Material information database updating method and device, electronic equipment and readable storage medium

Abstract

The application provides a material information database updating method and device, electronic equipment and readable storage medium, wherein the method comprises the following steps: after detecting that the material moves to the first conveying belt through the contact sensor, the upper surface of the first conveying belt is photographed through the camera to obtain a first image of the material currently located on the first conveying belt; a first foreground image in the first image is extracted; the classification of the current material is determined according to the first foreground image; the feature information corresponding to the first foreground image is inquired from the material information database according to the classification of the current material; the material information database is updated according to the search result of the feature information, and the current material is sorted. Through the method, the efficiency and accuracy of the overall sorting are enhanced.

Description

Technical Field

[0001] This application relates to the field of material information management technology, and in particular to a method, apparatus, electronic device, and readable storage medium for updating a material information database. Background Technology

[0002] Automated sorting machines are devices that use automation technology to quickly and accurately classify materials. They are widely used in logistics warehousing, express delivery, manufacturing, e-commerce, and other fields, significantly improving sorting efficiency, reducing labor costs, and decreasing sorting error rates. Their core function is to automatically allocate different categories of materials to corresponding areas or conveyor lines based on preset rules (such as weight, size, barcode, destination, etc.).

[0003] Automatic sorting machines can significantly improve automation and save labor costs. However, before an automatic sorting machine can operate, a material information database needs to be generated. Only when the material information database can completely and clearly record the information of each item can the automatic sorting machine perform sorting work normally. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a method, apparatus, electronic device and readable storage medium for updating a material information database. When determining the material classification, a contact sensor is used for assistance, which avoids the camera being in continuous photo-taking mode and reduces the cost of recognition. At the same time, based on the first foreground image obtained by taking a photo, the corresponding feature information is searched in the material information database, and the material information database is updated and the current materials are sorted based on the search results, thereby improving the overall sorting efficiency and accuracy.

[0005] In a first aspect, embodiments of this application provide a method for updating a material information database. The method operates on a material conveying system, which includes: a support mechanism, a drive mechanism fixedly mounted on the support mechanism, a first conveyor belt and a second conveyor belt driven by the drive mechanism; the first conveyor belt and the second conveyor belt are connected end-to-end; an array of contact sensors is disposed on the upper surface of the first conveyor belt; and a camera is disposed above the first conveyor belt. The method includes:

[0006] After the contact sensor detects that the material has moved onto the first conveyor belt, the camera takes a picture of the upper surface of the first conveyor belt to obtain a first image of the material currently on the first conveyor belt.

[0007] Extract the first foreground image from the first image;

[0008] Based on the first foreground image, determine the current material's classification;

[0009] Based on the current material classification, query the feature information corresponding to the first foreground image in the material information database;

[0010] Based on the search results of the feature information, the material information database is updated, and the current material is sorted.

[0011] In conjunction with the first aspect, embodiments of this application provide a first possible implementation of the first aspect, wherein extracting the first foreground image from the first image includes:

[0012] Based on a pre-captured background image, the range of the second foreground image is determined from the first image;

[0013] The range of the second foreground image is adjusted according to the location of the contact sensor in the triggered state to determine the first foreground image.

[0014] In conjunction with the first aspect, this application provides a second possible implementation of the first aspect, wherein determining the classification of the current material based on the first foreground image includes:

[0015] Based on the pre-obtained quantity of materials to be conveyed, quantity of materials to be analyzed, and preset system calculation efficiency, the current calculation method is determined; the calculation method includes an efficiency calculation method and a precision-priority method.

[0016] If the current calculation method is an efficiency calculation method, the method further includes: extracting the edge contour of the material and the pixel information of a first pixel point whose distance from the edge contour is less than a preset distance from the first foreground image; and determining the classification of the current material based on the edge contour and the pixel information of the first pixel point.

[0017] If the current calculation method is a precision-first method, the method further includes: extracting the position information and color information of multiple pixels in the first foreground image; determining the color distribution of the current material based on the position information and color information between different pixels; determining the specification information of the material in the first foreground image based on the position information of the contact sensor in the triggered state and the color distribution of the current material; and determining the classification of the current material based on the color distribution and the specification information.

[0018] In conjunction with the second possible implementation of the first aspect, this application provides a third possible implementation of the first aspect, wherein the step of extracting the position information and color information of multiple pixels in the first foreground image; determining the color distribution of the current material based on the position information and color information between different pixels; and determining the specification information of the material in the first foreground image based on the position information of the contact sensor in the triggered state and the color distribution of the current material; includes:

[0019] The position and color information of multiple pixels are extracted from the first foreground image using downsampling. The first color distribution of the current material is determined based on the position and color information of the downsampled pixels. The specification information of the material in the first foreground image is determined based on the position information of the contact sensor in the triggered state and the first color distribution. If the determined specification information does not meet the preset requirements, the second color distribution of the current material is determined based on the position and color information of all pixels in the first foreground image. The specification information is then updated based on the position information of the contact sensor in the triggered state and the second color distribution.

[0020] In conjunction with the first aspect, this application provides a fourth possible implementation of the first aspect, wherein, after the contact sensor detects that material has moved onto the first conveyor belt, the camera takes a picture of the upper surface of the first conveyor belt to obtain a first image of the material currently located on the first conveyor belt, including:

[0021] After the target contact sensor is triggered, a timer starts. If the detection results of the target contact sensor and the contact sensors around it do not change within a predetermined time, the camera is triggered to take a picture of the upper surface of the first conveyor belt after the timer reaches the predetermined time to obtain the first image of the material currently on the conveyor belt.

[0022] If the detection results of the target contact sensor and the surrounding contact sensors show regular changes within the predetermined time, then after the triggered contact sensor reaches the predetermined position, the camera is triggered to continuously take pictures of the upper surface of the first conveyor belt, so as to generate a first image based on the images obtained from the continuous pictures.

[0023] In conjunction with the first aspect, this application provides a fifth possible implementation of the first aspect, wherein updating the material information database based on the search result of the feature information includes:

[0024] If the database information of the current material is found, the actual attribute information of the current material is compared with the database information of the current material. If the comparison is successful, the current material is processed according to the disposal method of the current material in the material information database. If the comparison fails and the deviation is less than a predetermined value, the database information of the current material in the material information database is updated according to the actual attribute information of the current material.

[0025] If the database information for the current material is not found, the current material is determined to be an abnormal material, and material information for the current material is created in the material information database.

[0026] In conjunction with the first aspect, this application provides a sixth possible implementation of the first aspect, wherein the color of the first conveyor belt and the color of the contact sensor are different.

