Hazardous chemical substance inspection method, system and device and storage medium

Abstract

The invention relates to a hazardous chemical substance checking method, system and device and a storage medium, and belongs to the technical field of hazardous chemical substance image recognition, and the method comprises the steps that an image collection terminal collects image data of hazardous chemical substances; the server generates a to-be-detected label graph according to the image data, wherein the to-be-detected label graph is an image of a local area where the hazardous chemical substance is located; the server further obtains a target label graph according to the to-be-detected label graph and a standard label graph, the standard label graph is a label graph obtained by training in advance according to a pictogram of the hazardous chemical substance, and the target label graph is a label graph used for indicating the category of the hazardous chemical substance. According to the method, the image data of the hazardous chemical substance is acquired through the image acquisition terminal, and the category of the hazardous chemical substance is intelligently checked by the server according to the image data, so that the manual checking process is reduced, and the effect of improving the checking efficiency of the hazardous chemical substance is achieved.

Description

Technical Field

[0001] The present application relates to the technical field of image recognition of hazardous chemicals, and in particular to a method, system, device and storage medium for inspecting hazardous chemicals. Background Art

[0002] Hazardous chemicals are chemicals with explosive, corrosive, toxic, flammable, and combustion-supporting properties, which are harmful to humans, equipment, and the environment. Therefore, the United Nations has formulated the Globally Harmonized System of Classification and Labeling of Chemicals, referred to as GHS, so that companies can correctly classify hazardous chemicals and affix GHS labels when producing, operating, importing and exporting hazardous chemicals, and then the import and export customs can check the compliance of hazardous chemicals through the company's declaration information and the GHS labels on the hazardous chemicals.

[0003] At present, the identification of hazardous chemicals at the customs and enterprise ends still adopts the manual review method, and there are few or no cases of intelligent verification of hazardous chemicals with the help of deep learning models. At the same time, for the identification of imported and exported hazardous chemicals, the manual review method requires a lot of manpower and material resources. For example, customs staff need to take the enterprise application documents to manually verify whether the GHS label attached to the hazardous chemicals is consistent with the GHS label declared by the enterprise. If the number of hazardous chemicals is large, the inspection will take longer. The time lag caused by the low inspection efficiency will cause the hazardous chemicals to be stranded in the enterprise, requiring more storage costs. Summary of the invention

[0004] In order to solve the problem of low efficiency in hazardous chemicals inspection, the present application provides a hazardous chemicals inspection method, system, device and storage medium.

[0005] In a first aspect of the present application, a method for inspecting hazardous chemicals is provided. The method comprises: The image acquisition terminal collects image data of hazardous chemicals; The server generates a label image to be tested according to the image data, where the label image to be tested is an image of a local area where the hazardous chemicals are located; The server also obtains a target label image based on the label image to be tested and the standard label image, wherein the standard label image is a label image trained in advance based on the pictogram of the hazardous chemicals, and the target label image is a label image used to indicate the category to which the hazardous chemicals belong.

[0006] By adopting the above technical solution, first, the inspector uses the image acquisition terminal to collect the image data of hazardous chemicals, freeing the inspector's hands while ensuring the personal safety of the inspector. Then, the server receives the image data, intelligently detects the local area image where the hazardous chemicals are located in the image data, and obtains the label image to be tested, and then matches the corresponding standard label image based on the obtained label image to be tested, so as to achieve the purpose of intelligent classification of hazardous chemicals, reduce the process of manual review, and thus improve the inspection efficiency of hazardous chemicals.

[0007] In a possible implementation manner: the server further obtains a target label image according to the label image to be tested and the standard label image, including: Identify a feature to be tested in the label image to be tested, where the feature to be tested is used to reflect the category to which the label image to be tested belongs; Calculating the similarity between the label image to be tested and the standard label image based on the feature to be tested; The standard label image with the highest similarity to the label image to be tested is selected as the target label image.

