Sorting method of sheet type material receiving machine, material receiving machine, program product and medium

Abstract

The invention provides a sorting method of a sheet type material receiving machine, the material receiving machine, a program product and a medium, and relates to the technical field of sheet type material sorting. By calculating the similarity between the current image data and the multiple pieces of template data, the multiple first similarities are obtained, when only one first similarity exceeds the preset threshold value, the material type is directly determined, and the processing efficiency is improved. When the plurality of first similarities exceed a threshold value, a feature extraction and reconstruction step is introduced to generate current reconstruction image data and template reconstruction data for enhancing features (distinguishing points) between the current image data and the template data and amplifying distinguishing features (distinguishing points) between different sheet materials; according to the method, the template data with the first similarities exceeding the threshold value show more obvious differences after reconstruction, and finally, the material types are determined based on the maximum average similarity, so that the classification accuracy of the sheet materials is improved.

Description

Technical Field

[0001] This application relates to the technical field of chip material sorting, and particularly to a sorting method, a receiving machine, a program product, and a medium for a chip receiving machine. Background Art

[0002] With the rapid development of industrial automation, chip receiving machines have been widely used in various industries. Chip receiving machines can automatically collect, sort, and store various sheet materials, greatly improving production efficiency and accuracy.

[0003] Currently, common sorting methods for chip receiving machines mainly rely on image recognition technology. This method obtains the image data of the current chip material and then compares it with the pre-stored standard images. By calculating the similarity between the images, the system determines which category the current material belongs to.

[0004] However, in practical applications, due to the characteristics of chip materials: flat and their appearances may be very similar. In this case, the comparison may be difficult to accurately distinguish different types of materials, resulting in a low classification accuracy of chip materials. Summary of the Invention

[0005] This application provides a sorting method, a receiving machine, a program product, and a medium for a chip receiving machine to improve the classification accuracy of chip materials.

[0006] In a first aspect, this application provides a sorting method for a chip receiving machine, including: obtaining the current image data of the current chip material; respectively calculating the similarities between the current image data and a plurality of stored template data to obtain a plurality of first similarities, where the template data is the standard image data of a type of chip material; when only one first similarity is greater than a preset similarity threshold, determining the first template data corresponding to the first similarity greater than the preset similarity threshold; determining the type of chip material corresponding to the first template data for the current chip material; when at least two first similarities are greater than the preset similarity threshold, performing feature extraction and reconstruction on the current image data to obtain current reconstructed image data; calculating the similarities between the current reconstructed image data and the reconstructed data corresponding to a plurality of stored template data to obtain a plurality of second similarities, where the reconstructed data is obtained by performing feature extraction and reconstruction on the template data; performing weighted averaging on the first similarities and the corresponding second similarities to obtain a plurality of average similarities; determining the second template data corresponding to the maximum average similarity; and determining the type of chip material corresponding to the second template data for the current chip material.

[0007] By adopting the above technical solution, by calculating the similarity between the current image data and multiple template data, multiple first similarities are obtained. When only one first similarity exceeds the preset threshold, the material type is directly determined, improving the processing efficiency. When multiple first similarities exceed the threshold, feature extraction and reconstruction steps are introduced, so that the template data with multiple first similarities exceeding the threshold show more obvious differences after reconstruction. Finally, the material type is determined based on the maximum average similarity to improve the classification accuracy of chip materials.

[0008] Combined with some embodiments of the first aspect, in some embodiments, after the step of determining the second template data corresponding to the maximum average similarity, the method further includes: determining whether the difference between the maximum average similarity and the second largest average similarity is greater than a preset difference threshold; if the difference is greater than the preset difference threshold, then performing the step of determining the current chip material as the chip material type corresponding to the second template data; if the difference is not greater than the preset difference threshold, then determining the third template data corresponding to the second largest average similarity; cropping the same regions of the current image data, the second template data, and the third template data, where the same regions are regions with the same position and the same features; respectively performing feature extraction and reconstruction on the cropped current image data, the second template data, and the third template data to obtain a sub-current reconstructed image data, a sub-second reconstructed data, and a sub-third reconstructed data; respectively calculating the similarity between the sub-current reconstructed image data and the sub-second reconstructed data and the sub-third reconstructed data to obtain a third similarity; determining the current chip material as the chip material type corresponding to the template data with the maximum third similarity.

[0009] By adopting the above technical solution, when the difference between the maximum and the second largest average similarities exceeds the preset threshold, the material type is directly determined, ensuring the processing efficiency. When the difference is insufficient, further analysis is performed. By cropping the same regions of the current image data and the relevant template data, the same features (common parts) are removed, focusing on the most distinguishable features (differentiating points). Feature extraction and reconstruction are performed on the cropped data to obtain more refined reconstructed data, the similarity between the sub-current reconstructed image data and the sub-template reconstructed data is calculated to obtain a third similarity, and finally the material type is determined based on the maximum third similarity. Through the fine comparison of local features, the problem that global reconstruction may mask local differences is effectively solved.

