CN115861688A - Drug packaging and container appearance modeling identification and counting method and system - Google Patents

Abstract

The invention discloses a method and a system for modeling, identifying and counting the shapes of medicine packages and containers, wherein the method comprises the following steps: the method comprises the steps of collecting packaging container shape image data of N toxic and anesthetic drugs in a preset image collection mode, setting an identification mode of the packaging container image of the toxic and anesthetic drugs, learning and modeling the packaging container shape image data of the N toxic and anesthetic drugs through a YOLOv5 machine learning algorithm based on the identification mode, obtaining an identification model, and loading the identification model into a terminal application program to perform batch identification, counting and recovery on the toxic and anesthetic drugs. The image data of the toxic and anesthetic glass containers and the recognition modes set for the image data are utilized to carry out recognition, learning and modeling on the image data by utilizing a machine learning algorithm so as to obtain recognition models to automatically carry out recognition technology recovery on the toxic and anesthetic glass containers in batches, so that the recovery efficiency is improved, and the time cost is greatly saved.

Description

A pharmaceutical packaging and container shape modeling recognition and counting method and system

technical field

The invention relates to the technical field of information processing, in particular to a drug packaging and container shape modeling recognition and counting method and system.

Background technique

Narcotic drugs are controlled drugs, manual management is time-consuming and labor-intensive, and intelligent medicine cabinet management is the trend, but precise recycling of empty containers is a key link. It is necessary to identify the type and number of containers, fixed-point delivery of empty containers, and centralized removal. Currently, the main methods for identifying recycling containers The method is the OCR code identification method on the container. The above method can only be applied to a single recycling of the container, which affects the work efficiency during the peak medication period. It needs to be processed separately and cannot be widely used in clinical work, which reduces the recycling efficiency.

Contents of the invention

In view of the problems shown above, the present invention provides a drug packaging and container shape modeling recognition and counting method and system to solve the problem that the traditional technology mentioned in the background technology can only be applied to a single recovery of the container, which affects peak drug use During the work efficiency, it needs to be processed separately and cannot be widely used in clinical work, which reduces the recovery efficiency.

A pharmaceutical packaging and container shape modeling recognition and counting method, comprising the following steps:

Collect N pieces of image data of the packaging container shape of narcotic drugs through a preset image collection method;

Set the recognition mode of the packaging container image of narcotic drugs;

Carry out learning and modeling on the packaging container outline image data of the N pieces of narcotics and narcotic drugs through the YOLOv5 machine learning algorithm based on the recognition pattern, and obtain the recognition model;

The recognition model is loaded into the terminal application program to carry out batch recognition, counting and recovery of narcotics and narcotics.

Preferably, the acquisition of N pieces of packaging container shape image data of narcotics and narcotics through a preset image acquisition method includes:

Setting image acquisition conditions according to the preset image acquisition mode, the image acquisition conditions include: lighting conditions and ground conditions;

Setting the shooting angle and focal length parameters of the camera according to the image acquisition conditions;

Detect the arrangement of the packaging containers of narcotic drugs in the preset arrangement sequence;

According to the arrangement, it is judged whether the requirements for image acquisition are met, and if so, the camera is controlled by the preset image acquisition mode to set parameters to acquire N pieces of image data of the packaging containers of narcotics and narcotics.

Preferably, said setting the recognition mode of the packaging container image of narcotic drugs includes:

Obtain the bottleneck parameters, visual parameters and state parameters of the solution or powder inside the container for narcotics and narcotics packaging containers;

Obtain the label pattern recognition parameters and shape recognition parameters of the narcotics packaging container according to the visual parameters; Drug use state identification parameters;

Determining various display states of the packaging container of narcotic drugs based on all identification parameters;

A recognition rule corresponding to each display state is constructed and a recognition pattern corresponding to the display state is generated.

