Conveyor belt article material identification method, system and device
The images of conveyor belt items are obtained and screened through microscope array and incremental distance sequence methods, and the machine learning model is used for identification, which solves the problems of high equipment costs and difficult to focus in real time in the prior art, and achieves high accuracy and low cost item material recognition.
Patent Information
- Application Number
- CN202411887534.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-20
- Publication Date
- 2025-05-06
AI Technical Summary
In the existing conveyor belt material recognition technology, the equipment is expensive, the deployment is complex, and the low-cost small electron microscope is difficult to focus on moving items in real time, affecting application performance.
The colored object image sequence was obtained through the microscope array, and the screening process was performed in combination with the incremental distance sequence method to obtain representative images. Then, a machine learning model is used to identify representative images to realize the identification of item materials.
It improves the accuracy and real-time identification of conveyor belt item material, reduces equipment costs and deployment complexity, and enhances the stability and reliability of the system.
Smart Images
Figure CN119942532A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of object material identification, and in particular to a method, system and device for identifying the material of an object on a conveyor belt. Background Art
[0002] Conveyor belt material recognition technology effectively improves production efficiency, quality control and logistics management through automated material detection and classification. Although some existing high-precision material recognition technologies, such as X-ray machines and near-infrared spectrometers, can accurately complete material recognition tasks, they are also accompanied by problems such as high equipment costs and complex deployment, and the optical signal needs to be isolated from the external environment. In recent years, low-cost small electron microscopes have gradually become popular, but the lenses of such microscopes are difficult to focus on moving objects in real time, which affects their application performance in conveyor belt material recognition. Summary of the invention
[0003] The present application provides a method, system and device for identifying the material of items on a conveyor belt to solve one or more technical problems existing in the prior art and at least provide a beneficial option or create conditions.
[0004] On the one hand, the present application provides a method for identifying the material of an item on a conveyor belt, comprising the following steps: Acquire a color object image sequence of the object to be identified through a microscope array; the color object image sequence includes a plurality of color images, each of which is independently photographed by a microscope in the microscope array; According to the color object image sequence, a screening process is performed in combination with an increasing distance series method to obtain a representative image; According to the representative image, a machine learning model is used to perform recognition to obtain a material recognition result of the object to be recognized.
[0005] Furthermore, the number of microscopes in the microscope array is greater than or equal to a height-to-depth ratio, which is obtained by dividing the height range of the object to be identified by the depth of field distance of a single microscope and rounding up the result.
[0006] Furthermore, the method of filtering and processing the color object image sequence in combination with an increasing distance series method to obtain a representative image includes: Generate a set of non-negative integer sequences that are monotonically increasing and gradually increasing in increments as the screening sequence; Sampling the color image using the screening sequence to obtain a pixel point sequence of the color image; According to the pixel point sequence of the color image, in combination with the RGB difference calculation method, the RGB difference value of the color image is calculated; The color image with the largest RGB difference value in the color object image sequence is used as the representative image.
[0007] Furthermore, in the screening sequence, the increase between any two adjacent elements is not an integer multiple of the increase between any other two adjacent elements; the values of all elements in the screening sequence are smaller than the length value of the color image; the values of all elements in the screening sequence are smaller than the width value of the color image.
[0008] Further, the pixel point sequence includes a horizontal point sequence, a vertical point sequence and a diagonal point sequence; The step of sampling the color image using the screening sequence to obtain a pixel point sequence of the color image includes: Using the screening sequence to perform sampling along the horizontal direction of the color image to obtain the horizontal point sequence; Using the screening sequence to sample along the vertical direction of the color image to obtain the vertical point sequence; The diagonal point sequence is obtained by sampling along the diagonal direction of the color image using the screening sequence.
[0009] Further, the pixel point sequence includes a horizontal point sequence, a vertical point sequence and a diagonal point sequence; The step of calculating the RGB difference value of the color image based on the pixel point sequence of the color image and the RGB difference calculation method comprises: Calculating RGB difference values between adjacent pixel pairs of the horizontal point sequence to obtain horizontal RGB difference values; Calculating RGB difference values between adjacent pixel pairs of the vertical point sequence to obtain vertical RGB difference values; Calculating RGB difference values between adjacent pixel pairs of the diagonal point sequence to obtain diagonal RGB difference values; The maximum value among the horizontal RGB difference value, the vertical RGB difference value and the diagonal RGB difference value is taken as the RGB difference value of the color image.
[0010] Furthermore, the method of using a machine learning model to perform recognition based on the representative image to obtain a material recognition result of the object to be recognized includes: Performing cropping on the representative image to obtain a representative sub-image set; According to the representative sub-image set, a machine learning model is used to perform identification to obtain a material identification result of the object to be identified.