[0010] In some embodiments in combination with some embodiments of the first aspect, after the step of determining the second template data corresponding to the maximum average similarity, the method further includes: determining whether the differences between the maximum average similarity and other average similarities are all greater than a preset difference threshold; if all the differences are greater than the preset difference threshold, performing the step of determining the current sheet material as the sheet material type corresponding to the second template data; if there is at least one difference not greater than the preset difference threshold, determining the fourth template data corresponding to all the average similarities whose differences from the maximum average similarity are not greater than the preset difference threshold; cropping the same regions of the current image data and the fourth template data according to a preset cropping rule, where the same regions are regions with the same position and the same features; respectively performing feature extraction and reconstruction on the cropped current image data and the fourth template data to obtain a sub-current reconstructed image data and a sub-fourth reconstructed data; respectively calculating the similarities between the sub-current reconstructed image data and the sub-fourth reconstructed data to obtain a fourth similarity; determining whether the differences between the maximum fourth similarity and other fourth similarities are all greater than the preset difference threshold; if all the differences are greater than the preset difference threshold, determining the fifth template data corresponding to the maximum fourth similarity; determining the current sheet material as the sheet material type corresponding to the fifth template data; if there is at least one difference not greater than the preset difference threshold, determining the fourth similarity as the average similarity, and performing the step of determining the fourth template data corresponding to all the average similarities whose differences from the maximum average similarity are not greater than the preset difference threshold.

[0011] By adopting the above technical solution, when judging the differences between the maximum average similarity and all other similarities. If all the differences are greater than the preset threshold, the material type is directly determined, ensuring the processing efficiency. When there are similar similarities, determining all the similar template data as the fourth template data. Cropping the current image data and the fourth template data according to a preset rule, focusing on the key feature regions. Performing feature extraction and reconstruction on the cropped data to obtain more refined sub-reconstructed data. Calculating the fourth similarity between the sub-reconstructed data for a new round of difference judgment. If there is an obvious maximum similarity at this time, the material type is determined; otherwise, taking the fourth similarity as the new average similarity and repeating the whole process. This progressive analysis method effectively solves the problem of high similarity among multiple template data. Through multiple focusings and comparisons, the best match can be gradually selected from the highly similar candidates, improving the classification accuracy.

[0012] In some embodiments in combination with some embodiments of the first aspect, the preset cropping rule is as follows: determine all continuous regions in the current image data and the fourth template data that have the same position and the same features; select the region with the largest area from all the continuous regions as the initial cropping region; perform the first cropping on the initial cropping region with a first preset ratio; perform subsequent cropping on the initial cropping region, and the ratio of each cropping is the first preset ratio.

[0013] By adopting the above technical solution, all continuous regions in the current image data and the fourth template data that have the same position and the same features are determined, the spatial structure information is maintained, and additional distinguishing features are avoided after subsequent cropping. The region with the largest area is selected from these regions as the initial cropping region, ensuring that the selected region is a relatively large region, which is convenient for subsequent cropping.

[0014] In some embodiments in combination with some embodiments of the first aspect, before the step of performing subsequent cropping on the initial cropping region with the ratio of each cropping being the first preset ratio, the method further includes: before each subsequent cropping is performed, determine whether the initial cropping region is the region with the largest selected area; if it is the region with the largest area, perform the step of performing subsequent cropping on the initial cropping region with the ratio of each cropping being the first preset ratio; if it is not the region with the largest area, perform the step of performing subsequent cropping on the initial cropping region with the ratio of each cropping being the first preset ratio; then select the region with the largest area from all the continuous regions as the subsequent cropping region; perform subsequent cropping on the subsequent cropping region, and the ratio of each cropping is the first preset ratio.

[0015] By adopting the above technical solution, before each subsequent cropping is performed, it is determined whether the current cropping region is still the region with the largest area, and the cropping is continued, maintaining the continuity of the cropping. If it is no longer the largest region, the region with the largest area is reselected as the new cropping object, maximizing the efficiency of removing the same features and maximizing the effect of focusing on the most distinctive part.

[0016] In some embodiments in combination with some embodiments of the first aspect, the step of cropping the same regions of the current image data and the fourth template data according to the preset cropping rule specifically includes: determining the regions in the current image data and each template data that have the same position and the same features to obtain the corresponding same regions; wherein, the same regions determined for different template data are independent of each other; cropping the same regions of the current image data and each template data.

[0017] By adopting the above technical solution, first, the regions where the current image data has the same position and features as each template data are determined. This independent region determination method allows for capturing specific similar features for different template data, enabling more precise identification of the local similarity between the current image and each template, which is used to expand the cropping region and facilitate subsequent multiple croppings.

[0018] In combination with some embodiments of the first aspect, in some embodiments, the steps of extracting features from and reconstructing the current image data to obtain the current reconstructed image data specifically include: using an encoder including multiple convolutional layers and pooling layers to extract features from the current image data to generate a compressed feature representation; using a decoder including an upsampling layer to reconstruct the image using the compressed feature representation to generate the current reconstructed image data, where the decoder is configured to expand the compressed feature representation back to the size of the original image.