Preferably, before learning and modeling the outer shape image data of the packaging containers of the N drugs and narcotics through the YOLOv5 machine learning algorithm based on the recognition pattern, and obtaining the recognition model, it also includes:

Obtain the model parameters of the preset YOLOv4 machine learning model;

Screen out model input terminal parameters, benchmark network parameters, Neck network parameters and Head output layer parameters in the model parameters;

Optimizing the model input parameters, reference network parameters, Neck network parameters and Head output layer parameters;

Obtain the YOLOv5 machine learning algorithm based on the optimized YOLOv4 machine learning model.

Preferably, based on the recognition pattern, the YOLOv5 machine learning algorithm is used to learn and model the shape image data of the packaging container of the N drugs and narcotics, and obtain a recognition model, including:

Divide N images of the packaging container shape of narcotic drugs into M training images and N-M verification images and generate a data training set and a data verification set respectively;

Divide the M training images into a plurality of training data groups according to the preset data load quantity of each training group, and set the maximum number of single batch processing data, hyperparameters and resolution of the training data group;

Writing the YOLOv5 machine learning algorithm and the recognition pattern into the preset neural network model, utilizing the preset neural network model to carry out six recognition trainings to the training images in a plurality of training data sets;

Perform error analysis on the output data of the preset neural network model after six trainings, determine whether the training model is qualified according to the analysis results, and if so, confirm it as the recognition model.

Preferably, the loading of the identification model into the terminal application program to perform batch identification, counting and recycling of narcotics includes:

Encoding the identification model to obtain an encoding result, and writing the encoding result into the terminal application program to generate a narcotics identification program;

Collect the target image of the narcotics packaging container to be identified, and use the narcotics identification program to identify the target image to determine the number of containers, the type of the container, and the use status of the container;

Classify and recycle the packaging containers of narcotic drugs in batches according to the number and type of containers and the state of use of the containers;

Add the counting identification information of narcotics and narcotics packaging containers to the filing system to match the information of the person who takes the drug and the person who returns it;

If it does not match, an information mismatch prompt is generated to inform the medication doctor and the drug administrator through the preset information platform.

Preferably, the error analysis is performed on the output data of the preset neural network model after the six trainings, and whether the training model is determined to be qualified according to the analysis results, if so, is confirmed as a recognition model, including:

Inputting the verification image in the data verification set to the trained preset neural network model for recognition, and obtaining the recognition result;

Determine the recognition error probability of the trained preset neural network model according to the recognition result, and analyze the recognition error of the trained preset neural network model according to the recognition error probability;

Confirm whether the recognition error is within the preset error threshold range, if yes, determine that the training model is qualified, if not, determine that the training model is unqualified;

When it is confirmed that the training model is qualified, it is confirmed as the recognition model.

Preferably, the method also includes:

Obtaining the model parameters of the recognition model, and obtaining the mean value of the loss function, the mean value of the detection loss, the mean value of the classification loss, the precision value and the threshold value of the recall rate of the recognition model according to the model parameters;

Determine the detection accuracy of the recognition model according to the loss function mean value and the detection loss mean value;

determining the classification accuracy of the recognition model based on the classification loss mean value;

The recognition accuracy of the recognition model is determined according to the precision value and the recall rate threshold, and whether the recognition model needs to continue training is determined based on the detection precision, classification precision and recognition precision.

Preferably, the batch sorting and recycling of the narcotics and narcotics packaging containers according to the number of containers, the type of containers and the state of use of the containers includes:

Divide the packaging containers of narcotics and narcotics to be identified according to the type of the container, and obtain the first division result;

Carry out used empty bottles, unused half bottles, and unused full bottles for each of the poisonous and narcotic drug packaging containers in the first classification result according to the use state of the container, and obtain the second classification result;

According to the first classification result and the second classification result, an adapted recovery mode is selected to perform batch recovery processing on the packaging containers of narcotics and narcotics.

A pharmaceutical packaging and container shape modeling recognition and counting system, the system includes:

The collection module is used to collect N pieces of packaging container shape image data of narcotic drugs through a preset image collection method;

The setting module is used to set the recognition mode of the packaging container image of narcotic drugs;

A modeling module, for learning and modeling the packaging container outline image data of the N pieces of narcotics and narcotics through the YOLOv5 machine learning algorithm based on the recognition pattern, to obtain a recognition model;

The loading module is used to load the recognition model into the terminal application program to perform batch recognition, counting and recovery of narcotics and narcotics.