[0011] Further, the representative sub-graph set includes a plurality of representative sub-graphs; The method of using a machine learning model to identify the material of the object to be identified according to the representative sub-graph set to obtain the material identification result of the object to be identified includes: Using a machine learning model to identify each of the representative subgraphs, and obtaining a recognition result of each of the representative subgraphs as a recognition result set; The recognition result with the highest frequency in the recognition result set is used as the material recognition result of the object to be recognized.
[0012] On the other hand, the present application provides a conveyor belt item material recognition system, including an image acquisition module, an image screening module and an image recognition module; The image acquisition module is used to acquire a color object image sequence of the object to be identified through a microscope array; the color object image sequence includes a plurality of color images, each of which is independently photographed by a microscope in the microscope array; The image screening module is used to perform screening processing based on the color object image sequence in combination with the increasing distance series method to obtain a representative image; The image recognition module is used to perform recognition based on the representative image using a machine learning model to obtain a material recognition result of the object to be recognized.
[0013] On the other hand, the present application provides a conveyor belt item material recognition device, including a microscope array and a computing device; The microscope array is arranged above the conveyor belt; The number of microscopes in the microscope array is greater than or equal to a height-depth ratio, where the height-depth ratio is obtained by dividing a height range of an object to be identified by a depth of field distance of a single microscope and rounding the result upwards; The microscope array is used to obtain a color object image sequence of the object to be identified; the color object image sequence includes a plurality of color images, each of which is independently photographed by a microscope in the microscope array; The computing device is used to perform screening processing based on the color object image sequence in combination with the increasing distance series method to obtain a representative image; The computing device is also used to perform recognition based on the representative image using a machine learning model to obtain a material recognition result of the object to be recognized.
[0014] The beneficial effects of the present application are as follows: the present application provides a method for identifying the material of items on a conveyor belt, including obtaining a color item image sequence of an item to be identified through a microscope array; the color item image sequence includes multiple color images, each color image is independently photographed by a microscope in the microscope array; based on the color item image sequence, screening and processing are performed in combination with an increasing distance series method to obtain a representative image; based on the representative image, identification is performed using a machine learning model to obtain a material identification result of the item to be identified. The present application effectively improves the accuracy and real-time performance of conveyor belt item material identification by reasonably arranging the microscope array and using an increasing distance series method for image screening. The present application also provides a corresponding system and device, and the beneficial effects of the system and device are similar to those of the method, so they will not be described repeatedly here.
[0015] Other features and advantages of the present application will be described in the following description, and partly become apparent from the description, or understood by practicing the present application. The purpose and other advantages of the present application can be realized and obtained by the structures specifically pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] The accompanying drawings are used to provide a further understanding of the technical solution of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the technical solution of the present invention and do not constitute a limitation on the technical solution of the present invention.
[0017] Figure 1 It is a flow chart of the conveyor belt item material identification method provided by the present application; Figure 2 It is a structural diagram of the conveyor belt object material recognition system provided by the present application; Figure 3 It is a schematic diagram of deploying a microscope array above a conveyor belt provided in the present application. DETAILED DESCRIPTION
[0018] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0019] The present application is further described below in conjunction with the accompanying drawings and specific embodiments. The described embodiments should not be regarded as limiting the present application, and all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application.
[0020] In the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.
[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.
[0022] In response to the problems and defects in the related technologies, the embodiments of the present application propose a method, system and device for identifying the material of items on a conveyor belt. The embodiments of the present application effectively improve the accuracy and real-time performance of identifying the material of items on a conveyor belt through the reasonable arrangement of the microscope array and the incremental distance series method. First, the microscope array can achieve multi-view high-resolution imaging to ensure that the features of items at each height and angle can be clearly captured, avoiding the problem of inaccurate recognition caused by the depth of field limitation of a single camera, thereby greatly improving the recognition accuracy. Secondly, by screening representative images through the incremental distance series method, not only the impact of noise and outliers is reduced, the stability and reliability of the system are enhanced, but also unnecessary image processing steps are reduced, and computational efficiency is improved.
[0023] First, the implementation steps of the conveyor belt object material identification method provided by the embodiment of the present application will be described in detail with reference to the accompanying drawings.
[0024] The conveyor belt item material identification method proposed in the embodiment of the present application can be applied to a terminal, a server, or software running in a terminal or a server. The terminal can be a tablet computer, a laptop computer, a desktop computer, etc., but is not limited thereto. The server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content distribution networks, and big data and artificial intelligence platforms.
[0025] Reference Figure 1 The implementation process of the conveyor belt item material identification method provided in the embodiment of the present application includes but is not limited to the following steps.
[0026] Step 101: Acquire a color object image sequence of an object to be identified through a microscope array.
[0027] It should be noted that the color object image sequence includes multiple color images, each of which is independently captured by a microscope in the microscope array. The arrangement of the microscope array effectively improves the accuracy and reliability of conveyor belt item material recognition, and has important application value in the food industry, logistics management, and waste sorting. By capturing the microstructure and texture features of the material surface, the microscope array can distinguish between materials with similar appearance but different materials, providing high-resolution multi-view imaging, ensuring that comprehensive image data can be obtained even if the object has a complex shape.