[0019] By adopting the above technical solution, first, an encoder including multiple convolutional layers and pooling layers is used to extract features from the current image data. This deep network structure can effectively capture the multi-level features of the image, from low-level textures to high-level semantic information. The encoding process generates a compressed feature representation, greatly reducing the data dimension and improving the efficiency of subsequent processing. Subsequently, a decoder including an upsampling layer is used to reconstruct the image using the compressed feature representation. The decoder is configured to expand the compressed feature representation back to the size of the original image. This process not only restores the spatial structure of the image but also retains the key feature information. Through this encoding-decoding process, the essential features of the image can be effectively extracted and reconstructed.

[0020] In a second aspect, the present application provides a sheet material collecting machine, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, and the computer program code includes computer instructions. The one or more processors call the computer instructions to cause the sheet material collecting machine to execute the method described in the first aspect and any possible implementation manner of the first aspect.

[0021] In a third aspect, the present application provides a computer program product including instructions, which, when running on a sheet material collecting machine, causes the sheet material collecting machine to execute the method described in the first aspect and any possible implementation manner of the first aspect.

[0022] In a fourth aspect, the present application provides a computer-readable storage medium including instructions, which, when running on a sheet material collecting machine, causes the sheet material collecting machine to execute the method described in the first aspect and any possible implementation manner of the first aspect.

[0023] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. By calculating the similarity between the current image data and multiple template data, multiple first similarities are obtained. When only one first similarity exceeds the preset threshold, the material type is directly determined, improving the processing efficiency. When multiple first similarities exceed the threshold, feature extraction and reconstruction steps are introduced to generate the current reconstructed image data and template reconstructed data, which are used to enhance the features (differentiating points) between the current image data and the template data, magnifying the differentiating features (differentiating points) between different chip materials, making the template data with multiple first similarities exceeding the threshold show more obvious differences after reconstruction. Finally, the material type is determined based on the maximum average similarity to improve the classification accuracy of chip materials.

[0024] 2. When the difference between the maximum and the second - largest average similarities exceeds the preset threshold, the material type is directly determined, ensuring the processing efficiency. When the difference is insufficient, further analysis is carried out. By cropping the same regions of the current image data and the relevant template data, the same features (common parts) are removed, focusing on the most differentiating features (differentiating points). Feature extraction and reconstruction are performed on the cropped data to obtain more refined reconstructed data, the similarity between the sub - current reconstructed image data and the sub - template reconstructed data is calculated to obtain the third similarity, and finally the material type is determined based on the maximum third similarity. Through the fine comparison of local features, the problem that global reconstruction may mask local differences is effectively solved.

[0025] 3. Judge the difference between the maximum average similarity and all other similarities. If all differences are greater than the preset threshold, the material type is directly determined, ensuring the processing efficiency. When there are similar similarities, all similar template data are determined as the fourth template data. The current image data and the fourth template data are cropped according to preset rules, focusing on the key feature regions. Feature extraction and reconstruction are performed on the cropped data to obtain more refined sub - reconstructed data. The fourth similarity between the sub - reconstructed data is calculated for a new round of difference judgment. If an obvious maximum similarity appears at this time, the material type is determined; otherwise, the fourth similarity is used as the new average similarity, and the whole process is repeated. This progressive analysis method effectively solves the problem of high similarity among multiple template data. Through multiple focusings and comparisons, the best match can be gradually selected from highly similar candidates, improving the classification accuracy. Description of the Drawings

[0026] Figure 1 is a flow schematic diagram of the sorting method of the chip receiving machine in an embodiment of the present application; Figure 2 is another flow schematic diagram of the sorting method of the chip receiving machine in an embodiment of the present application; Figure 3 is another flow schematic diagram of the sorting method of the chip receiving machine in an embodiment of the present application; Figure 4 is another schematic flowchart of the sorting method of the sheet material collector in the embodiments of the present application; Figure 5 is an exemplary hardware structure schematic diagram of the material collector in the embodiments of the present application. Detailed implementation manners

[0027] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to limit the present application. As used in the specification and claims of the present application, the singular forms "a", "an", "the", "above-mentioned", "said", and "this" are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term " / and / " used in the present application refers to and includes any or all possible combinations of one or more of the listed items.

[0028] Hereinafter, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as implying or suggesting relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of the present application, unless otherwise specified, the meaning of "a plurality" is two or more.

[0029] Please refer to Figure 1 , which is a schematic flowchart of the sorting method of the sheet material collector in the embodiments of the present application.

[0030] S101. Obtain the current image data of the current sheet material; The current sheet material is the object currently being classified; In some embodiments, the image acquisition device takes a picture or scans the current sheet material to be classified to obtain its digital image.

[0031] S102. Calculate the similarities between the current image data and several stored template data respectively to obtain several first similarities, where the template data is the standard image data of a kind of sheet material; The template data refers to the standard image data of various known sheet materials pre-stored therein, which is used as a comparison benchmark.

[0032] In some embodiments, all the pre-stored template data is retrieved from the database. Then, a suitable image similarity calculation algorithm is selected to compare the current image data with each template data and calculate the similarity between them.

[0033] In some specific embodiments, the image similarity calculation algorithm is as follows: extract the feature vectors of the current image data and the template data; calculate the Euclidean distance between the feature vectors; convert the Euclidean distance into a similarity value between 0 and 1; store the calculation result as the first similarity array, which is not limited here.