Additional features and advantages of the invention will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention may be realized and attained by the structure particularly pointed out in the written description and appended drawings.

The technical solutions of the present invention will be described in further detail below with reference to the accompanying drawings and embodiments.

Description of drawings

The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the description, and are used together with the embodiments of the present invention to explain the present invention, and do not constitute a limitation to the present invention.

Fig. 1 is a working flow chart of a drug packaging and container shape modeling recognition and counting method provided by the present invention;

Fig. 2 is another work flow chart of a kind of pharmaceutical packaging and container shape modeling identification and counting method provided by the present invention;

Fig. 3 is yet another work flow chart of a kind of pharmaceutical packaging and container shape modeling identification and counting method provided by the present invention;

Fig. 4 is a structural schematic diagram of a pharmaceutical packaging and container shape modeling recognition and counting system provided by the present invention.

Detailed ways

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numerals in different drawings refer to the same or similar elements unless otherwise indicated. The implementations described in the following exemplary examples do not represent all implementations consistent with the present disclosure. Rather, they are merely examples of apparatuses and methods consistent with aspects of the present disclosure as recited in the appended claims.

Narcotic drugs are controlled drugs, manual management is time-consuming and labor-intensive, and intelligent medicine cabinet management is the trend, but precise recycling of empty containers is a key link. It is necessary to identify the type and number of containers, fixed-point delivery of empty containers, and centralized removal. Currently, the main methods for identifying recycling containers The method is the OCR code identification method on the container. The above method can only be applied to a single recycling of the container, which affects the work efficiency during the peak medication period. It needs to be processed separately and cannot be widely used in clinical work, which reduces the recycling efficiency. In order to solve the above-mentioned problems, the present embodiment discloses a method for identifying and counting pharmaceutical packaging and container shape modeling.

A pharmaceutical packaging and container shape modeling recognition and counting method, as shown in Figure 1, comprises the following steps:

Step S101, collecting N pieces of image data of the packaging containers of narcotics and narcotics through a preset image collection method;

Step S102, setting the recognition mode of the packaging container image of narcotic drugs;

Step S103, based on the recognition pattern, use the YOLOv5 machine learning algorithm to learn and model the shape image data of the N packages of narcotics and narcotics, and obtain a recognition model;

Step S104, loading the recognition model into the terminal application program to perform batch recognition, counting and recycling of narcotics.

In this embodiment, the preset image acquisition method may be scanning, shooting, transmission, uploading, etc.;

In this embodiment, the recognition mode includes appearance recognition, shape recognition, color recognition and state recognition of the packaging container of narcotic drugs;

In this embodiment, the loading method may be loaded into the terminal application in the form of a small program or in the form of an application.

The working principle of the above-mentioned technical solution is as follows: collect N pieces of packaging container shape image data of narcotics and narcotics through a preset image collection method, set the recognition mode of the packaging container image of narcotics and narcotics, based on the recognition mode, use the YOLOv5 machine learning algorithm to The shape image data of the packaging containers of the N pieces of narcotics and narcotics are learned and modeled to obtain a recognition model, and the recognition model is loaded into the terminal application program to perform batch recognition, counting and recovery of narcotics and narcotics.

The beneficial effect of the above technical solution is: by using the image data of the glass container of narcotics and narcotics and the recognition mode set for it, the machine learning algorithm is used to carry out recognition and learning modeling on the image data, and then the recognition model is obtained to automatically identify batches of narcotics and narcotics. Recycling of drug glass containers with identification technology improves recycling efficiency and greatly saves time and cost. It solves the problem that traditional technology can only be applied to a single container recycling, which affects work efficiency during peak drug use. It needs to be processed separately and cannot be widely used in clinical work. , reducing the problem of recovery efficiency.