[0028] In step 101, each microscope in the microscope array independently photographs the object to be identified, ensuring that detailed image information is captured from multiple heights and angles. This avoids the problem of partial blurring of a single camera due to depth of field limitations, and provides more comprehensive and accurate object features. Each microscope is able to provide high-resolution color images, retaining rich color and texture information, and providing a solid foundation for subsequent image processing and classification. The microscope array is reasonably arranged according to the height range of the object, ensuring that even if the height of the object is not fixed, at least one microscope can provide a clear image, thereby adapting to objects of different sizes and shapes.
[0029] Step 102, based on the color object image sequence, a screening process is performed in combination with an increasing distance series method to obtain a representative image.
[0030] It should be noted that the increasing distance series method generates an increasing distance series, performs multi-directional sampling on color images, and selects representative images from the color object image sequence. The increasing distance series method is a non-periodic sampling technique. Instead of selecting sample points at fixed intervals, the sampling positions are dynamically adjusted according to specific increasing rules to ensure that the distance between adjacent samples is not constant. This non-uniformly distributed sampling method can cover a wider range, reduce the similarity of adjacent samples, improve the representativeness of sampling, and effectively capture changes in different areas of the image, especially where the changes are effective. The increasing distance series method is highly adaptable and can be flexibly adjusted according to specific application scenarios. It is suitable for complex, non-repetitive pattern tasks such as image recognition and natural language processing. It not only reduces redundant information and avoids the overfitting problem that may be caused by periodic sampling, but also enhances the stability and reliability of the system. It is an important tool in modern image processing and data analysis.
[0031] In step 102, the image sequence is screened by using the increasing distance series method to obtain a representative image as the basis for subsequent recognition, which reduces unnecessary image processing steps, reduces the demand for computing resources, and improves system efficiency. The screened representative image is optimized and can effectively reflect the key features of the object, reduce the impact of noise and outliers, and enhance the stability of the system and the reliability of the recognition results.
[0032] Step 103, based on the representative image, use the machine learning model to perform recognition to obtain the material recognition result of the object to be recognized.
[0033] It should be noted that the machine learning model improves the accuracy and robustness of conveyor belt material recognition by automatically learning features from a large amount of labeled data, and can capture subtle differences in images. Its good generalization ability enables it to adapt to different types of objects and complex background environments, and is suitable for many fields such as food industry, logistics, and garbage classification. The machine learning model not only realizes full process automation, reduces manual intervention, and improves production efficiency, but also is good at processing high-dimensional data and dealing with the diversity problems caused by a wide variety of objects, reducing development and maintenance costs, and has real-time and scalability.
[0034] In step 103, a machine learning model is used to automatically extract and classify features in the representative image to achieve fast and accurate object material recognition and obtain a material recognition result of the object to be identified.
[0035] In some embodiments of the present application, the number of microscopes in the microscope array is greater than or equal to the height-to-depth ratio, which is obtained by dividing the height range of the object to be identified by the depth of field distance of a single microscope and rounding the result upward.
[0036] It should be noted that the depth of field of a microscope refers to the distance range from the nearest to the farthest distance that can maintain a clear image in one imaging. High magnification and large numerical aperture usually result in a shallow depth of field, while low magnification and small numerical aperture provide a deeper depth of field. In addition, the wavelength of light and the refractive index of the medium will also affect the depth of field. Short wavelength light and high refractive index media will make the depth of field shallower. Therefore, the reasonable arrangement of the microscope array is important to ensure clear imaging of different height positions of the sample.
[0037] The number of microscopes in the microscope array is calculated based on the height range of the object to be identified and the depth of field of a single microscope, ensuring that every possible height level of the object within the entire height range can be clearly imaged by at least one microscope. This not only fully covers the height changes of the object, improves the recognition accuracy and reliability, but also reduces misjudgments through multi-view capture. This design optimizes resource allocation, enhances the robustness and real-time response capabilities of the system, and provides a solid guarantee for the efficient and accurate recognition of conveyor belt items.
[0038] In some embodiments of the present application, in step 102, the implementation process of obtaining a representative image by screening the color object image sequence in combination with the increasing distance series method includes but is not limited to the following steps.
[0039] Step 201, generating a set of non-negative integer sequences that are monotonically increasing and have gradually increasing increments as screening sequences.
[0040] In step 201, a set of non-negative integer sequences that are monotonically increasing and gradually increasing in amplitude are generated as screening sequences. The values in the screening sequence increase with increasing positions, and the amplitude of each increase gradually increases. Such a sequence is used in the subsequent sampling process to ensure that when the image is non-uniformly sampled in the subsequent steps, a wider range can be covered, while reducing the similarity between adjacent samples, thereby improving the representativeness of the sampling.