[0034] In some specific embodiments, the image similarity calculation algorithm is as follows: convert the current image data and the template data into grayscale images; calculate the structural similarity index of the two images; use the structural similarity index value as the similarity; record the calculation result as the first similarity list, which is not limited here.

[0035] S103. When only one first similarity is greater than the preset similarity threshold, determine the first template data corresponding to the first similarity greater than the preset similarity threshold. The preset similarity threshold refers to a pre-set judgment criterion used to determine whether the similarity is high enough to make a classification decision.

[0036] In some embodiments, if it is found that only one first similarity is greater than this threshold, it is considered to be a clear matching result. Then, the template data corresponding to this first similarity exceeding the threshold will be determined and marked as the first template data.

[0037] S104. Determine the type of the current chip material as the chip material type corresponding to the first template data. In some embodiments, access the pre-established template database, search for the relevant information of the first template data, and then extract the chip material type information associated with the template data to complete the classification of the current chip material.

[0038] S105. When at least two first similarities are greater than the preset similarity threshold, perform feature extraction and reconstruction on the current image data to obtain the current reconstructed image data. In some embodiments, it is confirmed that multiple template data have high similarity with the current image. Then, start the feature extraction algorithm, such as a convolutional neural network (CNN), to extract key features from the current image data. These features may include information such as edges, textures, and shapes. Then, use these extracted features to reconstruct the image through a reconstruction algorithm (such as an autoencoder). Finally, obtain a current reconstructed image data with enhanced features. This data usually can highlight the key features (differentiating points) of the material more than the original image.

[0039] It should be noted that performing feature extraction and reconstruction on the current image data to obtain the current reconstructed image data is used to enhance the features (differentiating points) between the current image data and the template data, and magnify the differentiating features (differentiating points) between different chip materials.

[0040] In some specific embodiments, step S105 specifically includes: S1051. Use an encoder including multiple convolutional layers and pooling layers to extract features from the current image data to generate a compressed feature representation; An encoder refers to a neural network structure used to convert an input image into a compressed representation.

[0041] In some embodiments, a deep neural network structure composed of multiple convolutional layers and pooling layers is used as the encoder. The current image data first passes through the convolutional layers, which can capture low-level features such as local textures and edges of the image, as well as more complex high-level features. Then, these features are downsampled through the pooling layers, gradually reducing the spatial size of the feature map while retaining the most significant feature information. Through this progressive processing, the information of the image is condensed into a highly abstract and compressed feature representation, which retains the key information of the original image but greatly reduces the data volume.

[0042] In some specific embodiments, a series of convolutional blocks are constructed. Each block contains 2 - 3 convolutional layers and a max pooling layer. Then, the image is processed through these blocks in sequence, gradually increasing the number of feature channels while reducing the spatial size. Finally, a fully connected layer or a global average pooling layer is used to obtain the final compressed feature representation, which is not limited herein.

[0043] S1052. Use a decoder including an upsampling layer to reconstruct the image using the compressed feature representation to generate the current reconstructed image data, where the decoder is configured to expand the compressed feature representation back to the size of the original image.

[0044] A decoder refers to a neural network structure used to restore the compressed feature representation to an image.

[0045] In some embodiments, a decoder network including multiple upsampling layers is used. The decoder receives the compressed feature representation output by the encoder as input and then gradually increases the spatial dimension of the feature map through a series of upsampling operations. In this process, the decoder learns how to convert the abstract feature information back to pixel-level image details. Upsampling may use methods such as transposed convolution (deconvolution), nearest neighbor interpolation, or bilinear interpolation. The last layer of the decoder usually uses a convolutional layer to convert the feature map into an output with the same number of channels as the original image. In this way, the decoder can generate a reconstructed image with the same size as the original image but possibly different in details.

[0046] In some specific embodiments, an upsampling hierarchy symmetric to the encoder structure is constructed. Then, after each upsampling step, the corresponding encoder feature map is concatenated with the current feature map to retain more detailed information. Finally, a series of convolutional layers are used to finely adjust the features to generate the final reconstructed image, which is not limited herein.

[0047] It can be seen that first, an encoder including multiple convolutional layers and pooling layers is used to extract features from the current image data. This deep network structure can effectively capture the multi-level features of the image, from low-level textures to high-level semantic information. The encoding process generates a compressed feature representation, greatly reducing the data dimension and improving the efficiency of subsequent processing. Subsequently, a decoder including upsampling layers is used to reconstruct the image using the compressed feature representation. The decoder is configured to expand the compressed feature representation back to the size of the original image. This process not only restores the spatial structure of the image but also retains the key feature information. Through this encoding-decoding process, the essential features of the image can be effectively extracted and reconstructed.

[0048] S106. Calculate the similarity between the current reconstructed image data and the reconstructed data corresponding to a number of stored template data to obtain a number of second similarities. The reconstructed data is obtained by performing feature extraction and reconstruction on the template data. Among them, the reconstructed data corresponding to the template data refers to the data obtained after performing the same feature extraction and reconstruction processing on each template data as on the current image data.

[0049] It should be noted that the principle and process of this step are similar to those of step S102. The relevant steps can refer to step S102, which is not limited herein.

[0050] S107. Perform weighted averaging on the first similarity and the corresponding second similarity to obtain a number of average similarities. The average similarity combines the results of the two comparisons to obtain a more comprehensive similarity evaluation.