[0008] In a possible implementation manner: the feature to be measured is located within a marking frame; The calculating the similarity between the label image to be tested and the standard label image according to the feature to be tested includes: Calculating the confidence between the feature to be tested and the standard feature in each identification frame, wherein the standard feature is used to reflect the category to which the standard label image belongs; Selecting a standard feature with the highest confidence level with the feature to be measured as a target feature of the identification frame; The number of target features contained in the standard label image is proportional to the similarity between the label image to be tested and the standard label image.

[0009] In a possible implementation manner: the features to be measured include text, color, contour, texture, and shape.

[0010] By adopting the above technical solution, various features to be tested, such as text, color, contour, texture, shape, etc. in the label image to be tested are used as the basis for selecting the target label image, thereby ensuring the accuracy of the selected target label image, and then ensuring the accuracy of the category to which the label image to be tested obtained by analysis belongs.

[0011] In a possible implementation, the method further includes: When the number of target features contained in multiple standard label images is the same: Calculating the confidence value of each of the plurality of standard label images, wherein the confidence value is the sum of the confidences between the feature to be measured in each identification frame and the target feature corresponding to the feature to be measured on the label image to be measured; A standard label image with the largest confidence value is used as the target label image.

[0012] By adopting the above technical solution, when there are multiple standard label images with the same number of target features, that is, when the similarity between the label image to be tested and the multiple standard label images is the same, the standard label image with the largest confidence value is used as the target label image, so that the selected target label image not only contains the most features that are the same as the label image to be tested, but also the identical features are the closest, thereby further ensuring the accuracy of the selected target label image.

[0013] In a possible implementation manner: the server generates a label image to be tested according to the image data, including: Using the local area image where the hazardous chemicals are located in the image data as the intermediate label image; The aspect ratio of the intermediate label image is adjusted to a preset ratio, and the interference information in the intermediate label image is suppressed by using a high temperature refinement and background suppression model to obtain the label image to be tested.

[0014] By adopting the above technical solution, firstly, the local area image where the hazardous chemicals are located is extracted as the intermediate label image to reduce the interference of other area images on the inspection results. At the same time, the width and height ratio of the intermediate label image is adjusted and the interference information is suppressed to further reduce the interference of other interference information on the inspection results and ensure the accuracy of the inspection results.

[0015] In a possible implementation manner: before the server generates a label image to be tested according to the image data, the method further includes: Collect pictograms of various hazardous chemicals to create training samples; Using the training samples to train the AI ​​recognition algorithm; The trained AI recognition algorithm is used to generate a label map to be tested based on the image data.

[0016] By adopting the above technical solution, this application creates training samples based on pictograms of various hazardous chemicals, and then uses the created training samples to train the AI ​​recognition algorithm to ensure the accuracy of the AI ​​recognition algorithm. Based on the trained AI recognition algorithm, the AI ​​recognition algorithm is used to intelligently analyze the category of hazardous chemicals based on image data, reducing the manual review process, thereby improving the inspection efficiency of hazardous chemicals.

[0017] In a second aspect of the present application, a hazardous chemicals inspection system is provided. The system comprises: Image acquisition terminal, used to collect image data of hazardous chemicals; A server, comprising a data generating module and a data determining module; The data generation module is used for the server to generate a label image to be tested according to the image data, and the label image to be tested is an image of a local area where the hazardous chemicals are located; The data determination module is used to obtain a target label image based on the label image to be tested and the standard label image, wherein the standard label image is a label image obtained by pre-training based on the pictogram of the hazardous chemicals, and the target label image is a label image used to indicate the category to which the hazardous chemicals belong.

[0018] In a third aspect of the present application, a hazardous chemical inspection device is provided, which includes a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the program, any of the above-mentioned hazardous chemical inspection methods is implemented.

[0019] In a fourth aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, any of the above-mentioned hazardous chemicals inspection methods is implemented.