[0027] Secondly, embodiments of this application also provide a material information database updating device, the device operating on a material conveying system, the material conveying system comprising: a support mechanism, a drive mechanism fixedly mounted on the support mechanism, a first conveyor belt and a second conveyor belt driven by the drive mechanism; the first conveyor belt and the second conveyor belt are connected end-to-end, an array of contact sensors is disposed on the upper surface of the first conveyor belt; a camera is disposed above the first conveyor belt; the device includes:

[0028] The camera module is used to take a picture of the upper surface of the first conveyor belt by the camera after the contact sensor detects that the material has moved onto the first conveyor belt, so as to obtain a first image of the material currently located on the first conveyor belt.

[0029] The extraction module is used to extract the first foreground image from the first image;

[0030] The determination module is used to determine the classification of the current material based on the first foreground image;

[0031] The query module is used to query the feature information corresponding to the first foreground image in the material information database according to the current material classification;

[0032] The update module is used to update the material information database based on the search results of the feature information, and to sort the current material.

[0033] Thirdly, embodiments of this application also provide an electronic device, including: a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the memory via the bus, and when the machine-readable instructions are executed by the processor, the steps in any of the possible implementations of the first aspect described above are performed.

[0034] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps in any of the possible implementations of the first aspect described above.

[0035] This application provides a method, apparatus, electronic device, and readable storage medium for updating a material information database. In determining the material classification, a contact sensor is used to assist in avoiding the camera being in continuous photo-taking mode, thus reducing the cost of identification. At the same time, based on the foreground image obtained from the photo, the corresponding feature information is searched in the material information database, and the material information database is updated and the current materials are sorted based on the search results, thereby improving the overall sorting efficiency and accuracy.

[0036] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0037] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 A schematic diagram of a material conveying system provided in an embodiment of this application is shown;

[0039] Figure 2 A schematic diagram of a contact sensor provided in an embodiment of this application is shown;

[0040] Figure 3 A flowchart illustrating a method for updating a material information database according to an embodiment of this application is shown;

[0041] Figure 4 A schematic diagram of the structure of a material information database updating device provided in an embodiment of this application is shown;

[0042] Figure 5A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0044] There are many types of automatic sorting machines. Some are specialized sorting machines, such as pharmaceutical sorting machines and ore sorting machines. These sorting machines primarily sort items based on specifications or size, as the specifications of these items are unlikely to change significantly. For example, when a pharmaceutical sorting machine is in operation, the prerequisite for mass production of pharmaceuticals, or even for a pharmaceutical item to pass through the sorting machine, is that the pharmaceuticals must undergo approval. This results in a limited variety of pharmaceuticals, and their packaging is generally standardized. This allows the material information database to pre-enter these limited types of pharmaceuticals, making sorting relatively easy. Ore sorting machines primarily sort ores based on size. Under current technological conditions, sorting by size is not particularly difficult and can therefore be completed relatively easily.

[0045] There is also a type of sorting machine called a non-dedicated sorting machine (general-purpose sorting machine). Non-dedicated sorting machines are characterized by irregular sorting objects, a wide variety of sorting objects, and rapid changes in sorting characteristics. The material information database update method provided in this application is designed for the material information database of this type of non-dedicated sorting machine. Its purpose is to ensure accurate object identification based on the material information database through updates, thereby maximizing identification efficiency and accuracy.

[0046] For non-dedicated sorting machines, the accuracy of the material information database determines the accuracy of identification. In the past, when there were fewer types of items, material information could be entered into the material information database before being transported to the sorting machine. This ensured that the identification and sorting of materials could be carried out entirely based on the information recorded in the database. However, with the increase in the types of items and the randomness of packaging, sorting efficiency has decreased or errors have occurred, and costs have become very high (requiring repeated training of the model for sorting, or the use of overly complex methods for sorting).

[0047] Based on this, this application provides a method for updating a material information database, aiming to balance sorting accuracy and efficiency to a certain extent. The material information database updating method provided in this application applies to material conveying systems, such as... Figure 1 and Figure 2 As shown, the material conveying system includes: a support mechanism, a drive mechanism fixedly mounted on the support mechanism, a first conveyor belt and a second conveyor belt driven by the drive mechanism, and a first sorting arm mounted on the first conveyor belt; the first conveyor belt and the second conveyor belt are connected end to end, and an array of contact sensors is provided on the upper surface of the first conveyor belt; a camera is provided above the first conveyor belt.

[0048] The support mechanism primarily serves to support the entire material conveying system; its shape is not specifically required, as long as it provides support. The drive mechanism typically consists of a rotating shaft centered around a motor. This rotating shaft can be divided into a driving shaft and a driven shaft. Generally, at least one rotating shaft exists in the first conveyor belt, and the second conveyor belt should also have one. The positional relationship between the first and second conveyor belts does not need to be as detailed. Figure 1 As shown in the diagram, as long as the material can smoothly move from the first conveyor belt to the second conveyor belt, there can be a height difference between the first and second conveyor belts; that is, the material conveyed by the first conveyor belt can fall directly onto the second conveyor belt. Alternatively, the two conveyor belts can be placed adjacent to each other. Figure 2 As shown, multiple contact sensors are installed on the first conveyor belt, and these contact sensors are arranged in an array. Figure 2 The arrangement shown is only one possible setup and does not necessarily have to be a square array; other rules can also be used, depending on the shape of the material. The main purpose of placing contact sensors only on the upper surface of the first conveyor belt is to save costs. Actual testing has shown that once the length of the first conveyor belt reaches a predetermined value (generally exceeding 3 meters), contact sensors only need to be placed on the first conveyor belt. The camera's primary purpose is to capture images of the first conveyor belt; therefore, the camera should be positioned above the first conveyor belt. Specifically, to avoid lighting issues during filming, the camera's normal should be perpendicular to the first conveyor belt, and the camera should be located at the end of the first conveyor belt.

[0049] like Figure 3 As shown, the method includes the following steps S101-S105:

[0050] S101, after detecting that the material has moved onto the first conveyor belt by the contact sensor, the upper surface of the first conveyor belt is photographed by the camera to obtain a first image of the material currently located on the first conveyor belt.

[0051] S102, Extract the first foreground image from the first image.

[0052] S103, Determine the classification of the current material based on the first foreground image.

[0053] S104. Based on the current material classification, query the feature information corresponding to the first foreground image in the material information database.

[0054] S105, based on the search results of the feature information, update the material information database and sort the current materials.