In one embodiment, as shown in FIG. 2, the acquisition of N pieces of packaging container shape image data of narcotics and narcotics through a preset image acquisition method includes:

Step S201, setting image acquisition conditions according to the preset image acquisition mode, the image acquisition conditions include: lighting conditions and ground conditions;

Step S202, setting the shooting angle and focal length parameters of the camera according to the image acquisition conditions;

Step S203, detecting the arrangement of the packaging containers of narcotic drugs in the preset arrangement sequence;

Step S204 , judging whether the image collection requirements are met according to the arrangement, and if so, control the camera through the preset image collection method to set parameters to collect N pieces of image data of the packaging containers of narcotics and narcotics.

In this embodiment, the arrangement status represents the correct placement of the packaging containers of narcotics and narcotics in the shooting arrangement sequence set by the camera.

The beneficial effects of the above technical solution are: by setting the image acquisition conditions, the image acquisition work can be realized more clearly and completely, and the work efficiency and stability are improved; further, by setting the shooting angle and focal length parameters of the camera, the acquisition can be further ensured The clarity of the image provides high-quality training samples for subsequent model training and improves the learning accuracy of the model during the training process.

In one embodiment, the setting of the recognition mode of the packaging container image of narcotic drugs includes:

Obtain the bottleneck parameters, visual parameters and state parameters of the solution or powder inside the container for narcotics and narcotics packaging containers;

Obtain the label pattern recognition parameters and shape recognition parameters of the narcotics packaging container according to the visual parameters; Drug use state identification parameters;

Determining various display states of the packaging container of narcotic drugs based on all identification parameters;

A recognition rule corresponding to each display state is constructed and a recognition pattern corresponding to the display state is generated.

In this embodiment, the bottleneck parameter is expressed as a state parameter when the bottleneck is complete or broken;

In this embodiment, the visual parameter is expressed as the visual parameter of the shape of the drug packaging container;

In this embodiment, the display state is represented as a drug carrying display state and an appearance display state of the packaging container of narcotics and narcotics.

The beneficial effect of the above technical solution is: by determining multiple parameters of the narcotic drug packaging container to determine its multiple display states and then generate a recognition pattern that can take all display forms of the narcotic drug packaging into consideration, so as to achieve no real-time sample analysis. Missed recognition further improves the recognition efficiency and stability.

In this embodiment, a recognition rule corresponding to each display state is constructed and a recognition pattern corresponding to the display state is generated, including:

Determine the viewing angle attributes of the narcotics packaging container in each display state;

Determining the descriptive factor of the narcotics packaging container in each display state based on the viewing angle attribute;

Constructing a recognition model of the narcotics packaging container in each display state through the description factors;

According to the identification model, the identification parameters of the drug packaging container in each display state are obtained;

Arranging and combining the identification parameters to generate M identification strategies, setting N preset identification rules, where M=N;

Fusing the preset recognition rules and recognition strategies to obtain N recognition strategy rules;

Extracting rule parameters and policy parameters of N identification policy rules and constructing a calculation model for each identification policy rule based on the two;

The convolutional neural network is used to perform adaptive learning processing on the identification parameters of the drug packaging container in each display state, and determine the corresponding calculation parameters of the calculation model of each identification strategy rule according to the processing results;

Generate a calculation expression for the identification policy rule according to the calculation parameters and the calculation model corresponding to the calculation model of each identification policy rule;

Execute the calculation expression of each identification strategy rule based on the identification parameters of the narcotics packaging container in each display state, and select the best identification strategy rule for the narcotics packaging container in each display state according to the execution result;

Determining the recognition loss difference of the drug packaging container in each display state under the optimal recognition strategy rule, and creating a recognition dictionary according to the recognition loss difference;

Obtain the constant matrix, embedding matrix and label matrix of the recognition dictionary;

According to the constant matrix, embedding matrix and label matrix, the corresponding recognition mode of the drug packaging container in each display state is set.

In this embodiment, the viewing angle attribute is expressed as a multi-viewing angle display attribute factor in each display state of the drug packaging container;

In this embodiment, the description factor is expressed as the description of the appearance of the narcotics packaging container in each display state;

In this embodiment, the identification parameters are expressed as shape identification parameters, color identification parameters, completeness identification parameters and state identification parameters of the drug packaging container in each display state;

In this embodiment, the recognition loss difference is expressed as the recognition error of the drug packaging container in each display state under the optimal recognition strategy rule or the ratio of the recognition image loss local area to the entire area;

In this embodiment, the recognition dictionary is represented as a network dictionary that performs recognition compensation for recognition loss differences.