[0020] In summary, the present application includes at least one of the following beneficial technical effects: First, the inspectors use the image acquisition terminal to collect image data of hazardous chemicals, freeing the inspectors' hands while ensuring their personal safety. Then, the server receives the image data, calls the pre-trained AI recognition algorithm, and uses the pre-trained AI recognition algorithm to intelligently detect the local area image where the hazardous chemicals are located in the image data, thereby obtaining the label image to be tested, and then matching the corresponding standard label image based on the obtained label image to be tested. Specifically, during matching, multiple features to be tested, such as text, color, contour, texture, shape, etc. in the label image to be tested are used as the basis for selecting the target label image, thereby ensuring the accuracy of the selected target label image, and then ensuring the accuracy of the category to which the analyzed label image belongs. Since the matching process is obtained through intelligent analysis by the AI ​​recognition algorithm, the manual review process is reduced, thereby improving the inspection efficiency of hazardous chemicals. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a schematic diagram of an exemplary operating environment of an embodiment of the present application.

[0022] Figure 2 It is a flow chart of the hazardous chemicals inspection method of an embodiment of the present application.

[0023] Figure 3 It is an example diagram of the pictogram of hazardous chemicals in the embodiment of the method of the present application.

[0024] Figure 4 It is an example diagram of a label diagram preset in advance in the embodiment of the method of the present application.

[0025] Figure 5 It is a block diagram of a hazardous chemicals inspection system according to an embodiment of the present application.

[0026] Explanation of the accompanying drawings: 10, image acquisition terminal; 20, server; 21, data generation module; 22, data determination module. DETAILED DESCRIPTION

[0027] This specific embodiment is merely an explanation of the present application and is not a limitation of the present application. After reading this specification, those skilled in the art may make modifications to the present embodiment without any creative contribution as needed. However, as long as it is within the scope of the claims of the present application, it shall be protected by the patent law.

[0028] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0029] Figure 1 A schematic diagram of an exemplary operating environment in which an embodiment of the present application can be implemented is shown, in which an image acquisition terminal 10 and a server 20 are provided, and the image acquisition terminal 10 and the server 20 communicate with each other via a wired connection or a wireless connection, such as a cable connection, or a wireless connection such as wifi, Bluetooth, 4G or 5G network.

[0030] The image acquisition terminal 10 is a terminal device such as AR glasses, mobile phones, tablets, cameras, etc. that has the function of collecting and uploading images. This example uses AR glasses. AR glasses are commonly used smart glasses on the market that can support shooting or recording by voice, blinking, pressing buttons, etc. In addition to the recording and shooting functions, AR glasses can also quickly locate the position of hazardous chemicals from a first-person perspective, and can provide a first-person perspective for non-persons to observe the hazardous chemicals inspection site, freeing the hands of customs and enterprise inspection personnel and ensuring the personal safety of inspection personnel.

[0031] The server 20 can be integrated with the image acquisition terminal 10, and the server 20 at this time is a micro server. The server 20 can also be independently set outside the image acquisition terminal 10 to receive images or videos uploaded by the image acquisition terminal 10. Specifically, the server 20 is based on the CPU, which supports the operation of the AI ​​recognition algorithm. The AI ​​recognition algorithm is a target detection algorithm based on deep learning, such as the YOLOV5 network model or Faster RCNN in the YOLO series. Among them, YOLOv5 uses CSPDarknet53 as the backbone network and introduces an adaptive training strategy, which can automatically adjust the parameters of the network model according to the hardware resources of the server 20 to improve the generalization ability of the network model. Faster RCNN shares convolutional features with the detection network through a fully convolutional network (region proposal network, RPN) to achieve the purpose of efficient operation. Specifically, RPN can simultaneously predict the score of the identification box and each identification box. The identification box is used to mark the feature information in the image, so as to achieve the purpose of efficiently selecting the regional image within the high-quality boundary box.

[0032] It should be noted that Figure 1 The operating environment shown is only illustrative and is not intended to limit the application or use of the embodiments of the present application. For example, the operating environment may include multiple image acquisition terminals 10 and multiple servers 20, each server 20 is used to interact with one or more image acquisition terminals 10.

[0033] Figure 2 A flow chart of a hazardous chemicals inspection method according to an embodiment of the present application is shown, and the main process of the method is described as follows.

[0034] Step S100: The image acquisition terminal 10 acquires image data of hazardous chemicals.