[0041] Step 202: Sample the color image using the screening sequence to obtain a pixel point sequence of the color image.
[0042] In step 202, the color image is sampled using the screening sequence to obtain a pixel sequence of the color image. In this way, the amount of data can be reduced while maintaining the image features, making the calculation more efficient and capturing the characteristics of different areas in the image, especially those places where the changes are effective.
[0043] Step 203: Calculate the RGB difference value of the color image according to the pixel sequence of the color image in combination with the RGB difference calculation method.
[0044] It should be noted that the RGB difference value of a color image is a quantitative indicator used to measure the color similarity or difference between two color images. It calculates the color distance between two pixels based on the values of the three color channels: red (R), green (G), and blue (B).
[0045] In step 203, the RGB difference value of the color image is calculated based on the pixel sequence and the RGB difference calculation method. The RGB difference value quantifies the color difference between images and can reflect the richness and diversity of the image colors. This step is very important for identifying and comparing color changes between images because it helps to determine which images are the most representative, that is, contain the most color information or changes.
[0046] Step 204: taking the color image with the largest RGB difference value in the color object image sequence as the representative image.
[0047] In step 204, the color image with the largest RGB difference value in the color object image sequence is used as the representative image. The image with the largest RGB difference value usually contains the most information or the richest color changes, so it can well summarize the characteristics of the entire image sequence, ensuring that the selected image can effectively display the core content or change trend of the original sequence.
[0048] In some embodiments of the present application, the increment between any two adjacent elements in the screening sequence is not an integer multiple of the increment between the other two adjacent elements. This design ensures the non-uniformity and diversity of the distribution of sampling points in the color image, avoids the appearance of periodic repetitive patterns, and thus improves the representativeness of the sampling and the comprehensiveness of the image feature capture. In this way, the system can more effectively cover the key areas in the image, reduce redundant information, and improve the accuracy and efficiency of subsequent processing steps.
[0049] In some embodiments of the present application, the values of all elements in the screening sequence are less than the length value of the color image and the values of all elements in the screening sequence are less than the width value of the color image. This design ensures that the sampling index of the color image is always within the valid range, avoiding the problem of exceeding the image boundary, thereby ensuring the stability and reliability of the sampling process. In this way, the system can efficiently sample and process the image without losing important information, thereby improving the accuracy and efficiency of the overall recognition.
[0050] In some embodiments of the present application, the pixel point sequence includes a horizontal point sequence, a vertical point sequence and a diagonal point sequence. In step 202, the process of sampling the color image using the screening sequence to obtain the pixel point sequence of the color image includes but is not limited to the following steps.
[0051] Step 301: Use the screening sequence to sample along the horizontal direction of the color image to obtain a horizontal point sequence.
[0052] In step 301, the screening sequence is used to sample along the horizontal direction of the color image to obtain a horizontal point sequence. This process ensures that the key information of each row of the image can be effectively extracted, provides a detailed description of the horizontal features of the image, helps capture color changes and texture information in the horizontal direction, and enhances the accuracy of image recognition.
[0053] Step 302: Use the screening sequence to sample along the vertical direction of the color image to obtain a vertical point sequence.
[0054] In step 302, the screening sequence is used to sample along the vertical direction of the color image to obtain a vertical point sequence. Through vertical sampling, the system can extract the key features of each column of the image, provide a detailed description of the longitudinal features, help capture the color changes and structural information in the vertical direction, and further enrich the feature representation of the image.
[0055] Step 303: Use the screening sequence to sample along the diagonal direction of the color image to obtain a diagonal point sequence.
[0056] In step 303, the screening sequence is used to sample along the diagonal direction of the color image to obtain a diagonal point sequence. Sampling in the diagonal direction supplements the deficiencies of sampling in the horizontal and vertical directions, can capture the oblique features and changes in the image, improve the comprehensiveness and diversity of image feature capture, and ensure the effective recognition of complex patterns and textures.
[0057] Through these multi-directional sampling steps, the embodiment of the present application can comprehensively capture the features of the color image from different angles, thereby improving the accuracy and reliability of subsequent processing and providing a solid foundation for efficient and accurate image recognition.
[0058] In some embodiments of the present application, in step 203, the implementation process of calculating the RGB difference value of the color image according to the pixel point sequence of the color image in combination with the RGB difference calculation method includes but is not limited to the following steps.
[0059] Step 401, calculating RGB difference values between adjacent pixel pairs in a horizontal point sequence to obtain horizontal RGB difference values.
[0060] In step 401, the RGB difference values between adjacent pixel pairs of the horizontal point sequence are calculated to obtain the horizontal RGB difference values. This process quantifies the color changes of the color image in the horizontal direction, captures the horizontal color gradient and texture features of the image, ensures the effective extraction of the horizontal features of the image, and provides an important basis for subsequent analysis.
[0061] Step 402, calculating RGB difference values between adjacent pixel pairs in a vertical point sequence to obtain vertical RGB difference values.
