belongs, and generate the corresponding optimization plan after locating the step.

[0057] It is worth mentioning here that in the production process of FRP molds, there are many processing defects, such as cracks, delamination, bubbles, scratches, uneven color and deformation, etc. Different types of defects generally involve one or more processing steps. Therefore, it is necessary to conduct a refined evaluation of processing defects and quality and locate the problem steps. And in some defects, it is difficult to identify them by relying on the characteristics of a single step. For example, in the two adjacent processing steps S1 and S2, there are bubbles in the mold, and they are only identified in S2 or subsequent steps, but they actually exist in the S1 step. At this time, based on the feature table of the entire processing process, the corresponding transition feature information can be effectively obtained from the adjacent steps, and the defect can be identified with high accuracy based on the transition features, so as to effectively locate the defect between the S1 and S2 steps, and then optimize and feedback the associated processing steps, realize real-time processing optimization and real-time feedback analysis of the entire processing process, improve processing quality and processing efficiency, and reduce the situation of relying on manual experience for processing evaluation. At the same time, during the processing process, the changes of minor quality problems can be effectively mined, and fast and efficient processing defect analysis and comparison can be achieved.

[0058] According to an embodiment of the present invention, according to the processing step optimization scheme, the second target FRP mold is processed and produced, and the processing quality is evaluated through the second target feature set and the full processing flow feature table collected in real time, and the real-time feedback analysis of the processing step optimization scheme is performed based on the evaluation, specifically: According to the processing step optimization plan, the processing steps of the second target FRP mold are optimized; The video data of the second target FRP mold is collected in real time and key frame extraction and image preprocessing are performed to generate a second target feature set; The second target feature set is compared with the feature table of the entire processing flow in real time for processing feature comparison and processing quality evaluation. The evaluation results are used to provide real-time feedback on the processing step optimization plan and real-time adjustment of the processing plan.

[0059] It should be noted that the corresponding production optimization plan of the first target FRP mold can be applied to the subsequent processing of the second target FRP mold.

[0060] The third aspect of the present invention also provides a computer-readable storage medium, which includes a FRP mold processing program based on feedback analysis. When the FRP mold processing program based on feedback analysis is executed by a processor, the steps of the FRP mold processing method based on feedback analysis as described in any one of the above items are implemented.

[0061] The present invention discloses a glass fiber reinforced plastic mold processing method and system based on feedback analysis. By acquiring video data of N processing steps, extracting key frames to form a data set, using Prewitt operator and color histogram to extract contour, texture and color features, and constructing a first feature set. Using an autoencoder network to learn and reconstruct data features, a second feature set is generated, and the feature changes are analyzed in combination with a ResNet model to form an intermediate change feature set, and a full processing flow feature table is constructed. The real-time target feature set and the feature table are compared to identify processing defects and locate problem steps, generate optimization solutions, realize accurate optimization of processing steps and real-time feedback analysis, improve processing quality and processing efficiency, and reduce the situation of relying on manual experience for processing evaluation.

[0062] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0063] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0064] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0065] A person of ordinary skill in the art can understand that: all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: a mobile storage device, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0066] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0067] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A glass fiber reinforced plastic mold processing method based on feedback analysis, characterized in that: include: Based on machine vision technology, obtain the video data of the preset FRP mold in N processing steps; Extract key frames from the video data to form a processed image data set, identify the FRP mold area on the processed image data set, extract contour, texture and color features through the Prewitt operator and color histogram to form the first feature set; Based on the autoencoder network, data feature learning and data reconstruction are performed on the first feature set to generate a second feature set, the second feature sets of adjacent processing steps are imported into the ResNet model to generate feature graphs, and feature change analysis is performed based on the feature graphs to generate intermediate change feature sets; A full processing flow feature table is constructed based on the first feature set, the second feature set and the corresponding intermediate change feature set, real-time video data of the target FRP mold in N processing steps are collected and key frame extraction and feature extraction are performed to form a target feature set, and feature comparison and processing defect identification are performed between the target feature set and the full processing flow feature table, the existing defects are located in steps, and a processing step optimization plan is generated; According to the processing step optimization plan, the second target FRP mold is processed and produced, and the processing quality is evaluated through the real-time collected second target feature set and the full processing flow feature table. Based on the evaluation, real-time feedback analysis of the processing step optimization plan is carried out.

2. The method for processing a glass fiber reinforced plastic mold based on feedback analysis according to claim 1, characterized in that: The method of obtaining the video data of the preset FRP mold in N processing steps based on machine vision technology is as follows: Based on machine vision technology, through a preset high-definition camera device, image data of the entire processing process of the preset fiberglass mold is acquired to obtain corresponding video data; The whole processing includes N processing steps.

3. The method for processing a glass fiber reinforced plastic mold based on feedback analysis according to claim 1, characterized in that: The key frames of the video data are extracted to form a processed image data set, the fiberglass mold area is identified on the processed image data set, and the contour, texture and color features are extracted by the Prewitt operator and the color histogram to form a first feature set, which is specifically: Extract key frames from video data to form processed image data; Perform image noise reduction, enhancement, and data standardization preprocessing on the processed image data; The FRP mold area is identified on the processed image data set to obtain the mold area image. The edge of the mold area image is detected by the Prewitt operator, and the contour features and texture features are extracted. Generate a color histogram based on the mold area image, and generate color feature data through statistical information of the color histogram; A first feature set is formed based on the contour feature, texture feature and color feature data.

4. The method for processing a glass fiber reinforced plastic mold based on feedback analysis according to claim 3 is characterized in that: Among the N processing steps, each processing step corresponds to a processed image data set.