[0035] In a scenario where hazardous chemicals need to be inspected, the inspector wears AR glasses, and after observing the location of the hazardous chemicals, controls the AR glasses to image the hazardous chemicals, or records the location of the hazardous chemicals to obtain video data, by voice, blinking, or pressing a button. Finally, the AR glasses use the captured image or recorded video data as image data, and upload the image data to the server 20.

[0036] Step S200: The server 20 generates a label image to be tested based on the image data. The label image to be tested is an image of a local area where the hazardous chemicals are located.

[0037] It can be seen from the embodiment of the operating environment that an AI recognition algorithm is set in the server 20, and the AI ​​recognition algorithm is one of the YOLOV5 network model and Faster RCNN. This example takes the YOLOV5 network model as an example for explanation.

[0038] Before using the YOLOV5 network model, you need to train the YOLOV5 network model first. The following is the training process.

[0039] First, collect pictograms to make training samples. First, collect pictograms of hazardous chemicals from the Internet or expert database. Currently, there are nine commonly used pictograms, namely explosive pictograms, flammable pictograms, compressed gas pictograms, corrosive pictograms, toxic pictograms, health pictograms, irritant pictograms, and environmental pictograms. Figure 3 As shown in the figure, the pictograms of the nine hazardous chemicals all have prismatic borders and a white background. Then, the collected pictograms are inserted into the pre-set label image. In a specific example, the format of the label image is as follows: Figure 4 As shown, the label includes the product name, the pictogram corresponding to the product name, the product identifier, the signal word, the supplier logo, and the hazard statement precautions. Among them, the product name refers to the name of the hazardous chemical. The pictogram corresponding to the product name is determined according to the chemical properties of the hazardous chemical. For example, the chemical property of liquid ammonia is toxic, and the pictogram corresponding to liquid ammonia is a toxic pictogram. The product identifier refers to the general term for various expressions and indications used to indicate hazardous chemicals. The product identifier is mainly manifested as the origin of the hazardous chemical, the name of the manufacturer, the address, the main ingredients, the specifications and models, etc. The product identifier can be marked on the hazardous chemical or on the packaging of the hazardous chemical. The signal word is the keyword involved in the hazardous chemical, the supplier logo is the logo of the manufacturer that produces the hazardous chemical or the logo of the source manufacturer of the hazardous chemical, and the hazard statement precautions are used to explain the precautions when using the hazardous chemical or the precautions when storing it. Of course, the label can also be customized to add matters related to hazardous chemicals as needed, such as adding a shelf life or deleting the signal word item.

[0040] In order to increase the richness of training samples, and because hazardous chemicals may have multiple chemical properties, multiple pictograms are combined and then inserted into the label image, so that one product name can correspond to multiple pictograms. For example, the flammable pictogram and the irritant pictogram are combined into a group of images, and then the group of images is inserted into the pictogram frame of the label image, and each pictogram occupies a pictogram frame. For another example, after the flammable pictogram and the irritant pictogram are combined into a group of images, a pictogram is selected from the explosive pictogram, compressed gas pictogram, corrosive pictogram, toxic pictogram, health hazard pictogram, and environmental hazard pictogram to reorganize with the group of images, so as to obtain a new image composed of three pictograms, and finally the new group of images is inserted into the label image.

[0041] For the label map of the inserted pictogram, the LabelImg tool is used to annotate the data and create a file containing nine types of identifiers. The format of the file is classes,txt. The nine identifiers are represented by baozhaowu, huoyan, yanghuaji, yasuoqiti, fushiwu, duxing, weihaijiankang, cijixing, and huanjing, respectively. Finally, the training samples of pictograms with annotated label identifiers are obtained.

[0042] Then, the YOLOV5 network model is trained using the training samples, and the number of iterations or the loss function is set. When the number of training times reaches the set number of iterations or the value output by the loss function meets the set value, the YOLOV5 network model completes the training.

[0043] Based on the YOLOV5 network model trained in advance, when the server 20 receives the image data uploaded by the image acquisition terminal 10, the YOLOV5 network model is called up, and the YOLOV5 network model generates a label map to be tested according to the image data. The specific processing flow is shown in steps S210 to S220.