[0062] In step 402, the RGB difference values between adjacent pixel pairs of the vertical point sequence are calculated to obtain the vertical RGB difference values. By evaluating the color change of the color image in the vertical direction, this step can capture the color gradient and structural information of the image in the vertical direction, supplement the deficiency of horizontal sampling, and further enrich the description of image features.
[0063] Step 403, calculating the RGB difference values between adjacent pixel pairs in the diagonal point sequence to obtain the diagonal RGB difference values.
[0064] In step 403, the RGB difference values between adjacent pixel pairs of the diagonal point sequence are calculated to obtain the diagonal RGB difference values. The difference values in the diagonal direction capture the color changes and complex patterns of the color image in the diagonal direction, enhance the comprehensiveness and diversity of the image feature representation, and ensure the effective fusion of multi-directional features.
[0065] Step 404: taking the maximum value among the horizontal RGB difference value, the vertical RGB difference value and the diagonal RGB difference value as the RGB difference value of the color image.
[0066] In step 404, the maximum value among the horizontal RGB difference value, the vertical RGB difference value and the diagonal RGB difference value is used as the RGB difference value of the color image, so as to highlight the most effective color change in the color image.
[0067] Through the above steps, the embodiment of the present application can comprehensively evaluate the color changes of color images from multiple directions, ensure the effective capture of key features, and provide a solid foundation for efficient and accurate image recognition.
[0068] In some embodiments of the present application, in step 103, the process of obtaining the material recognition result of the object to be identified by using a machine learning model for identification based on the representative image includes but is not limited to the following steps.
[0069] Step 501: crop the representative image to obtain a representative sub-image set.
[0070] In step 501, the representative image is cropped to obtain a representative sub-image set. This process ensures that each sub-image can focus on different key areas of the image by dividing the representative image into multiple sub-images, thereby improving the precision and accuracy of feature extraction and providing richer local information for subsequent recognition.
[0071] Step 502: Based on the representative sub-graph set, a machine learning model is used to perform identification to obtain a material identification result of the object to be identified.
[0072] In step 502, the machine learning model is used to identify the material of the object to be identified based on the representative sub-image set. By analyzing these sub-images through the machine learning model, the system can capture more subtle feature differences, thereby improving the accuracy and robustness of recognition. The machine learning model can not only automatically process large amounts of data, but also continuously optimize the recognition logic, adapt to different types of objects and complex background environments, and ensure the reliability and efficiency of the final recognition results.
[0073] In some embodiments of the present application, the representative sub-image set includes multiple representative sub-images. In step 502, the process of obtaining the material recognition result of the object to be recognized by using the machine learning model for recognition according to the representative sub-image set includes but is not limited to the following steps.
[0074] Step 601, using a machine learning model to identify each representative sub-graph, and obtaining a recognition result of each representative sub-graph as a recognition result set.
[0075] In step 601, each representative sub-image is identified using a machine learning model, and the identification results of each representative sub-image are obtained as a set of identification results. This process ensures the effective capture of local details by analyzing the features in each sub-image, and improves the accuracy and robustness of recognition. The machine learning model can automatically extract and classify subtle differences in the image, providing basic data for subsequent comprehensive evaluation.
[0076] Step 602: The recognition result with the highest frequency in the recognition result set is used as the material recognition result of the object to be recognized.
[0077] In step 602, the recognition result with the highest frequency in the recognition result set is used as the final recognition result of the material of the object to be recognized. By counting the frequency of the recognition results of each sub-graph and selecting the recognition result with the highest number of occurrences, the influence of misjudgment and outliers can be effectively reduced, ensuring the reliability and consistency of the final result. This method not only improves the accuracy of overall recognition, but also enhances the anti-noise ability and stability of the system.
[0078] Through the above two steps, the embodiment of the present application realizes a complete process from local feature recognition to global result summary, ensuring high-precision and high-reliability recognition of the material of the object to be identified, is suitable for a variety of complex application scenarios, and effectively improves the performance and efficiency of the conveyor belt item material recognition system.
[0079] Secondly, refer to Figure 2 , an embodiment of the present application provides a conveyor belt item material recognition system, including an image acquisition module 701, an image screening module 702 and an image recognition module 703.
[0080] The image acquisition module 701 is used to acquire a color object image sequence of the object to be identified through the microscope array. The color object image sequence includes multiple color images, each of which is independently photographed by a microscope in the microscope array.
[0081] The image screening module 702 is used to perform screening processing based on the color object image sequence in combination with the increasing distance series method to obtain a representative image.
[0082] The image recognition module 703 is used to perform recognition based on the representative image using a machine learning model to obtain a material recognition result of the object to be recognized.
[0083] Furthermore, an embodiment of the present application provides a conveyor belt item material recognition device, including a microscope array and a computing device.
[0084] An array of microscopes is arranged above the conveyor belt.