[0055] In step S101, after the contact sensor detects that material has moved onto the first conveyor belt, a camera can be used to take a picture of the upper surface of the first conveyor belt to obtain the first image. Generally, the triggering of a contact sensor indicates that material has moved onto the first conveyor belt. However, in some cases, if only a few contact sensors are triggered, it does not necessarily mean that material has moved onto the first conveyor belt. This is mainly determined by the density of the contact sensors and the size of the material. In actual use, the density of contact sensors should be determined based on the size of the smallest material. It should be ensured that the smallest material, when laid flat, can trigger at least 3-4 contact sensors to avoid false triggering.

[0056] In step S102, after determining the first image, the first foreground image can be directly extracted from the first image, thereby reducing the complexity of subsequent processing steps and improving overall efficiency. Specifically, before extracting the first foreground image, the first conveyor belt should be photographed in advance (without material on the first conveyor belt) to obtain a background image. Only then can the first foreground image be extracted from the first image based on this background image. To improve the accuracy of recognition, the colors of the first conveyor belt and the contact sensor should be set to be different from the color of the material.

[0057] In step S103, some features of the current material (the material in the first image) can be directly extracted based on the first foreground image, and the classification of the current material can be determined based on these features. Specifically, the features of the current material can be divided into two types: image features (which can be directly extracted from the image based on the first foreground image) and information features (which can be obtained through other means, such as weight, size, material, etc.). The main feature to be used should be determined according to the specific material situation (such as the number of types of materials and the differences between different materials). Generally speaking, at least image features should exist (mainly because most different materials can be well identified by image features). Information features are only needed to distinguish different materials when the image similarity between different materials is too high, that is, when image features alone are insufficient to distinguish different materials. In other words, it is not necessary to distinguish all different materials; it is only necessary to ensure that the two materials that need to be distinguished are distinguishable. Specifically, if large and small bottles of beverages are sorted to the same location on the sorting machine, then there is no need to distinguish between them; however, if large and small bottles of beverages are sorted to different locations on the sorting machine, then they need to be distinguished.

[0058] The information features here can be detected using other detection equipment. For example, weight can be detected using a belt scale, and size can be detected using contact sensors (how many contact sensors are triggered), or it can be detected using projection methods. The reason why information features are auxiliary is mainly because it is impossible to distinguish all materials using only information features, but image features can be used to distinguish them very well.

[0059] In step S104, after determining the classification of the current material, the characteristic information of the material can be queried from the material information database.

[0060] In step S105, if the corresponding feature information can be found, the current material can be sorted according to the sorting mechanism corresponding to the feature information. If the feature information cannot be found, there are two possibilities: first, there is a problem in the detection process; second, the current material is indeed newly arrived. The appropriate handling should be performed for each of these two scenarios. Generally, in both cases, the material can be sorted to a designated sorting area. After manual comparison, if it is the first case, the material information database can be updated based on the foreground image, and the classification mechanism determined in step S103 can be updated accordingly. If it is the second case, a new material category needs to be created in the material information database, and the feature information corresponding to the foreground image and the sorting strategy need to be input for subsequent use.

[0061] That is, step S105 can be implemented according to the following steps 1051-1052:

[0062] Step 1051: If the database information of the current material is found, the actual attribute information of the current material is compared with the database information of the current material. If the comparison is successful, the current material is processed according to the disposal method of the current material in the material information database. If the comparison fails and the deviation of the comparison is less than a predetermined value, the database information of the current material in the material information database is updated according to the actual attribute information of the current material.

[0063] Step 1052: If the database information for the current material is not found, the current material is determined to be an abnormal material, and a new database information for the current material is created in the material information database.

[0064] The actual attribute information in step 1051 refers to the shallow attribute information in the first foreground image or carried on the surface of the material. A comparison is made between the database information and the actual attribute information, primarily for verification. If the comparison passes, the material can be processed directly according to the handling method in the material information database. If the comparison fails, two scenarios need to be considered: if the deviation is small, it is considered the same material, and the database information is updated. If the deviation is large, manual judgment is required to determine if it is the same material. If it is, the database information is updated accordingly; if they are different, the method in step 1052 is required to create a new database for that material in the material information database. This process also requires manual intervention.

[0065] The material information database update method provided in this application utilizes a contact sensor to assist in determining material classification, thereby avoiding the need for the camera to continuously take pictures and reducing the cost of recognition. At the same time, based on the first foreground image obtained from the picture, the corresponding feature information is searched in the material information database, and the material information database is updated and the current materials are sorted based on the search results, which enhances the overall sorting efficiency and accuracy.

[0066] Specifically, step S102 can be implemented according to the following steps 1021-1022:

[0067] Step 1021: Determine the range of the second foreground image from the first image based on the pre-captured background image.

[0068] Step 1022: Adjust the range of the second foreground image according to the location of the contact sensor in the triggered state to determine the first foreground image.

[0069] In step 1021, the background image is typically a picture taken while the first conveyor belt is in operation. In practice, this step can be implemented in two ways. The first is when the colors of the contact sensor and the upper surface of the first conveyor belt do not affect the material. In this case, the colors of the contact sensor and the upper surface of the first conveyor belt are consistent, and their colors are clearly different from the material's color. This requires stricter requirements on the material; for example, both the contact sensor and the first conveyor belt must be black, and the material must be a color other than black. Thus, after capturing the first image, the pre-captured background image can be directly compared with the first image to determine the range of the second foreground image (the area where the material is located is determined based on image recognition technology).

[0070] The second scenario doesn't require the material's color, but demands a clear distinction between the contact sensor's color and the first conveyor belt's color. In this case, the contact sensor also assists in positioning; therefore, there must be a significant color contrast between the contact sensor and the first conveyor belt to ensure the contact sensor can be easily identified. After capturing the background image, a large model is used for training, and then the trained model is used to complete this step. A model framework can be built using PyTorch / TensorFlow, with OpenCV used for image processing, and the captured background images used as training samples. After obtaining the background images, a mask image corresponding to each background image needs to be manually generated (mainly due to the complexity of the material types, this step can only be done manually). After ensuring that most shapes of materials have appeared in the background images, the background images and corresponding masks can be used to generate training and validation sets. Due to the complexity of the materials, it is difficult to determine a precise loss function and optimization strategy. Furthermore, different types of locations have different material preferences; therefore, adjustments are generally made on-the-spot based on the actual training results. Since the materials in this scheme are all objects with volume, edge optimization techniques are basically unnecessary, and the training cost can be appropriately reduced.

[0071] The range obtained in step 1021 is not required to be absolutely accurate. Then, in step 1022, the range of the second foreground image is adjusted based on the triggering status of the contact sensors, thus obtaining the first foreground image. Generally, the second foreground image should cover all the contact sensors in the triggered state. Therefore, this step does not simply draw a range based on the position of the triggered contact sensors and take the union of that range with the range of the second foreground image. Instead, it uses the triggered contact sensors to verify the range of the second foreground image. If the second foreground image does not contain all the contact sensors in the triggered state, a verification mechanism should be triggered, requiring manual intervention or the current material being sorted to a temporary area.