[0051] It should be noted that considering that the reconstructed image may have certain distortion or information loss and does not rely solely on the second similarity, the first similarity and the corresponding second similarity are weighted averaged to obtain a number of average similarities. For different types of samples, the reconstruction process helps to magnify the differential features between them, making the average similarity further decrease. For the same type of samples, even if the reconstruction process may introduce minor changes, the average similarity can still be maintained at a relatively high level.

[0052] S108. Determine the second template data corresponding to the maximum average similarity. Among them, the maximum average similarity refers to the template data with the highest numerical value among all the calculated average similarities.

[0053] S109. Determine the type of the current chip material corresponding to the second template data.

[0054] In some embodiments, access a pre-established template database, search for relevant information of the second template data, and then extract the type information of the chip material associated with the template data to complete the classification of the current chip material.

[0055] It can be seen that by calculating the similarity between the current image data and multiple template data, multiple first similarities are obtained. When only one first similarity exceeds the preset threshold, the material type is directly determined, improving the processing efficiency. When multiple first similarities exceed the threshold, feature extraction and reconstruction steps are introduced, so that the template data with multiple first similarities exceeding the threshold show more obvious differences after reconstruction. Finally, the material type is determined based on the maximum average similarity to improve the classification accuracy of the chip material.

[0056] However, in the actual use process, due to the excessive similarity of the chip materials, the local inconsistency between the current image data and the template data may be masked by the global reconstruction process. Even if there are local differences, the overall reconstruction error may still be small, resulting in insignificant differences in the average similarity. In this case, it is difficult to accurately identify and select the optimal reconstruction data, leading to low classification accuracy.

[0057] Please refer to Figure 2 , Figure 2 which is another flow schematic diagram of the sorting method of the chip material collector in the embodiments of the present application; In some embodiments, after step S108, it further includes: S201. Determine whether the difference between the maximum average similarity and the second largest average similarity is greater than a preset difference threshold; The second largest average similarity refers to the similarity value second only to the maximum average similarity. The preset difference threshold is used to represent the standard value for judging whether the similarity difference is significant.

[0058] It should be noted that it is used to distinguish whether there is sufficient distinguishability between the template data corresponding to the maximum average similarity and the template data corresponding to the second largest average similarity.

[0059] If the difference is greater than the preset difference threshold, execute step S109; If the difference is not greater than the preset difference threshold, S202. Determine the third template data corresponding to the second largest average similarity; It should be noted that when it is found that the difference does not exceed the preset threshold, it means that the difference between the two most similar template data is not obvious enough to directly make a reliable classification decision. In this case, the template data corresponding to the second largest average similarity will be identified and marked as the third template data.

[0060] S203. Crop the same regions of the current image data, the second template data, and the third template data, where the same regions are regions with the same position and the same features. In some embodiments, it is necessary to identify regions with the same position and similar features in the current image data, the second template data, and the third template data. This may involve techniques such as image registration and feature matching. Once the same regions are determined, these regions will be cropped from the three image data. The purpose of cropping is to focus the comparison on the most representative and discriminative parts and reduce the interference that may be brought by other regions.

[0061] In some specific embodiments, corresponding feature points in the three images are found through a feature matching algorithm; based on the matched feature points, a common region of interest is determined, and the region of interest is jointly cropped from the current image data, the second template data, and the third template data, which is not limited herein.

[0062] S204. Respectively perform feature extraction and reconstruction on the cropped current image data, the second template data, and the third template data to obtain sub-current reconstructed image data, sub-second reconstructed data, and sub-third reconstructed data. It should be noted that the principle and process of this step are similar to those of step S106, and the relevant steps can refer to step S106, which is not limited herein.

[0063] S205. Respectively calculate the similarities between the sub-current reconstructed image data and the sub-second reconstructed data and the sub-third reconstructed data to obtain the third similarity. It should be noted that the principle and process of this step are similar to those of step S102, and the relevant steps can refer to step S102, which is not limited herein.

[0064] S206. Determine the type of the current sheet material as the sheet material type corresponding to the template data with the largest third similarity.

[0065] It should be noted that the principle and process of this step are similar to those of step S104, and the relevant steps can refer to step S104, which is not limited herein.

[0066] It can be seen that when the difference between the maximum and the second-largest average similarity exceeds a preset threshold, the material type is directly determined, ensuring the processing efficiency. When the difference is insufficient, further analysis is carried out. By cropping the same regions of the current image data and the relevant template data, the same features (common parts) are removed, focusing on the parts with the most distinctive features (differentiating points). Feature extraction and reconstruction are performed on the cropped data to obtain more refined reconstructed data, and the similarity between the current reconstructed image data and the sub-template reconstructed data is calculated to obtain the third similarity. Finally, the material type is determined based on the maximum third similarity. Through the fine comparison of local features, the problem that global reconstruction may mask local differences is effectively solved.

[0067] In the actual usage process, in practical applications, a large amount of template data may exhibit highly similar average similarities. Even after subsequent cropping processing to increase the differences, the similarity difference may still remain within a small range. This clustering phenomenon of similarities makes it difficult to accurately identify and select the most suitable reconstructed data among numerous candidates, which may affect the accuracy of classification.