[0085] The number of microscopes in the microscope array is greater than or equal to a quotient of a height range of the objects to be identified divided by a depth of field distance of the microscopes.
[0086] The microscope array is used to obtain a color object image sequence of the object to be identified. The color object image sequence includes multiple color images, each of which is independently photographed by a microscope in the microscope array.
[0087] The computing device is used to perform screening processing based on the color object image sequence in combination with the increasing distance series method to obtain a representative image.
[0088] The computing device is also used to perform recognition based on the representative image using a machine learning model to obtain a material recognition result of the object to be recognized.
[0089] The implementation of the conveyor belt object material identification device proposed in the embodiment of the present application will be described in detail below.
[0090] First, refer to Figure 3 , on the conveyor belt Microscope array deployed above Specifically, according to the conveyor belt Any item to be identified The possible height range and the depth of field distance of a single microscope that can clearly image an object not only ensure that at least one microscope can clearly observe the surface of the object at any height in the range above the conveyor belt, but also ensure that the lighting conditions on the conveyor belt surface are basically consistent.
[0091] Optionally, the number of microscopes in the microscope array satisfies the following formula (1): (1); In formula (1), represents the number of microscopes in the microscope array; Indicates the maximum height of items on the conveyor belt; Indicates the minimum height of items on the conveyor belt; Represents the depth of field distance of a single microscope; It represents the height-to-depth ratio, which is the result of rounding up the quotient of the height range of the object to be identified divided by the depth of field distance of a single microscope.
[0092] Secondly, the computing device monitors the images collected by each microscope in the microscope array in real time. If the computing device observes from the collected images that there are no objects on the conveyor belt, it will continue to observe and wait for the objects to appear. Once the computing device observes from the collected images that an object appears on the conveyor belt, it controls each microscope to take pictures independently. Each microscope takes a color image to form a color object image sequence. , the number of color images in the color object image sequence is consistent with the number of microscopes in the microscope array.
[0093] Furthermore, for the color object image sequence , the computing device generates a set of non-negative integer sequences that are discontinuous, monotonically increasing, and gradually increasing in magnitude As a filter array, the number of elements in the filter array At least 3, It can be adjusted according to the performance of the computing device. In the screening sequence, the increment between any two adjacent elements is not an integer multiple of the increment between the other two adjacent elements. The values of all elements in the screening sequence are smaller than the length and width of the color image.
[0094] Optionally, the implementation process of the generation algorithm of the screening sequence includes but is not limited to the following steps.
[0095] Step (1), let the length of the color image be and width value is , let the first element of the screening sequence be a non-negative integer , and guarantee and .
[0096] Step (2), randomly select a positive integer for The growth value of , so that , while ensuring and .
[0097] Step (3), randomly select a positive integer for Growth value, guaranteed , so that , while ensuring and .
[0098] Similarly, similar to step (3), generate , get the screening sequence .
[0099] The computing device then uses the filtered sequence The color image is sampled to obtain the pixel point sequence of the color image. The pixel point sequence includes a horizontal point sequence, a vertical point sequence and a diagonal point sequence. and For example, using Along Sampling in the horizontal direction, we get middle The corresponding horizontal point sequence ;use Along Sampling in the vertical direction, we get middle The corresponding vertical point sequence ;use Along Sampling in the diagonal direction of middle The corresponding diagonal point sequence .
[0100] Similarly, the computing device utilizes the filtering sequence For color images Sampling is performed along the horizontal, vertical and diagonal directions to obtain a color image. The horizontal point sequence, vertical point sequence and diagonal point sequence of A sequence of pixels.
[0101] Furthermore, the computing device uses the color image The pixel sequence is combined with the RGB difference calculation method to calculate the color image RGB difference value . In pixels and pixels For example, pixel and pixels The RGB difference between them satisfies the following formula (2): (2); In formula (2), Represents pixel and pixels The RGB difference between Represents pixel R value, Represents pixel R value, Represents pixel G value, Represents pixel G value, Represents pixel The B value, Represents pixel The B value.
[0102] Similarly, similar to formula (2), the computing device calculates the color image The horizontal RGB difference value sequence, vertical RGB difference value sequence and diagonal RGB difference value sequence of the color image are obtained. The horizontal RGB difference value sequence selects the largest element value as the color image The horizontal RGB difference values from the color image The vertical RGB difference value sequence is selected to obtain the largest element value as the color image The vertical RGB difference values from the color image The diagonal RGB difference value sequence selects the largest element value as the color image The diagonal RGB difference values of The maximum value among the horizontal RGB difference value, vertical RGB difference value and diagonal RGB difference value is selected as the color image RGB difference value Similar to the above steps, the computing device calculates the color object image sequence The RGB difference value of each color image in .
[0103] The computing device then extracts the color object image sequence Select representative images. Specifically, Select the largest element value and record its subscript as , which means that the color image Has the largest RGB difference value, indicating a color image With the most special optical characteristics, it is just right for the microscope Observed, so the color image It is denoted as the representative image.