[0072] Step S103 can be implemented in detail according to the following steps 1031-1033:

[0073] Step 1031: Determine the current calculation method based on the pre-acquired quantity of materials to be conveyed, quantity of materials to be analyzed, and preset system calculation efficiency; the calculation method includes efficiency calculation method and accuracy priority method.

[0074] Step 1032, if the current calculation method is an efficiency calculation method, the update method further includes: extracting the edge contour of the material and the pixel information of the first pixel point whose distance to the edge contour is less than a preset distance from the first foreground image; and determining the classification of the current material based on the edge contour and the pixel information of the first pixel point.

[0075] Step 1033: If the current calculation method is a precision-first method, the update method further includes: extracting the position information and color information of multiple pixels in the first foreground image; determining the color distribution of the current material based on the position information and color information between different pixels; determining the specification information of the material in the first foreground image based on the position information of the contact sensor in the triggered state and the color distribution of the current material; and determining the classification of the current material based on the color distribution and specification information.

[0076] Step 1031 primarily determines which method to use for subsequent processing. The greater the quantity of materials to be conveyed, the more efficient the calculation method should be. Similarly, the greater the quantity of materials to be analyzed (the first foreground image in step 103), the more efficient the calculation method should be. Since system calculation efficiency is typically within a certain range, it can be considered a fixed value. Theoretically, the lower the system calculation efficiency, the more efficient the calculation method should be. The quantity of materials to be conveyed refers to the quantity of materials about to be conveyed onto the first conveyor belt. This can be analyzed based on data from upstream transport vehicles or upstream conveyor belts. This analysis does not need to be highly precise; a general range is sufficient.

[0077] In efficiency mode, the analysis primarily focuses on extracting the edge contours of the data within the image and analyzing pixels located within and close to these edge contours to determine the classification. This approach is chosen mainly for two reasons:

[0078] 1. The presence of trademarks or similar corporate identifiers in the center of the materials can significantly impact identification accuracy. This is because a company's materials may not employ the same sorting strategy, but materials packaged in the same way usually use the same sorting strategy (sorting strategies can be efficient or robust; efficient sorting is fast but can cause some impact on the goods, while robust sorting is relatively slow). When a corporate logo or trademark occupies a large area of ​​the image (usually located in the center of the materials, i.e., the center of the first foreground image), it is highly likely that materials with the same corporate logo will be classified as the same type of material, thus affecting the accuracy of classification.

[0079] 2. Less data is processed, and some materials can even be identified simply by their edge contours, eliminating the need for auxiliary analysis of pixel information from the first pixel point whose distance from the edge contour is less than a preset distance.

[0080] Specifically, materials such as plush toys can usually be sorted using a more efficient sorting mechanism, as the shape of plush toys is relatively fixed (the edges and contours are relatively fixed). Conversely, when the material is a relatively regular shape (such as a cube, cuboid, etc.), a paper-based handling method should be used.

[0081] In precision-first mode, to improve calculation accuracy, the position and color information of all pixels in the first foreground image can be used for processing. That is, the color distribution is determined based on these two pieces of information, and then the material classification is determined by comparing the color distribution. Specifically, this step is mainly designed for situations where different types of materials have excessively high image similarity. Theoretically, after full-pixel recognition, the material classification should be determined. However, since the material's orientation cannot be accurately determined—that is, the material may be facing away from its back or side—only one side of the material is displayed in the first image, and this side is difficult to guarantee is the front. In this case, additional recognition methods are needed after full-pixel recognition to determine its classification. Specifically, these additional recognition methods mainly include weight and projected area recognition. Weight recognition can be achieved using a belt scale, and the projected area is determined by the position information of a contact sensor in a triggered state (this can be determined through other methods, such as point cloud data, but the cost is too high). In other words, this step should begin by analyzing the color distribution (this analysis should be performed at low resolution). Then, based on the color distribution and other material attribute information (weight and projection information), the material specifications should be determined, and finally, the classification should be determined using the specifications. However, if this information cannot accurately determine the specifications, then a high-resolution image analysis method should be considered (this method has a higher computational cost) to determine the color distribution before determining the specifications and classification.

[0082] In other words, the method provided in this application involves extracting the position and color information of multiple pixels in the first foreground image during the execution steps; determining the color distribution of the current material based on the position and color information between different pixels; and determining the specification information of the material in the first foreground image based on the position information of the contact sensor in the triggered state and the color distribution of the current material. This can be implemented as follows:

[0083] The position and color information of multiple pixels are extracted from the first foreground image using downsampling. The first color distribution of the current material is determined based on the position and color information of the downsampled pixels. The specification information of the material in the first foreground image is determined based on the position information of the contact sensor in the triggered state and the first color distribution. If the determined specification information does not meet the preset requirements (generally, if the specification information cannot point to a unique sorting method, it is considered that the specification information does not meet the preset requirements), the second color distribution of the current material is determined based on the position and color information of all pixels in the first foreground image. The specification information is then updated based on the position information of the contact sensor in the triggered state and the second color distribution.

[0084] In step S104, the determined classification is used to find the corresponding feature information, which includes sorting strategy (efficiency-oriented or robust), sorting destination, and may further include product information of the materials. Afterwards, the determined feature information can be used to execute subsequent control processes.

[0085] The specific step S101 can be implemented according to the following steps 1011-1012:

[0086] Step 1011: After the target contact sensor is triggered, a timer is started. If the detection results of the target contact sensor and the contact sensors around it do not change within a predetermined time, the camera is triggered to take a picture of the upper surface of the first conveyor belt after the timer reaches the predetermined time to obtain the first image of the material currently located on the conveyor belt.

[0087] Step 1012: If the detection results of the target contact sensor and the contact sensors around it show regular changes within a predetermined time, then after the triggered contact sensor reaches the predetermined position, the camera is triggered to take continuous pictures of the upper surface of the first conveyor belt, so as to generate a first image based on the images obtained from the continuous pictures.

[0088] Step 1011 mainly involves taking photos after the material has stably stopped on the first conveyor belt. The target contact sensor refers to a contact sensor whose trigger state changes within one operating cycle (a change in the sensor's trigger state within one revolution of the conveyor belt indicates normal operation). At this point, a timing method can be used to trigger the camera to take a picture. Alternatively, the camera could be triggered when the contact sensor detects the material approaching a certain position; however, in practice, due to the uncertain size of the material, some small materials may not trigger the contact sensor. Therefore, the timing method is more reliable.