[0044] Step S210: extract the area where the hazardous chemicals are located in the image data as an intermediate label image.

[0045] First, parse the image data. Since the image data may be an image or a video, after obtaining the image data, it is necessary to parse the image data first.

[0046] Then, determine whether the image data is an image. If it is an image, the local area image where the hazardous chemicals are located in the image is used as the intermediate label image, and the images in other areas are considered invalid areas and deleted. If the image data contains a video, the most representative video frame in the video stream is extracted as the image of the hazardous chemical, and then the local area image where the hazardous chemicals are located in the image is used as the intermediate label image, and the images in other areas are considered invalid areas and deleted. The image recognition technology used to identify the local area where the hazardous chemicals are located is established in advance and integrated in the YOLOV5 network model. Similarly, the selection of the most representative video frame from the video stream is also established in advance and integrated in the YOLOV5 network model, such as first using YOLOV5 to identify the video frame with the largest amount of information, and then using the frame filtering technology to remove other video frames, and only retaining the video frame with the largest amount of information.

[0047] Step S220: preprocess the intermediate label image to obtain a label image to be tested.

[0048] First, the width and height ratio of the intermediate label image is adjusted to the preset ratio, and then the high temperature refinement and background suppression model is used to suppress the interference information in the intermediate label image to obtain the label image to be tested. The high temperature refinement and background suppression model consists of a high temperature refinement sub-model and a background suppression sub-model. The high temperature refinement sub-model allows the model to learn appropriate feature scales and improve the representation ability of various features, while the background suppression sub-model uses the classification confidence to divide the intermediate label image into foreground and background, and suppresses the features in the low confidence area, thereby improving the model's discrimination ability and achieving the effect of suppressing the interference information in the intermediate label image.

[0049] During the preprocessing process, in addition to the above-mentioned preprocessing steps, it can also include normalizing the pixels on the intermediate label image to the same range, or performing data enhancement operations such as random enhancement, random flipping or rotation on the intermediate label image, so as to improve the quality of the obtained label image to be tested.

[0050] In step S300, the server 20 also obtains a target label image based on the label image to be tested and the standard label image. The standard label image is a label image trained in advance based on the pictograms of hazardous chemicals, and the target label image is a label image used to indicate the category to which the hazardous chemicals belong.

[0051] First, the YOLOV5 network model identifies the features to be tested in the label image to be tested. The features to be tested are used to reflect the category to which the label image to be tested belongs. The label image to be tested is an image of the local area where hazardous chemicals are located, so the category to which the label image to be tested belongs is also the category to which hazardous chemicals belong. In a specific example, the features to be tested are information such as text, color, contour, texture, shape, etc. in the label image to be tested. After identifying the features to be tested, the YOLOV5 network model uses an identification frame to mark each feature to be tested, that is, draws an identification frame around each feature to be tested. Therefore, the more features to be tested contained in the label image to be tested, the more corresponding identification frames there are.

[0052] Then, retrieve the standard label map. The standard label map is pre-trained using the trained YOLOV5 network model, and then reviewed and revised by experts in related fields. The standard label map contains standard features such as product name, pictogram corresponding to the product name, product identifier, signal word, supplier logo, hazard statement precautions, etc. Other standard features can also be customized according to actual needs.

[0053] Calculate the similarity between the label map to be tested and the standard label map. Specifically, calculate the confidence between the feature to be tested and the standard feature in each identification frame. Taking the feature to be tested as text as an example, if the semantics of the text in the identification frame is the same as or closer to the semantics of the text on the standard label map, the higher the confidence of the two. Then select the standard feature with the highest confidence in the feature to be tested as the target feature of the identification frame, and other standard features are eliminated to reduce false detection. At the same time, the coordinates of the identification frame in which the target feature has been determined will be output, and the identification frame will be decrypted and converted into the actual coordinates on the label map to be tested to bind the identification frame that determines the target feature and the target feature to prevent repeated calculations. It can be seen that each identification frame corresponds to a target feature, and the more target features are contained in the standard label map, the higher the similarity with the label map to be tested. Conversely, when the number of target features contained in the standard label map is zero, the similarity between the standard label map and the label map to be tested is considered to be zero.