5. The method for processing a glass fiber reinforced plastic mold based on feedback analysis according to claim 4 is characterized in that: Based on the autoencoder network, data feature learning and data reconstruction are performed on the first feature set to generate a second feature set, the second feature sets of adjacent processing steps are imported into the ResNet model to generate feature graphs, and feature change analysis is performed based on the feature graphs to generate intermediate change feature sets, specifically: Construct an autoencoder network, import the first feature set as raw data into the autoencoder network, map the imported raw data to a low-dimensional hidden representation through the encoder, and perform the mapping process through a multi-layer perceptron to obtain a low-dimensional output; The low-dimensional output is reconstructed through the decoder to generate a high-dimensional output. A preset loss function is introduced to compare the high-dimensional output with the original data input, and the autoencoder network parameters are corrected. The data reconstruction is repeated until the loss function converges to the preset value. The high-dimensional output of the autoencoder network at this time is recorded to obtain the second feature set.

6. The method for processing a glass fiber reinforced plastic mold based on feedback analysis according to claim 5, characterized in that: The second feature set of the adjacent processing steps is imported into the ResNet model to generate a feature map, and feature change analysis is performed based on the feature map to generate an intermediate change feature set, specifically: The second feature sets of two adjacent processing steps are respectively imported into the ResNet model for feature learning and forward propagation of the input to the output layer, and feature maps of different levels are generated, and a plurality of first feature maps and a plurality of second feature maps are respectively marked based on the two adjacent processing steps; Based on the same level, the first feature map and the second feature map are respectively added and multiplied, the feature maps are fused, and a transition feature map is generated; Through multi-level feature map operations, multiple transition feature maps are generated. Based on the ResNet model, corresponding feature data is generated for all transition feature maps to form an intermediate change feature set. The feature variation analysis is performed on every two adjacent processing steps to generate multiple variation feature sets.

7. The method for processing a glass fiber reinforced plastic mold based on feedback analysis according to claim 6, characterized in that: The full processing flow feature table is constructed according to the first feature set, the second feature set and the corresponding intermediate change feature set, the real-time video data of the target FRP mold in N processing steps are collected and key frame extraction and feature extraction are performed to form a target feature set, the target feature set is compared with the full processing flow feature table and processing defect identification is performed, the existing defects are located in steps, and a processing step optimization plan is generated, specifically: Based on the processing sequence, the first feature set, the second feature set and the corresponding intermediate change feature set are sorted, and the processing node information of each feature set is associated to form a feature table of the entire processing process; Collect real-time video data of the target FRP mold in N processing steps and perform key frame extraction and multi-dimensional feature extraction, integrate the extracted features to form a target feature set; The target feature set is compared with the feature set in the feature table of the entire processing flow to identify processing defects. The comparison process analyzes the difference of features based on the standard Euclidean distance, and the feature set in the feature table of the entire processing flow is used as the standard feature to identify processing defects and locate the processing steps where the defects are located, generating processing defect information; The comparison process also includes a difference analysis between the target feature set and the intermediate change feature set in the full processing flow feature table; According to the processing defect information, the corresponding processing steps are optimized and the processing step optimization plan is generated.

8. The method for processing a glass fiber reinforced plastic mold based on feedback analysis according to claim 7, characterized in that: According to the processing step optimization plan, the second target FRP mold is processed and produced, and the processing quality is evaluated through the second target feature set and the full processing flow feature table collected in real time, and the real-time feedback analysis of the processing step optimization plan is performed based on the evaluation, specifically: According to the processing step optimization plan, the processing steps of the second target FRP mold are optimized; The video data of the second target FRP mold is collected in real time and key frame extraction and image preprocessing are performed to generate a second target feature set; The second target feature set is compared with the feature table of the entire processing flow in real time for processing feature comparison and processing quality evaluation. The evaluation results are used to provide real-time feedback on the processing step optimization plan and real-time adjustment of the processing plan.

9. A glass fiber reinforced plastic mold processing system based on feedback analysis, characterized in that: The system includes: a memory and a processor, wherein the memory includes a glass fiber reinforced plastic mold processing program based on feedback analysis, and the glass fiber reinforced plastic mold processing program based on feedback analysis is executed by the processor to implement the following steps: Based on machine vision technology, obtain the video data of the preset FRP mold in N processing steps; Extract key frames from the video data to form a processed image data set, identify the FRP mold area on the processed image data set, extract contour, texture and color features through the Prewitt operator and color histogram to form the first feature set; Based on the autoencoder network, data feature learning and data reconstruction are performed on the first feature set to generate a second feature set, the second feature sets of adjacent processing steps are imported into the ResNet model to generate feature graphs, and feature change analysis is performed based on the feature graphs to generate intermediate change feature sets; A full processing flow feature table is constructed based on the first feature set, the second feature set and the corresponding intermediate change feature set, real-time video data of the target FRP mold in N processing steps are collected and key frame extraction and feature extraction are performed to form a target feature set, and feature comparison and processing defect identification are performed between the target feature set and the full processing flow feature table, the existing defects are located in steps, and a processing step optimization plan is generated; According to the processing step optimization plan, the second target FRP mold is processed and produced, and the processing quality is evaluated through the real-time collected second target feature set and the full processing flow feature table. Based on the evaluation, real-time feedback analysis of the processing step optimization plan is carried out.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a FRP mold processing program based on feedback analysis. When the FRP mold processing program based on feedback analysis is executed by a processor, the steps of the FRP mold processing method based on feedback analysis as described in any one of claims 1 to 8 are implemented.

Images