[0068] Please refer to Figure 3 , Figure 3 which is another flowchart of the sorting method of the chip feeder in the embodiments of the present application; In some embodiments, after step S108, it further includes: S301. Determine whether the differences between the maximum average similarity and other average similarities are all greater than a preset difference threshold; It should be noted that the principle and process of this step are similar to those of step S201. The relevant steps can refer to step S201 and are not limited here.

[0069] It should be noted that the triggering conditions for this step are one of the following situations: the maximum average similarity is relatively close to multiple other average similarities, or the difference between the maximum average similarity and the second-largest average similarity is very small. The specific threshold and the number of similarities considered can be flexibly set according to the actual application and are not limited here.

[0070] If all the differences are greater than the preset difference threshold, then execute step S109; If there is at least one difference not greater than the preset difference threshold, S302. Determine the fourth template data corresponding to all the average similarities whose differences from the maximum average similarity are not greater than the preset difference threshold; The fourth template data represents the candidate templates that need to be further compared after preliminary screening.

[0071] In some embodiments, all the average similarities whose differences from the maximum average similarity do not exceed the preset threshold are found, and the template data corresponding to these similarities are determined.

[0072] S303. Crop the same regions of the current image data and the fourth template data according to a preset cropping rule, where the same regions are regions with the same position and the same features; It should be noted that the principle and process of this step are similar to those of step S203. Relevant steps can refer to step S203 and are not limited here.

[0073] In some specific embodiments, step S303 specifically includes: S601. Determine the regions with the same position and the same features between the current image data and each template data to obtain corresponding same regions; among them, the same regions determined for different template data are independent of each other; Independent of each other means that there is no association between the same regions obtained for different template data.

[0074] S602. Crop the same regions of the current image data and each template data.

[0075] In some embodiments, precise cropping operations will be performed on the current image and each corresponding template image according to the position and size information of the previously determined same regions.

[0076] It can be seen that first, the regions with the same position and the same features between the current image data and each template data are determined. This independent region determination method allows specific similar features to be captured for different template data, can more accurately identify the local similarity between the current image and each template, is used to expand the cropping region, and facilitates subsequent multiple croppings.

[0077] S304. Respectively perform feature extraction and reconstruction on the cropped current image data and the fourth template data to obtain sub-current reconstructed image data and sub-fourth reconstructed data; It should be noted that the principle and process of this step are similar to those of step S105. Relevant steps can refer to step S105 and are not limited here.

[0078] S305. Respectively calculate the similarity between the sub-current reconstructed image data and the sub-fourth reconstructed data to obtain a fourth similarity; It should be noted that the principle and process of this step are similar to those of step S102. Relevant steps can refer to step S102 and are not limited here.

[0079] S306. Determine whether the difference between the largest fourth similarity and other fourth similarities is greater than a preset difference threshold; If all differences are greater than the preset difference threshold, S307. Determine the fifth template data corresponding to the largest fourth similarity; The fifth template data represents the best matching template determined after further comparison.

[0080] In some embodiments, if the difference between the maximum fourth similarity and all other fourth similarities is greater than a preset threshold, this matching result is considered to have a sufficiently high confidence level. Therefore, the template data corresponding to the maximum fourth similarity is determined as the fifth template data, i.e., the final matching template.

[0081] S308. Determine the type of the current sheet material corresponding to the fifth template data; It should be noted that the principle and process of this step are similar to those of step S104. Relevant steps can refer to step S104 and are not limited here.

[0082] If there is at least one difference not greater than the preset difference threshold, the fourth similarity is determined as the average similarity, and step S302 is executed.

[0083] Among them, the fourth similarity represents the degree of matching between the reconstructed images, and the fourth template data represents the candidate templates that need to be further compared.

[0084] It should be noted that it is used to handle the situation where a clear result cannot be obtained in a more refined comparison. Specifically, if there is at least one difference between a fourth similarity and the maximum fourth similarity that is not greater than the preset threshold, all fourth similarities are redefined as the average similarity, and then the previous screening process is repeated. The purpose of this step is to try to obtain a more clear result by expanding the comparison range or changing the comparison method when a clear decision cannot be made.

[0085] It can be seen that when judging the difference between the maximum average similarity and all other similarities. If all differences are greater than the preset threshold, the material type is directly determined, ensuring the processing efficiency. When there are similar similarities, all similar template data are determined as the fourth template data. The current image data and the fourth template data are cropped according to preset rules, focusing on the key feature areas. Feature extraction and reconstruction are performed on the cropped data to obtain more refined sub-reconstructed data. The fourth similarity between the sub-reconstructed data is calculated, and a new round of difference judgment is carried out. If there is an obvious maximum similarity at this time, the material type is determined; otherwise, the fourth similarity is used as the new average similarity, and the whole process is repeated. This progressive analysis method effectively solves the problem of high similarity among multiple template data. Through multiple focusings and comparisons, the best match can be gradually screened out among the highly similar candidates, improving the classification accuracy.