[0054] Finally, the standard label image with the highest similarity to the label image to be tested is selected as the target label image. The product name on the target label image is the name of the hazardous chemical, and the pictogram on the target label image is the chemical property of the hazardous chemical, which also represents the category to which the hazardous chemical belongs. It should be noted that when there are multiple standard label images with the same number of target features, that is, when the similarity between the label image to be tested and multiple standard label images is the same, the confidence value of each standard label image with the same similarity is calculated. The confidence value here is the sum of the confidence values ​​between the feature to be tested in each identification box on the label image to be tested and the target feature corresponding to the feature to be tested. The standard label image with the largest confidence value is then used as the target label image.

[0055] The server 20 combines the obtained target label image, the confidence values ​​of the target label image and the label image to be tested, and the position coordinates of the label image to be tested into the inspection result data, and saves the inspection result data into the memory. At the same time, the inspection result data is also returned to the AR glasses. The AR glasses retrieve the target label image based on the location of the label image to be tested and superimpose it on the real scene, so that the inspectors can quickly verify whether the hazardous chemicals are consistent with the hazardous chemicals declared by the enterprise, so as to shorten the time for manual inspection of hazardous chemicals by the customs, shorten the customs clearance time for imported and exported hazardous chemicals, and greatly reduce the manpower inspection cost of customs inspectors. At the same time, the shortening of customs clearance time reduces the storage cost of enterprises when detaining hazardous chemicals, greatly reduces the operating cost of the enterprise side, and reduces the management risk of hazardous chemicals at the enterprise side.

[0056] In addition, after receiving the inspection result data, the AR glasses can also retrieve the logo corresponding to the target label image, which is one or more of the above-mentioned baozhaowu, huoyan, yanghuaji, yasuoqiti, fushiwu, duxing, weihaijiankang, cijixing, and huanjing, so that the inspectors can know the logo of the hazardous chemicals they are inspecting.

[0057] In summary, the implementation principle of a hazardous chemicals inspection method in the embodiment of the present application is as follows: First, AR glasses are used to collect image data of hazardous chemicals for inspectors, freeing the hands of inspectors while ensuring the personal safety of inspectors. Then, the server 20 receives the image data, and intelligently detects the local area image where the hazardous chemicals are located in the image data by calling the AI ​​recognition algorithm trained in advance, and matches the corresponding standard label image for the area image, so as to achieve the purpose of intelligently classifying hazardous chemicals, thereby improving the inspection efficiency of hazardous chemicals.

[0058] Figure 5A block diagram of a hazardous chemicals inspection system according to an embodiment of the present application is shown. The system includes the above-mentioned AR glasses and a server 20, and the server 20 can be divided into a data generation module 21 and a data determination module 22.

[0059] The data generation module 21 is used to generate a label image to be tested based on the image data, where the label image to be tested is an image of a local area where the hazardous chemicals are located.

[0060] The data determination module 22 is used for the server 20 to obtain a target label image based on the label image to be tested and the standard label image. The standard label image is a label image trained in advance based on the pictograms of hazardous chemicals, and the target label image is a label image used to indicate the category to which the hazardous chemicals belong.

[0061] The modules involved in the embodiments of the present application may be implemented by software or hardware. The modules described may also be set in a processor, for example, it may be described as: a processor includes a data generation module 21 and a data determination module 22. The names of these modules do not constitute a limitation on the modules themselves in some cases, for example, the data generation module 21 may also be described as "a module for generating a label image to be tested based on image data".

[0062] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the described module can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0063] In order to better execute the procedures of the above method, the present application also provides a hazardous chemicals inspection device, which includes a memory and a processor.

[0064] The memory may be used to store instructions, programs, codes, code sets or instruction sets. The memory may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function, and instructions for implementing the above-mentioned hazardous chemicals inspection method, etc.; the data storage area may store data involved in the above-mentioned hazardous chemicals inspection method, etc.