[0104] Furthermore, the computing device is based on the representative image , using the machine learning model to identify and obtain the material recognition result of the object to be identified. Specifically, the computing device first uses a random position cropping method to crop the representative image Cut and get The representative subgraph set , where represents the subgraph set In the figure, the number of subgraphs is , which means the length of the subgraph is , which means the width of the sub-image is The computing device then uses the machine learning model to represent the subgraph set Recognition is performed to obtain the recognition results of each representative sub-graph. For example, representing a subgraph The recognition result Satisfies the following formula (3): (3); In formula (3), Represents a machine learning model; based on the recognition results of each representative subgraph, a set of recognition results is obtained Finally, the computing device collects the recognition results Count the frequency of occurrence of each element and select the element value with the largest number of occurrences as the item to be identified The final recognition result of the material.
[0105] Optionally, the machine learning model includes a ResNet50 network model, and the training process of the ResNet50 network model includes but is not limited to the following steps.
[0106] First, define the set of item material categories that need to be identified.
[0107] Secondly, put a variety of different types of test items on the conveyor belt. Taking item C as an example, the computing device controls the microscope array to shoot item C to obtain a color item image sequence; the computing device generates a screening sequence to sample the color item image sequence, and combines the RGB difference calculation method to screen the representative image from the color item image sequence; the computing device uses the center cropping method and the random position cropping method to crop the representative image to obtain a representative sub-image set of the representative image.
[0108] Furthermore, the ResNet50 network model is used to identify the representative sub-graph set to obtain the material recognition result of object C.
[0109] For different types of test items, similar to the above steps, their material recognition results are obtained. Their material recognition results are compared with their actual material categories, and the recognition accuracy of the ResNet50 network model is calculated.
[0110] Finally, the loss function is used to optimize the ResNet50 network model. When the ResNet50 network model achieves satisfactory performance on the test items, it is deployed to the actual conveyor belt item material recognition system to start automatically classifying and identifying items being transmitted in real time.
[0111] In summary, the embodiments of the present application provide the following technical effects.
[0112] The embodiment of the present application achieves multi-angle, high-definition object shooting and obtains a sequence of color object images by arranging a microscope array above the conveyor belt. The microscope array ensures that the features of the object at each height and angle can be clearly captured, avoiding the problem of inaccurate recognition caused by the depth of field limitation of a single camera, and effectively improving the image quality and the accuracy of feature extraction. The embodiment of the present application uses the increasing distance series method to perform multi-directional sampling on the color object image sequence, and combines the RGB difference calculation method to screen and obtain representative images, thereby reducing redundant data and optimizing the computing load, thereby enhancing the processing speed and efficiency of the system. In addition, the embodiment of the present application uses a machine learning model to identify the representative image to obtain the material recognition result of the object to be identified, thereby enhancing the robustness and adaptability of the system, being able to cope with factors such as lighting changes and object position offsets, and ensuring stable performance.
[0113] In some optional embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation schematic diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the application is provided by way of example, for the purpose of providing a more comprehensive understanding of the technology. The disclosed method is not limited to the operation and logic flow presented herein. The optional embodiment is expected, wherein the order of various operations is changed and the sub-operation of a part of the larger operation is wherein described is performed independently.
[0114] In addition, although the present application is described in the context of functional modules, it should be understood that, unless otherwise specified, one or more of the functions and / or features can be integrated into a single physical device and / or software module, or one or more functions and / or features can be implemented in a separate physical device or software module. It can also be understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present application. More specifically, in view of the properties, functions and internal relationships of the various functional modules in the devices disclosed herein, the actual implementation of the module will be understood within the conventional techniques of engineers. Therefore, those skilled in the art can implement the present application as set forth in the claims using ordinary techniques. It can also be understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present application, which is determined by the full scope of the attached claims and their equivalents.
[0115] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several programs to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.
[0116] The logic and / or steps represented in the flowchart or otherwise described herein, for example, may be considered as an ordered list of executable programs for implementing the logical functions, and may be embodied in any computer-readable medium for use by a program execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch and execute a program from a program execution system, device or apparatus), or in conjunction with such program execution systems, devices or apparatuses. For purposes of this specification, a "computer-readable medium" may be any device that can contain, store, communicate, propagate or transmit a program for use by a program execution system, device or apparatus, or in conjunction with such program execution systems, devices or apparatuses.
[0117] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in a suitable manner thereof, and then stored in a computer memory.
[0118] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable program execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0119] In the above description of this specification, the description with reference to the terms "one embodiment / implementation", "another embodiment / implementation" or "certain embodiments / implementations" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiments or examples are included in the embodiments or examples of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.
[0120] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.
[0121] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of the present invention.