[0086] Please refer to Figure 4 , Figure 4 which is another flow schematic diagram of the sorting method of the sheet material collecting machine in the embodiments of the present application; The preset cropping rule is as follows: S401. Determine all continuous regions in the current image data and the fourth template data that have the same position and the same features; Among them, the continuous region is a set of pixels that are adjacent in the image and have similar features.

[0087] In some embodiments, the current image and the candidate template image are analyzed simultaneously to find regions that correspond in position and have similar visual features in both images. These regions are usually the key parts that can reflect the material features, such as marks, special shapes, or important structures, etc. By finding these regions, it can ensure that the subsequent cropping and comparison focus on the most meaningful parts of the image.

[0088] In some specific embodiments, image segmentation and matching methods can be used. The current image and the template image are respectively segmented into multiple sub-regions. Calculate the feature similarity between these sub-regions, such as using color histograms or texture features. According to the position correspondence relationship and the feature similarity threshold, determine the continuous regions that meet the conditions, which are not limited here.

[0089] S402. Select the region with the largest area from all continuous regions as the initial cropping region; It should be noted that the region for selection is a relatively large region, which is convenient for subsequent cropping.

[0090] S403. Perform the first cropping on the initial cropping region, and the cropping ratio is the first preset ratio; In some embodiments, the first reduction operation is performed on the initial cropping region, and the degree of cropping is determined by the first preset ratio. For example, if the first preset ratio is 0.8, then the area of the cropped region will be 80% of the original region. The purpose of this step is to gradually reduce the comparison region so as to more accurately focus on the most critical feature parts in the subsequent steps.

[0091] In some specific embodiments, the central cropping method can be used. By calculating the center point coordinates of the initial cropping region, calculate the new width and height according to the first preset ratio, and use the new width and height to crop a new region centered on the center point, which is not limited here.

[0092] In some other specific embodiments, the cropping method that maintains the aspect ratio can be used. Calculate the aspect ratio of the initial cropping region, calculate the new width and height according to the first preset ratio and the original aspect ratio, and start from the upper left corner of the region, and use the new width and height to crop a new region, which is not limited here.

[0093] S404. Before each subsequent cropping, determine whether the initial cropping region is the region with the largest selected area; It should be noted that the principle and process of this step are similar to those of step S402. For related steps, reference can be made to step S402, and no specific limitations are imposed here.

[0094] If it is the region with the largest area, in S405, subsequent cropping is performed on the initial cropping region, and the cropping ratio each time is the first preset ratio.

[0095] It can be seen that by determining all continuous regions in the current image data and the fourth template data that have the same position and the same features, maintaining the spatial structure information, and avoiding the appearance of additional distinct features after subsequent cropping, and selecting the region with the largest area from these regions as the initial cropping region, it is ensured that the selected region is a relatively large region, facilitating subsequent cropping.

[0096] If it is not the region with the largest area, in S406, the region with the largest area is selected from all continuous regions as the subsequent cropping region; In S407, subsequent cropping is performed on the subsequent cropping region, and the cropping ratio each time is the first preset ratio.

[0097] It can be seen that before each execution of subsequent cropping, it is judged whether the current cropping region is still the region with the largest area, and cropping continues on it, maintaining the continuity of cropping. If it is no longer the largest region, the region with the largest area is reselected as the new cropping object, maximizing the efficiency of removing the same features and maximizing the effect of focusing on the most distinct feature part. Next, an exemplary receiving machine 500 provided by the embodiments of the present application is introduced. Figure 5 It is an exemplary hardware structure diagram of the receiving machine 500 provided by the embodiments of the present application.

[0098] In some embodiments, the receiving machine 500 is a computer device or the receiving machine 500 includes a computer device. The computer device includes a processor, a memory, and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data. The network interface of the computer device is used to communicate with other external terminals or servers through a network connection. In some embodiments, the network interface can be a wired network interface, and in some embodiments, the network interface can also be a wireless network interface. The computer program, when executed by the processor, implements the method in the embodiments of the present application.

[0099] Those skilled in the art can understand, Figure 5The structure shown is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have a different component layout.

[0100] As described above, the above embodiments are only used to illustrate the technical solutions of this application, rather than to limit them; although this application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0101] As used in the above embodiments, depending on the context, the term "when..." can be interpreted to mean "if...", or "after...", or "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if detecting (the stated condition or event)" can be interpreted to mean "if determining...", or "in response to determining...", or "when detecting (the stated condition or event)", or "in response to detecting (the stated condition or event)".

[0102] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of this application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (such as coaxial cable, optical fiber, digital subscriber line) or wirelessly (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer, or a data storage device such as a server or data center that includes one or more integrated available media. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid-state drive), etc.

[0103] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by relevant hardware instructed by a computer program. This program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above method embodiments. The foregoing storage medium includes: various media that can store program codes such as ROM or random access memory RAM, magnetic disks, or optical discs.