The beneficial effect of the above technical solution is: by constructing the identification model of the narcotics packaging container in each display state, the identification parameters of the narcotics packaging container in each display state can be quickly obtained, and according to its different strategy rules The test results to select the best recognition strategy rules, making the screening rules more objective and practical, further, by creating a recognition dictionary and then setting the recognition mode according to the recognition loss difference of the screening recognition strategy rules, it can overcome the problems caused by the recognition rules. The error or loss is used to carry out more accurate and stable identification work on the packaging containers of narcotics and narcotics, which improves the stability while ensuring the identification accuracy and improving the practicability.

In one embodiment, before learning and modeling the shape image data of the packaging container of the N drugs and narcotics through the YOLOv5 machine learning algorithm based on the recognition pattern, and obtaining the recognition model, it also includes:

Obtain the model parameters of the preset YOLOv4 machine learning model;

Screen out model input terminal parameters, benchmark network parameters, Neck network parameters and Head output layer parameters in the model parameters;

Optimizing the model input parameters, reference network parameters, Neck network parameters and Head output layer parameters;

Obtain the YOLOv5 machine learning algorithm based on the optimized YOLOv4 machine learning model.

The beneficial effect of the above-mentioned technical solution is: by optimizing the model parameters, it can ensure that the recognition accuracy and recognition speed of the machine learning algorithm can be greatly improved, and the recognition efficiency and practicability are further improved.

In one embodiment, as shown in FIG. 3 , the N pieces of drug packaging container outline image data are learned and modeled based on the recognition pattern through the YOLOv5 machine learning algorithm, and the recognition model is obtained, including:

Step S301. Divide the N images of the packaging containers of narcotics and anesthetics into M training images and N-M verification images, and generate a data training set and a data verification set respectively;

Step S302, divide the M training images into multiple training data groups according to the preset data load quantity of each training group, and set the maximum number of single batch processing data, hyperparameters and resolution of the training data group;

Step S303, writing the YOLOv5 machine learning algorithm and the recognition pattern into the preset neural network model, and using the preset neural network model to perform six recognition trainings on the training images in the multiple training data sets;

Step S304 , analyzing the error of the output data of the preset neural network model after six times of training, and determining whether the training model is qualified according to the analysis result, and if so, confirming it as the recognition model.

In this embodiment, the hyperparameter can be a non-zero positive integer;

In this embodiment, the six identification trainings are to ensure the optimal training times for the preset neural network model to achieve convergence.

The beneficial effect of the above technical solution is: the model can be trained and verified at the same time by generating the data training set and the data verification set, which further improves the practicability; further, by dividing the training array and the batch of image data Processing parameters can enable the model to perform orderly training of data, further improving stability.

In one embodiment, the loading of the identification model into the terminal application program to perform batch identification, counting and recycling of narcotics includes:

Encoding the identification model to obtain an encoding result, and writing the encoding result into the terminal application program to generate a narcotics identification program;

Collect the target image of the narcotics packaging container to be identified, and use the narcotics identification program to identify the target image to determine the number of containers, the type of the container, and the use status of the container;

Classify and recycle the packaging containers of narcotic drugs in batches according to the number and type of containers and the state of use of the containers;

Add the counting identification information of narcotics and narcotics packaging containers to the filing system to match the information of the person who takes the drug and the person who returns it;

If it does not match, an information mismatch prompt is generated to inform the medication doctor and the drug administrator through the preset information platform.

In this embodiment, the narcotics identification program is represented as a narcotics identification application program that can run on a computer or a server terminal.

The beneficial effects of the above-mentioned technical solution are as follows: intelligent identification, classification and recycling of glass containers to be identified can be carried out through terminals such as computers, which improves intelligence and recycling efficiency, saves labor costs, and further improves practicability. Further, through Adding the counting and identification information of the narcotics packaging container to the filing system and matching the information of the person taking and returning the narcotics can record the usage information of each narcotic drug and realize the precise monitoring of the use process of each narcotic drug , further improving the practicality.