Claims
1. A method for identifying the material of items on a conveyor belt, characterized in that: The steps include: Acquire a color object image sequence of the object to be identified through a microscope array; the color object image sequence includes a plurality of color images, each of which is independently photographed by a microscope in the microscope array; According to the color object image sequence, a screening process is performed in combination with an increasing distance series method to obtain a representative image; According to the representative image, a machine learning model is used to perform recognition to obtain a material recognition result of the object to be recognized.
2. The method for identifying the material of items on a conveyor belt according to claim 1, characterized in that: The number of microscopes in the microscope array is greater than or equal to a height-to-depth ratio, which is obtained by dividing the height range of the object to be identified by the depth of field distance of a single microscope and rounding up the result.
3. The method for identifying the material of items on a conveyor belt according to claim 1, characterized in that: The method of screening and processing the color object image sequence in combination with the increasing distance series method to obtain a representative image includes: Generate a set of non-negative integer sequences that are monotonically increasing and gradually increasing in increments as the screening sequence; Sampling the color image using the screening sequence to obtain a pixel point sequence of the color image; According to the pixel point sequence of the color image, in combination with the RGB difference calculation method, the RGB difference value of the color image is calculated; The color image with the largest RGB difference value in the color object image sequence is used as the representative image.
4. The method for identifying the material of items on a conveyor belt according to claim 3, characterized in that: In the screening sequence, the increase between any two adjacent elements is not an integer multiple of the increase between any other two adjacent elements; the values of all elements in the screening sequence are smaller than the length value of the color image; the values of all elements in the screening sequence are smaller than the width value of the color image.
5. The method for identifying the material of items on a conveyor belt according to claim 3, characterized in that: The pixel point sequence includes a horizontal point sequence, a vertical point sequence and a diagonal point sequence; The step of sampling the color image using the screening sequence to obtain a pixel point sequence of the color image includes: Using the screening sequence to perform sampling along the horizontal direction of the color image to obtain the horizontal point sequence; Using the screening sequence to sample along the vertical direction of the color image to obtain the vertical point sequence; The diagonal point sequence is obtained by sampling along the diagonal direction of the color image using the screening sequence.
6. The method for identifying the material of items on a conveyor belt according to claim 3, characterized in that: The pixel point sequence includes a horizontal point sequence, a vertical point sequence and a diagonal point sequence; The step of calculating the RGB difference value of the color image based on the pixel point sequence of the color image and the RGB difference calculation method comprises: Calculating RGB difference values between adjacent pixel pairs of the horizontal point sequence to obtain horizontal RGB difference values; Calculating RGB difference values between adjacent pixel pairs of the vertical point sequence to obtain vertical RGB difference values; Calculating RGB difference values between adjacent pixel pairs of the diagonal point sequence to obtain diagonal RGB difference values; The maximum value among the horizontal RGB difference value, the vertical RGB difference value and the diagonal RGB difference value is taken as the RGB difference value of the color image.
7. The method for identifying the material of items on a conveyor belt according to claim 1, characterized in that: The method of using a machine learning model to identify the representative image to obtain a material identification result of the object to be identified includes: Performing cropping processing on the representative image to obtain a representative sub-image set; According to the representative sub-image set, a machine learning model is used to perform identification to obtain a material identification result of the object to be identified.
8. The method for identifying the material of items on a conveyor belt according to claim 7, characterized in that: The representative sub-graph set includes a plurality of representative sub-graphs; The method of using a machine learning model to identify the material of the object to be identified according to the representative sub-graph set to obtain the material identification result of the object to be identified includes: Using a machine learning model to identify each of the representative subgraphs, and obtaining a recognition result of each of the representative subgraphs as a recognition result set; The recognition result with the highest frequency in the recognition result set is used as the material recognition result of the object to be recognized.
9. Conveyor belt item material recognition system, characterized in that: It includes an image acquisition module, an image screening module and an image recognition module; The image acquisition module is used to acquire a color object image sequence of the object to be identified through a microscope array; the color object image sequence includes a plurality of color images, each of which is independently photographed by a microscope in the microscope array; The image screening module is used to perform screening processing based on the color object image sequence in combination with the increasing distance series method to obtain a representative image; The image recognition module is used to perform recognition based on the representative image using a machine learning model to obtain a material recognition result of the object to be recognized.
10. A conveyor belt article material identification device, characterized in that: Includes microscope arrays and computing equipment; The microscope array is arranged above the conveyor belt; The number of microscopes in the microscope array is greater than or equal to a height-depth ratio, where the height-depth ratio is obtained by dividing a height range of an object to be identified by a depth of field distance of a single microscope and rounding the result upwards; The microscope array is used to obtain a color object image sequence of the object to be identified; the color object image sequence includes a plurality of color images, each of which is independently photographed by a microscope in the microscope array; The computing device is used to perform screening processing based on the color object image sequence in combination with the increasing distance series method to obtain a representative image; The computing device is also used to perform recognition based on the representative image using a machine learning model to obtain a material recognition result of the object to be recognized.