[0089] Step 1012 mainly refers to the situation where the material rolls on the first conveyor belt. In this case, the contact sensors on the first conveyor belt are usually triggered sequentially from front to back. The first contact sensor that comes into contact with the material is already in an untriggered state when the material is transported into the camera's field of view. At this time, the camera can be triggered to work and take continuous pictures based on the appearance of the contact sensors in the triggered state in the camera's field of view. Then, the first image is generated by image stitching.

[0090] Specifically, when performing image stitching, the first step is to identify the feature points in different images (in a given image, the image features of a feature point are clearly different from the image features of its surrounding pixels). Then, feature point matching is used to stitch the different images together and complete the subsequent recognition. In fact, the recognition accuracy of this stitched image is very high.

[0091] Building upon the solution provided in this application, a systematic intelligent governance method and process for material master data based on AI technology can be further developed. This addresses a series of problems arising from the traditional material management process, which focuses solely on the material digitization and coding process. Specific problems include:

[0092] 1. The construction did not take into account relevant national and industry standards, and relied entirely on experience, which does not conform to national standards.

[0093] 2. The coding process did not incorporate relevant expert and industry knowledge, and the material architecture lacked scientific rigor and uniqueness.

[0094] 3. Lack of governance after the material construction is completed results in an excessive number of master data, multiple items with one code, and one item with multiple codes, which in turn leads to problems such as incorrect or missing material information attributes.

[0095] 4. The application process is difficult for applicants, as there are no effective reference examples or pathways, and timely verification of any deficiencies is necessary.

[0096] To address the aforementioned issues, specific solutions include constructing large-scale material standard models, industry-specific models, standard processing models, and intelligent data entry models. This will enable the standardization, scientification, and intelligentization of the material system, effectively resolving prominent problems encountered in traditional construction processes such as multiple codes for a single item, multiple materials, attribute errors and omissions, and difficulties in data entry.

[0097] The specific technology may include the following parts:

[0098] 1. Collection of existing material information

[0099] Material data collection is the first and crucial step in material cleansing. This stage primarily involves assessing the relevant business scope of the group using the material master data system, collecting, processing, and standardizing material data. It mainly includes the following steps:

[0100] Business scope assessment: Conduct a business scope assessment based on the Group's existing business operations and requirements, such as those involving materials, equipment, raw materials, etc.

[0101] Data collection: Collect material data based on the customer's current situation, including but not limited to existing master data platforms, material systems, procurement systems, ERP, documents, etc., to ensure the comprehensiveness of the data.

[0102] Data processing: The collected data is processed according to a standard architecture to provide a foundation for subsequent steps.

[0103] Second, the creation of a discovery model for national and industry standards for materials.

[0104] Based on the collected material data, a national and industry standard model for enterprise materials is constructed. This model can identify the national and industry standards that the enterprise's existing material data needs to comply with, providing a guarantee for the subsequent establishment of material architecture and material attributes. The specific construction process is as follows:

[0105] Data standardization processing: Preprocessing operations such as deduplication and standardization are performed on the collected material data to improve data quality.

[0106] Classification planning: Based on the standardized material information, classification planning is carried out.

[0107] Model training: The results of classification planning are used as input to train a large model so that national and industry standards that each classification should meet can be discovered.

[0108] Third, analytical models for national and industry standards for materials.

[0109] Based on the discovered national and industry standards, after manual verification, a national and industry standard library is constructed using the analytical capabilities of a large model. This ensures the standardization of subsequent material architecture and attribute settings, as well as attribute values. The relevant steps are as follows:

[0110] Standard analysis: The standards that have been manually reviewed are analyzed through a large model to extract the relevant classifications, indicators and parameters of the model design.

[0111] Structuring: The data after standard parsing is processed in a structured manner according to indicators and related parameters.

[0112] Fourth, the leaf attribute discovery model

[0113] Based on the relevant material architecture, the parameters and related descriptions that the corresponding leaf type should possess can be found. The relevant steps are as follows:

[0114] After inputting the relevant leaf class, the large model will output a list of relevant attributes and the range to be filled based on the leaf class information and relevant national and industry standards.

[0115] 5. Attribute Synonym Discovery Model

[0116] Attribute synonym discovery models are crucial in the material cleaning process. They are used to identify and process different representations of the same attribute in material data, thereby supporting data standardization models. The relevant steps of attribute synonym discovery models are as follows:

[0117] Attribute acquisition: Obtain the attribute value of a specified attribute for a specified leaf class.

[0118] Grouping processing: Combine with an industry dictionary and group the attribute values according to relevant grouping algorithms.

[0119] Synonym discovery: According to the grouping information, feed the grouped content to a large model for synonym discovery and elect a synonym standard at the same time.

[0120] Thesaurus construction: Structurally process the results of synonym discovery.

[0121] (1) The discovered synonyms have the following characteristics:

[0122] Missing unit: such as 7.5; 7.5kw.

[0123] Case difference: St; st.

[0124] Chinese and English comparison: kilogram; KG.

[0125] Suspected inclusion: grease; grease lubrication.

[0126] Range overlap: ≤5mm; 0 - 5.

[0127] Typo: bobbin paper; sand tube paper.

[0128] Cardinal numbers: three poles / three levels, 3 poles / 3 levels.

[0129] Abbreviations: Unisplendour Corporation, Unisplendour Corporation Limited

[0130] Format difference: Q235 - B; Q235B

[0131] Technical and common descriptions: polyvinyl chloride; PVC

[0132] Professional terms: OS, operating system, etc.

[0133] (2) Accuracy guarantee measures

[0134] To improve the accuracy of the thesaurus:

[0135] Industry dictionary creation: Based on relevant classifications and standards, establish an industry dictionary to improve the synonym recognition rate and grouping accuracy.

[0136] Specific guarantees are as follows:

[0137] Optimization of the understanding of proper noun phrases:

[0138] Implementation method: Through web crawling and corpus construction, collect context information related to proper nouns.

[0139] Technical Implementation: Leveraging the contextual understanding capabilities of large language models to enhance the semantic parsing of specialized terms.

[0140] Supporting measures: Establish a domain knowledge graph and improve the relationship network of proper nouns.

[0141] Standardization of company name abbreviations:

[0142] Core algorithm: Automatic abbreviation recognition is achieved by calculating semantic similarity and setting dynamic thresholds.

[0143] Knowledge base construction: Construct a database mapping relationships between full company names and abbreviations.

[0144] System integration: Integrate the abbreviation dictionary into the thesaurus management system to enable real-time query and matching.

[0145] Intelligent typo correction:

[0146] Error correction mechanism: Utilizes the text error correction capabilities of large model to achieve automatic detection and correction.