Claims

1. A sorting method for a sheet material collecting machine, characterized in that, Including: Obtain the current image data of the current chip material; Calculate the similarity between the current image data and a number of stored template data respectively to obtain a number of first similarities, where the template data is the standard image data of a kind of chip material; When only one of the first similarities is greater than a preset similarity threshold, determine the first template data corresponding to the first similarity greater than the preset similarity threshold; Determine the current chip material as the chip material type corresponding to the first template data; When no less than two of the first similarities are greater than the preset similarity threshold, perform feature extraction and reconstruction on the current image data to obtain current reconstructed image data; Calculate the similarity between the current reconstructed image data and the reconstructed data corresponding to the stored a number of template data to obtain a number of second similarities, where the reconstructed data is obtained by performing feature extraction and reconstruction on the template data; Perform weighted averaging on the first similarity and the corresponding second similarity to obtain a number of average similarities; Determine the second template data corresponding to the largest average similarity; Determine the current chip material as the chip material type corresponding to the second template data.

2. The method according to claim 1, characterized in that, After the step of determining the second template data corresponding to the largest average similarity, the method further includes: Judge whether the difference between the largest average similarity and the second largest average similarity is greater than a preset difference threshold; If the difference is greater than the preset difference threshold, execute the step of determining the current chip material as the chip material type corresponding to the second template data; If the difference is not greater than the preset difference threshold, determine the third template data corresponding to the second largest average similarity; Crop the same regions of the current image data, the second template data, and the third template data, where the same regions are regions with the same position and the same features; Perform feature extraction and reconstruction on the cropped current image data, the second template data, and the third template data respectively to obtain sub-current reconstructed image data, sub-second reconstructed data, and sub-third reconstructed data; Calculate the similarity between the sub-current reconstructed image data and the sub-second reconstructed data and the sub-third reconstructed data respectively to obtain a third similarity; Determine the current chip material as the chip material type corresponding to the template data with the largest third similarity.

3. The method according to claim 1, wherein After the step of determining the second template data corresponding to the largest average similarity, the method further includes: Judge whether the differences between the largest average similarity and other average similarities are all greater than a preset difference threshold; If all the differences are greater than the preset difference threshold, execute the step of determining the current chip material as the chip material type corresponding to the second template data; If there is at least one difference not greater than the preset difference threshold, determine the fourth template data corresponding to all average similarities whose differences from the largest average similarity are not greater than the preset difference threshold; Crop the same regions of the current image data and the fourth template data according to a preset cropping rule, where the same regions are regions with the same position and the same features; Extract features and perform reconstruction on the cropped current image data and the fourth template data respectively to obtain sub-current reconstructed image data and sub-fourth reconstructed data; Calculate the similarity between the sub-current reconstructed image data and the sub-fourth reconstructed data respectively to obtain a fourth similarity; Determine whether the difference between the largest fourth similarity and the other fourth similarities is greater than the preset difference threshold; If all the differences are greater than the preset difference threshold, determine the fifth template data corresponding to the largest fourth similarity; Determine the type of the current sheet material as the sheet material type corresponding to the fifth template data; If there is at least one difference not greater than the preset difference threshold, determine the fourth similarity as the average similarity, and perform the step of determining all the fourth template data corresponding to the average similarities whose differences from the largest average similarity are not greater than the preset difference threshold; 4. The method according to claim 3, wherein The preset cropping rule is; Determine all the continuous regions with the same position and the same features in the current image data and the fourth template data; Select the region with the largest area from all the continuous regions as the initial cropping region; Perform the first cropping on the initial cropping region, and the cropping ratio is the first preset ratio; Perform subsequent cropping on the initial cropping region, and the cropping ratio for each time is the first preset ratio.

5. The method according to claim 4, characterized in that, Before the step of performing subsequent cropping on the initial cropping region, and the cropping ratio for each time is the first preset ratio, the method further includes: Before each subsequent cropping, determine whether the initial cropping region is the region with the largest selected area; If it is the region with the largest area, perform the step of performing subsequent cropping on the initial cropping region, and the cropping ratio for each time is the first preset ratio; If it is not the region with the largest area, perform the step of performing subsequent cropping on the initial cropping region, and the cropping ratio for each time is the first preset ratio; Then select the region with the largest area from all the continuous regions as the subsequent cropping region; Perform subsequent cropping on the subsequent cropping region, and the cropping ratio for each time is the first preset ratio.

6. The method according to claim 3, characterized in that, The step of cropping the same regions of the current image data and the fourth template data according to a preset cropping rule specifically includes; Determine the regions with the same position and the same features in the current image data and each template data to obtain corresponding same regions; wherein, the same regions determined for different template data are independent of each other; Crop the same regions of the current image data and each template data.

7. The method according to claim 1, wherein The step of extracting features and performing reconstruction on the current image data to obtain current reconstructed image data specifically includes: Use an encoder including multiple convolutional layers and pooling layers to extract features from the current image data to generate a compressed feature representation; Using a decoder including an upsampling layer, reconstruct an image using the compressed feature representation to generate the current reconstructed image data, wherein the decoder is configured to expand the compressed feature representation back to the dimensions of the original image.

8. A sheet material collecting machine, characterized in that, The sheet feeder includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to cause the sheet feeder to execute the method according to any one of claims 1-7.

9. A computer program product comprising instructions, characterized in that, When the computer program product runs on a sheet feeder, it causes the sheet feeder to execute the method according to any one of claims 1-7.

10. A computer-readable storage medium, comprising instructions, characterized in that, When the instruction runs on a sheet feeder, it causes the sheet feeder to execute the method according to any one of claims 1-7.

Images