In this embodiment, the counting identification information of the narcotics packaging container is added to the filing system, specifically:

Obtaining the node status information of the storage server corresponding to the filing system;

determining idle nodes in the storage server according to the node state information, and calling the remaining storage resources of each idle node;

Determining the memory value required to store count identification information for narcotics packaging containers;

Calculate the recommendation degree of each idle node according to the remaining storage resources of each idle node and the memory value required to store the count identification information of the drug packaging container:

Among them, F _i represents the recommendation degree of the i-th idle node, D _i represents the data transmission efficiency of the i-th idle node, T _1i represents the remaining storage resources of the i-th idle node, and T _2i represents the The total allocated storage resources of idle nodes, α is expressed as the gain factor of the allocated resource ratio of the node to the data transmission efficiency, log is expressed as a logarithm, and p ^' is expressed as the memory value required to store the counting identification information of the drug packaging container, p _i represents the allocated memory value of the i-th idle node, e represents a natural constant with a value of 2.72, a _i represents the processing capability value of the i-th idle node, and B represents the count identification of the drug packaging container The amount of data corresponding to the information, B _i represents the maximum storage data volume of the i-th idle node in a single cycle, D _i represents the transmission delay index of the i-th idle node, V _i represents the current network of the i-th space node Bandwidth, V ^' is expressed as a preset reference bandwidth threshold;

Select the target idle node with the highest recommendation degree as the data transmission node;

The counting identification information of the narcotics and narcotics packaging container is correspondingly stored in the storage server through the target idle node filing system.

The beneficial effect of the above technical solution is: by calculating the recommendation degree of each idle node according to the state parameters and working parameters, it can be based on the digestion and transmission efficiency of each idle node for the counting identification information of the narcotics packaging container and the transmission efficiency. Multiple dimensions such as time delay and transmission capacity are used to comprehensively evaluate the fitness of each idle node, so as to select the idle node with the highest recommendation for data transmission, which not only ensures the efficiency of data transmission, but also ensures the stability of data transmission performance and memory, improving usability.

In one embodiment, the error analysis is performed on the output data of the preset neural network model after six trainings, and whether the training model is determined to be qualified according to the analysis results, if so, is confirmed as a recognition model, including:

Inputting the verification image in the data verification set to the trained preset neural network model for recognition, and obtaining the recognition result;

Determine the recognition error probability of the trained preset neural network model according to the recognition result, and analyze the recognition error of the trained preset neural network model according to the recognition error probability;

Confirm whether the recognition error is within the preset error threshold range, if yes, determine that the training model is qualified, if not, determine that the training model is unqualified;

When it is confirmed that the training model is qualified, it is confirmed as the recognition model.

In this embodiment, the recognition error probability is expressed as the comprehensive probability of the trained preset neural network model identifying label errors, identifying shape errors, and identifying drug state errors in the bottle;

In this embodiment, the recognition error is expressed as the ratio of the number of failed recognitions to the number of successful recognitions of the packaging containers of narcotics and narcotics by the preset neural network model after training.

The beneficial effect of the above technical solution is: by analyzing the recognition error of the training model according to the recognition error probability of the training model, the recognition effect of the training model can be quickly determined, and then it can be selected whether to continue training, which ensures the accuracy and accuracy of the training model. Training effect, further improving stability and practicality.

In one embodiment, the method also includes:

Obtaining the model parameters of the recognition model, and obtaining the mean value of the loss function, the mean value of the detection loss, the mean value of the classification loss, the precision value and the threshold value of the recall rate of the recognition model according to the model parameters;

Determine the detection accuracy of the recognition model according to the loss function mean value and the detection loss mean value;

determining the classification accuracy of the recognition model based on the classification loss mean value;

The recognition accuracy of the recognition model is determined according to the precision value and the recall rate threshold, and whether the recognition model needs to continue training is determined based on the detection precision, classification precision and recognition precision.