[0147] Knowledge accumulation: Establish a database of misspelled words and their evolution, recording common error patterns and their correct forms.

[0148] Continuous optimization: Expand the typo knowledge base through user feedback mechanisms.

[0149] Unit standardization:

[0150] Model training: Train a unit normalized classifier based on multi-unit sample data.

[0151] Conversion Engine: Develop an intelligent unit conversion system that integrates international standard conversion rules.

[0152] Dynamic updates: Maintains an extensible unit conversion rule library and supports the addition of custom units.

[0153] Standardize terminology in both Chinese and English:

[0154] Alignment Algorithm: Based on co-occurrence analysis and contextual similarity calculation, this algorithm achieves Chinese-English matching.

[0155] Terminology database construction: Building a bilingual database of professional terms in the field.

[0156] Intelligent recommendation: Enables automatic recommendation and replacement of Chinese and English terminology.

[0157] Standardization of comparison objects:

[0158] Sample construction: Create a training sample set based on multi-dimensional data (attribute names, material information, etc.).

[0159] Model development: Training a dedicated model for object recognition and transformation.

[0160] Rule base construction: Establish a comparison object transformation rule engine to support complex scenario processing.

[0161] Deduplication of proper nouns:

[0162] Data preparation: Construct a sample library of proper nouns containing multiple expressions.

[0163] Model training: Develop a deep learning-based model for discriminating representations.

[0164] Knowledge integration: Establish a standard dictionary of proper nouns and achieve automatic merging.

[0165] The system combines the results from multiple models to arrive at a final conclusion. Weights and thresholds can be set to adjust the results according to specific needs.

[0166] VI. Rule Base Creation

[0167] Rule base creation is a key function of intelligent material cleaning, which mainly includes: general rules, proprietary rules, escape rules, and national and industry standard verification to ensure the quality of material data.

[0168] (1) General rules

[0169] The general rules primarily apply to all materials, ensuring the standardization of material data globally. These include: for example, the default length unit is millimeters (mm), which does not need to be entered into the system; other units should use lowercase letters (e.g., um, cm, m, km, kg); the first letter of unit symbols derived from personal names must be capitalized (e.g., Hz (Hertz), V (Volt), A (Ampere), Pa (Pascal)); special conventions: liters are capitalized as L, milliliters as ml; all symbols in the system should be in half-width characters (e.g., #φ*-), and multiplication signs should be *; spaces should not be used for separation, but rather half-width commas or hyphens should be used, and pauses should always be half-width commas, etc., to ensure the standardization and normalization of material attribute entry.

[0170] (2) Proprietary rule base

[0171] The main purpose of this standardization is to further regulate data in designated categories to achieve more precise management. Taking conveyor belts as an example, their classification can only be: V-belts, synchronous belts, winding machine belts, and winding machine belts; their unit of measurement is uniformly standardized as "strip," and their specifications are uniformly standardized as: thickness * width * length.

[0172] (3) National and industry standards database

[0173] The main purpose is to analyze relevant national and industry standards to determine the relevant filling requirements, restrictions, or scope.

[0174] Example: Taking a circuit breaker as an example, its rating is: 0P 1P 1P+N 2P 3P 3P+N 4P 4P+N; the rating of a small circuit breaker is below 125A, while that of a medium or large circuit breaker is 125A and above.

[0175] (4) Required fields

[0176] Definition: Used to indicate whether the material attribute of a specified leaf type is required.

[0177] Function: Attributes must be filled in when applying for materials; provides basic protection for escape verification.

[0178] (5) Escape Thesaurus

[0179] Definition: These are mainly random, meaningless words, such as: none, none, null, etc.

[0180] Function: To provide support for discovering those who evade mandatory field checks by filling in meaningless words.

[0181] (6) Synonyms

[0182] Definition: It mainly uses a large model to analyze and summarize the different manifestations of material properties.

[0183] Function: To conduct quality inspection and management of material properties, so that the material properties are all in accordance with the same standard.

[0184] (7) Merge rule bases

[0185] Definition: This mainly sets up the corpus used in the material cleaning process. The system defaults to required fields as the corpus, and supports online settings, modifications, shutdowns, and scope settings according to categories.

[0186] Function: To provide a basis for merging materials during the cleaning and merging process.

[0187] 7. Rule-based cleaning model

[0188] After establishing the rule base, the material data quality inspection and cleanup work can begin, mainly from the following aspects:

[0189] General rule inspection: Clean all materials and produce materials that do not conform to the general rules, along with details of the reasons for non-conformity.

[0190] Proprietary rule inspection: All materials are cleaned according to their specific proprietary rules, and the results of materials that do not conform to the proprietary rules for specified leaf types are provided along with details of the reasons for non-conformity.

[0191] National and industry standard testing: All materials are tested according to the national and industry standards they need to comply with, and any non-compliance is identified.

[0192] Escape detection: Combining the required field database and the escape word database, it detects cases where applicants bypass the required field validation by using meaningless words and generates non-compliance details.

[0193] 8. Material Standard Processing Model

[0194] The data, after being cleaned according to rules, will be further improved in quality through a material standard processing model. The relevant steps are as follows:

[0195] (1) Standardization process: mainly includes removing spaces, converting uppercase and lowercase letters, and unifying formats to ensure that materials are on the same standard.

[0196] (2) Synonymization: Combine the thesaurus and relevant industry standards to reprocess the data, such as standardizing the series 1p and 1, os and operating system.

[0197] The above steps further standardized and improved the quality of the data.

[0198] 9. Material Repeat Verification Model

[0199] The material cleaning model can clean material data based on the material classification and related merging rules to identify instances of one item having multiple codes or multiple items having one code. The relevant steps are as follows:

[0200] (1) Call the material standard processing model to standardize the material data.

[0201] (2) Collect the merge rule library in the system and see if there are merge rules for the leaf classes that need to be merged. If there are, use them directly; otherwise, the model will automatically obtain merge rules based on the basic and technical attributes of the leaf classes.

[0202] (3) The standardized data and the discovered merging rules are sent to the material merging model for merging and cleaning.

[0203] (4) After cleaning, return any materials that are suspected of being duplicates.

[0204] 10. Material Library Iteration

[0205] Based on the details of suspected duplicate materials, and after manual review of the material data and confirmation by the declaring unit according to the agreed template, the material library is iterated, mainly including:

[0206] Materials suspected of being duplicates are manually verified.

[0207] The details after manual review will be sent to the applicant for confirmation, mainly including whether the merger is agreed upon and the confirmation of the materials to be used after the merger.

[0208] The data confirmed by the reporting units will be collected and summarized to form the latest details of duplicate and merged materials.