[0065] The processor may include one or more processing cores. The processor calls the data stored in the memory by running or executing the instructions, programs, code sets or instruction sets stored in the memory, performs various functions of the present application and processes data. The processor may be at least one of a special purpose integrated circuit, a digital signal processor, a digital signal processing device, a programmable logic device, a field programmable gate array, a central processing unit, a controller, a microcontroller and a microprocessor. It is understandable that for different devices, the electronic device used to implement the above-mentioned processor function can also be other, and the embodiments of the present application are not specifically limited.

[0066] The present application also provides a computer-readable storage medium, for example, including: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes. The computer-readable storage medium stores a computer program that can be loaded by a processor and execute the above-mentioned hazardous chemicals inspection method.

[0067] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the aforementioned disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present application (but not limited to) by each other to form a technical solution.

Claims

1. A method for inspecting hazardous chemicals, characterized in that: include: The image acquisition terminal (10) acquires image data of hazardous chemicals; The server (20) generates a label image to be tested based on the image data, wherein the label image to be tested is an image of a local area where the hazardous chemicals are located; The server (20) also obtains a target label image based on the label image to be tested and the standard label image, wherein the standard label image is a label image obtained by pre-training based on the pictogram of the hazardous chemical, and the target label image is a label image used to indicate the category to which the hazardous chemical belongs.

2. The method for inspecting hazardous chemicals according to claim 1, characterized in that: The server (20) further obtains a target label map according to the label map to be tested and the standard label map, including: Identify a feature to be tested in the label image to be tested, where the feature to be tested is used to reflect the category to which the label image to be tested belongs; Calculating the similarity between the label image to be tested and the standard label image based on the feature to be tested; The standard label image with the highest similarity to the label image to be tested is selected as the target label image.

3. The method for inspecting hazardous chemicals according to claim 2, characterized in that: The feature to be measured is located in the identification frame; The calculating the similarity between the label image to be tested and the standard label image according to the feature to be tested includes: Calculating the confidence between the feature to be tested and the standard feature in each identification frame, wherein the standard feature is used to reflect the category to which the standard label image belongs; Selecting a standard feature with the highest confidence level with the feature to be measured as a target feature of the identification frame; The number of target features contained in the standard label image is proportional to the similarity between the label image to be tested and the standard label image.

4. The method for inspecting hazardous chemicals according to claim 3, characterized in that: The features to be measured include text, color, outline, texture, and shape.

5. The method for inspecting hazardous chemicals according to claim 3, characterized in that: The method further comprises: When the number of target features contained in multiple standard label images is the same: Calculating the confidence value of each of the plurality of standard label images, the confidence value being the sum of the confidences between the feature to be measured in each identification frame and the target feature corresponding to the feature to be measured on the label image to be measured; A standard label image with the largest confidence value is used as the target label image.

6. The method for inspecting hazardous chemicals according to claim 1, characterized in that: The server (20) generates a label image to be tested according to the image data, including: Using the local area image where the hazardous chemicals are located in the image data as the intermediate label image; The width-to-height ratio of the intermediate label image is adjusted to a preset ratio, and the interference information in the intermediate label image is suppressed by using a high temperature refinement and background suppression model to obtain the label image to be tested.

7. The method for inspecting hazardous chemicals according to claim 1, characterized in that: Before the server (20) generates a label image to be tested according to the image data, the method further comprises: Collect pictograms of various hazardous chemicals to create training samples; Using the training samples to train the AI ​​recognition algorithm; The trained AI recognition algorithm is used to generate a label map to be tested based on the image data.

8. A hazardous chemicals inspection system, characterized in that: include: An image acquisition terminal (10) is used to acquire image data of hazardous chemicals; A server (20), comprising a data generating module (21) and a data determining module (22); The data generation module (21) is used for the server (20) to generate a label image to be tested according to the image data, wherein the label image to be tested is an image of a local area where the hazardous chemicals are located; The data determination module (22) is used to obtain a target label image based on the label image to be tested and a standard label image, wherein the standard label image is a label image obtained by pre-training based on the pictogram of the hazardous chemical, and the target label image is a label image used to indicate the category to which the hazardous chemical belongs.

9. A hazardous chemicals inspection device, characterized in that: The method comprises a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images