[0209] The material library is iterated based on the material duplicate material details obtained from (3).

[0210] 11. Material verification capability

[0211] The various validation capabilities of the rule base are encapsulated and standardized validation capabilities are provided to external parties to facilitate the material entry of planning and reporting units and to verify the duplication of materials in real time.

[0212] Mainly includes:

[0213] General rules, special rules, national and industry standards, and escape inspection.

[0214] Verify whether the material is frozen before adding it.

[0215] Verify whether the materials entered are duplicates of those in the inventory. If duplicates are found, provide relevant details.

[0216] 12. Freeze monitoring

[0217] This function is mainly designed to prevent frozen data from being added through the "new" method. The idea is to compare the newly added data with the frozen database to determine whether it is newly added data after freezing.

[0218] Thirteen, Repair the monitoring system

[0219] This function is mainly for monitoring the materials submitted by the applicants. It monitors the cleaning issues, the coverage of repairs, and the type of repair (freezing, repair), allowing for a direct view of the material quality and repair status of each applicant.

[0220] The key to the above technology is that, by leveraging the learning and logical reasoning algorithms of large models and combining relevant national standards, industry standards and industry knowledge, models are constructed for each link of material management. These models mainly include standard discovery, attribute discovery, rule establishment, rule cleaning, material standard processing, and material duplication verification, thereby reducing the complexity of the reporting units' filling out and ensuring the data standardization and accuracy of the unified material library.

[0221] The corresponding technological value is:

[0222] (1) A complete material architecture system was constructed based on national standards, industry standards and industry knowledge.

[0223] (2) By leveraging the capabilities of large models, the difficulty of application for applicants can be reduced, and accompanying services can be provided.

[0224] (3) Establish a material quality system and management rules to ensure the integrity and quality of material data.

[0225] (4) Comprehensive verification and cleaning to reduce manual input and improve treatment effect and efficiency.

[0226] Based on the same technical concept, this application also provides a material information database updating device. The device operates on a material conveying system, which includes: a support mechanism, a drive mechanism fixedly mounted on the support mechanism, and a first conveyor belt and a second conveyor belt driven by the drive mechanism. The first and second conveyor belts are connected end-to-end. An array of contact sensors is disposed on the upper surface of the first conveyor belt. A camera is disposed above the first conveyor belt. Figure 4 As shown, the device includes:

[0227] The camera module 401 is used to take a picture of the upper surface of the first conveyor belt by the camera after the contact sensor detects that the material has moved onto the first conveyor belt, so as to obtain a first image of the material currently located on the first conveyor belt.

[0228] Extraction module 402 is used to extract the first foreground image from the first image;

[0229] The determining module 403 is used to determine the classification of the current material based on the first foreground image;

[0230] The query module 404 is used to query the feature information corresponding to the first foreground image in the material information database according to the current material classification;

[0231] The update module 405 is used to update the material information database based on the search results of the feature information, and to sort the current material.

[0232] Optionally, when extracting the first foreground image from the first image, the extraction module 402 is specifically used for:

[0233] Based on a pre-captured background image, the range of the second foreground image is determined from the first image;

[0234] The range of the second foreground image is adjusted according to the location of the contact sensor in the triggered state to determine the first foreground image.

[0235] Optionally, when determining the classification of the current material based on the first foreground image, the determining module 403 is specifically used for:

[0236] Based on the pre-obtained quantity of materials to be conveyed, quantity of materials to be analyzed, and preset system calculation efficiency, the current calculation method is determined; the calculation method includes an efficiency calculation method and a precision-priority method.

[0237] If the current calculation method is an efficiency calculation method, the determining module 403 is further configured to: extract the edge contour of the material and the pixel information of a first pixel point whose distance from the edge contour is less than a preset distance from the first foreground image; and determine the classification of the current material based on the edge contour and the pixel information of the first pixel point.

[0238] If the current calculation method is a precision-first method, the determining module 403 is further configured to: extract the position information and color information of multiple pixels in the first foreground image; determine the color distribution of the current material based on the position information and color information between different pixels; determine the specification information of the material in the first foreground image based on the position information of the contact sensor in the triggered state and the color distribution of the current material; and determine the classification of the current material based on the color distribution and the specification information.

[0239] Optionally, when the determining module 403 is used to extract the position and color information of multiple pixels in the first foreground image; determine the color distribution of the current material based on the position and color information between different pixels; and determine the specification information of the material in the first foreground image based on the position information of the contact sensor in the triggered state and the color distribution of the current material, it is specifically used for:

[0240] The position and color information of multiple pixels are extracted from the first foreground image using downsampling. The first color distribution of the current material is determined based on the position and color information of the downsampled pixels. The specification information of the material in the first foreground image is determined based on the position information of the contact sensor in the triggered state and the first color distribution. If the determined specification information does not meet the preset requirements, the second color distribution of the current material is determined based on the position and color information of all pixels in the first foreground image. The specification information is then updated based on the position information of the contact sensor in the triggered state and the second color distribution.

[0241] Optionally, when the imaging module 401 is used to take a picture of the upper surface of the first conveyor belt by the camera after the contact sensor detects that the material has moved onto the first conveyor belt, in order to obtain a first image of the material currently located on the first conveyor belt, it is specifically used for:

[0242] After the target contact sensor is triggered, a timer starts. If the detection results of the target contact sensor and the contact sensors around it do not change within a predetermined time, the camera is triggered to take a picture of the upper surface of the first conveyor belt after the timer reaches the predetermined time to obtain the first image of the material currently on the conveyor belt.

[0243] If the detection results of the target contact sensor and the surrounding contact sensors show regular changes within the predetermined time, then after the triggered contact sensor reaches the predetermined position, the camera is triggered to continuously take pictures of the upper surface of the first conveyor belt, so as to generate a first image based on the images obtained from the continuous pictures.

[0244] Optionally, when the update module 405 updates the material information database based on the search results of the feature information, it is specifically used for:

[0245] If the database information of the current material is found, the actual attribute information of the current material is compared with the database information of the current material. If the comparison is successful, the current material is processed according to the disposal method of the current material in the material information database. If the comparison fails and the deviation is less than a predetermined value, the database information of the current material in the material information database is updated according to the actual attribute information of the current material.

[0246] If the database information for the current material is not found, the current material is determined to be an abnormal material, and a new database information for the current material is created in the material information database.

[0247] Optionally, the color of the first conveyor belt and the color of the contact sensor are different.

[0248] Figure 5 A schematic diagram of an electronic device provided in this application embodiment includes: a processor 501, a memory 502, and a bus 503. The memory 502 stores machine-readable instructions executable by the processor 501. When the electronic device runs the above-described information processing method, the processor 501 and the memory 502 communicate through the bus 503. The processor 501 executes the machine-readable instructions to perform the steps of the method described in Embodiment 1.

[0249] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps described in Embodiment 1.

[0250] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, electronic devices, and computer-readable storage media described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0251] In the several embodiments provided in this application, it should be understood that the disclosed methods, apparatuses, electronic devices, and computer-readable storage media can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or modules may be electrical, mechanical, or other forms.

[0252] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0253] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0254] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0255] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The scope of protection of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of protection of the claims.

Claims

1. A method for updating a material information database, characterized in that, The method applies to a material conveying system, which includes: a support mechanism, a drive mechanism fixedly mounted on the support mechanism, a first conveyor belt and a second conveyor belt driven by the drive mechanism; the first conveyor belt and the second conveyor belt are connected end-to-end, and an array of contact sensors is disposed on the upper surface of the first conveyor belt; a camera is disposed above the first conveyor belt; the method includes: After the contact sensor detects that the material has moved onto the first conveyor belt, the camera takes a picture of the upper surface of the first conveyor belt to obtain a first image of the material currently on the first conveyor belt. Extract the first foreground image from the first image; Based on the first foreground image, determine the current material's classification; Based on the current material classification, query the feature information corresponding to the first foreground image in the material information database; Based on the search results of the feature information, the material information database is updated, and the current material is sorted. Determining the classification of the current material based on the first foreground image includes: Based on the pre-obtained quantity of materials to be conveyed, quantity of materials to be analyzed, and preset system calculation efficiency, the current calculation method is determined; the calculation method includes an efficiency calculation method and a precision-priority method. If the current calculation method is an efficiency calculation method, the method further includes: extracting the edge contour of the material and the pixel information of a first pixel point whose distance from the edge contour is less than a preset distance from the first foreground image; and determining the classification of the current material based on the edge contour and the pixel information of the first pixel point. If the current calculation method is a precision-first method, the method further includes: extracting the position information and color information of multiple pixels in the first foreground image; determining the color distribution of the current material based on the position information and color information between different pixels; determining the specification information of the material in the first foreground image based on the position information of the contact sensor in the triggered state and the color distribution of the current material; and determining the classification of the current material based on the color distribution and the specification information.

2. The method according to claim 1, characterized in that, Extracting the first foreground image from the first image includes: Based on a pre-captured background image, the range of the second foreground image is determined from the first image; The range of the second foreground image is adjusted according to the location of the contact sensor in the triggered state to determine the first foreground image.

3. The method according to claim 1, characterized in that, The steps include: extracting the position and color information of multiple pixels in the first foreground image; determining the color distribution of the current material based on the position and color information of different pixels; and determining the specification information of the material in the first foreground image based on the position information of the contact sensor in the triggered state and the color distribution of the current material. The position and color information of multiple pixels are extracted from the first foreground image using downsampling. The first color distribution of the current material is determined based on the position and color information of the downsampled pixels. The specification information of the material in the first foreground image is determined based on the position information of the contact sensor in the triggered state and the first color distribution. If the determined specification information does not meet the preset requirements, the second color distribution of the current material is determined based on the position and color information of all pixels in the first foreground image. The specification information is then updated based on the position information of the contact sensor in the triggered state and the second color distribution.

4. The method according to claim 1, characterized in that, After the contact sensor detects that the material has moved onto the first conveyor belt, the camera takes a picture of the upper surface of the first conveyor belt to obtain a first image of the material currently on the first conveyor belt, including: After the target contact sensor is triggered, a timer starts. If the detection results of the target contact sensor and the contact sensors around it do not change within a predetermined time, the camera is triggered to take a picture of the upper surface of the first conveyor belt after the timer reaches the predetermined time to obtain the first image of the material currently on the conveyor belt. If the detection results of the target contact sensor and the surrounding contact sensors show regular changes within the predetermined time, then after the triggered contact sensor reaches the predetermined position, the camera is triggered to continuously take pictures of the upper surface of the first conveyor belt, so as to generate a first image based on the images obtained from the continuous pictures.

5. The method according to claim 1, characterized in that, The step of updating the material information database based on the search results of the feature information includes: If the database information of the current material is found, the actual attribute information of the current material is compared with the database information of the current material. If the comparison is successful, the current material is processed according to the disposal method of the current material in the material information database. If the comparison fails and the deviation is less than a predetermined value, the database information of the current material in the material information database is updated according to the actual attribute information of the current material. If the database information for the current material is not found, the current material is determined to be an abnormal material, and a new database information for the current material is created in the material information database.

6. The method according to claim 1, characterized in that, The color of the first conveyor belt is different from the color of the contact sensor.

7. A device for updating a material information database, characterized in that, The device operates on a material conveying system, which includes: a support mechanism, a drive mechanism fixedly mounted on the support mechanism, a first conveyor belt and a second conveyor belt driven by the drive mechanism; the first conveyor belt and the second conveyor belt are connected end-to-end, and an array of contact sensors is provided on the upper surface of the first conveyor belt; a camera is provided above the first conveyor belt; the device includes: The camera module is used to take a picture of the upper surface of the first conveyor belt by the camera after the contact sensor detects that the material has moved onto the first conveyor belt, so as to obtain a first image of the material currently located on the first conveyor belt. The extraction module is used to extract the first foreground image from the first image; The determination module is used to determine the classification of the current material based on the first foreground image; The query module is used to query the feature information corresponding to the first foreground image in the material information database according to the current material classification; The update module is used to update the material information database based on the search results of the feature information, and to sort the current material. When the determining module is used to determine the classification of the current material based on the first foreground image, it is specifically used for: Based on the pre-obtained quantity of materials to be conveyed, quantity of materials to be analyzed, and preset system calculation efficiency, the current calculation method is determined; the calculation method includes an efficiency calculation method and a precision-priority method. If the current calculation method is an efficiency calculation method, the determining module is further configured to: extract the edge contour of the material and the pixel information of a first pixel point whose distance from the edge contour is less than a preset distance from the first foreground image; and determine the classification of the current material based on the edge contour and the pixel information of the first pixel point. If the current calculation method is a precision-first method, the determining module is further configured to: extract the position information and color information of multiple pixels in the first foreground image; determine the color distribution of the current material based on the position information and color information between different pixels; determine the specification information of the material in the first foreground image based on the position information of the contact sensor in the triggered state and the color distribution of the current material; and determine the classification of the current material based on the color distribution and the specification information.

8. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is in operation, the processor communicates with the memory via the bus, and the machine-readable instructions, when executed by the processor, perform the steps of the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the method as described in any one of claims 1 to 6.

Images