The beneficial effects of the above technical solution are: by determining the precision of the recognition model in each functional dimension according to the model parameters, the recognition effect of the model can be further determined, and the precision and recognition accuracy of the model can be further ensured.

In one embodiment, the batch sorting and recycling of the narcotics and narcotics packaging containers according to the number of containers, the types of containers and the state of use of the containers includes:

Divide the packaging containers of narcotics and narcotics to be identified according to the type of the container, and obtain the first division result;

Carry out used empty bottles, unused half bottles, and unused full bottles for each of the poisonous and narcotic drug packaging containers in the first classification result according to the use state of the container, and obtain the second classification result;

According to the first classification result and the second classification result, an adapted recovery mode is selected to perform batch recovery processing on the packaging containers of narcotics and narcotics.

The beneficial effect of the above-mentioned technical scheme is: by dividing the narcotics and narcotics packaging containers twice, all the states of the narcotics and narcotics packaging containers can be determined comprehensively, and then the best recycling mode for each narcotics and narcotics packaging container can be selected, further The recovery efficiency is guaranteed.

This embodiment also discloses a drug packaging and container shape modeling recognition and counting system, as shown in Figure 4, the system includes:

The collection module 401 is used to collect N pieces of packaging container shape image data of narcotic drugs through a preset image collection method;

The setting module 402 is used to set the recognition mode of the packaging container image of narcotic drugs;

The modeling module 403 is used to learn and model the shape image data of the packaging container of the N drugs and narcotics through the YOLOv5 machine learning algorithm based on the recognition pattern, and obtain the recognition model;

The loading module 404 is used to load the recognition model into the terminal application program to perform batch recognition, counting and recovery of narcotics.

The working principles and beneficial effects of the above technical solutions have been described in the method claims and will not be repeated here.

Other embodiments of the present disclosure will be readily apparent to users skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any modification, use or adaptation of the present disclosure, and these modifications, uses or adaptations follow the general principles of the present disclosure and include common knowledge or conventional technical means in the technical field not disclosed in the present disclosure . The specification and examples are to be considered exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

It should be understood that the present disclosure is not limited to the precise constructions which have been described above and shown in the drawings, and various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims (10)

1. A pharmaceutical packaging and container shape modeling identification and counting method, is characterized in that, comprises the following steps: Collect N pieces of image data of the packaging container shape of narcotic drugs through a preset image collection method; Set the recognition mode of the packaging container image of narcotic drugs; Carry out learning and modeling on the packaging container outline image data of the N pieces of narcotics and narcotic drugs through the YOLOv5 machine learning algorithm based on the recognition pattern, and obtain the recognition model; The recognition model is loaded into the terminal application program to carry out batch recognition, counting and recovery of narcotics and narcotics. 2. According to claim 1, the drug packaging and container shape modeling identification and counting method is characterized in that, the collection of N pieces of packaging container shape image data of narcotics and narcotics through a preset image collection method includes: Setting image acquisition conditions according to the preset image acquisition mode, the image acquisition conditions include: lighting conditions and ground conditions; Setting the shooting angle and focal length parameters of the camera according to the image acquisition conditions; Detect the arrangement of the packaging containers of narcotic drugs in the preset arrangement sequence; According to the arrangement, it is judged whether the requirements for image acquisition are met, and if so, the camera is controlled by the preset image acquisition mode to set parameters to acquire N pieces of image data of the packaging containers of narcotics and narcotics. 3. According to the pharmaceutical packaging and container shape modeling identification and counting method according to claim 1, it is characterized in that, the identification mode of the packaging container image of the described setting of narcotic drugs includes: Obtain the bottleneck parameters, visual parameters and state parameters of the solution or powder inside the container for narcotics and narcotics packaging containers; Obtain the label pattern recognition parameters and shape recognition parameters of the narcotics packaging container according to the visual parameters; Drug use state identification parameters; Determining various display states of the packaging container of narcotic drugs based on all identification parameters; A recognition rule corresponding to each display state is constructed and a recognition pattern corresponding to the display state is generated. 4. according to the described drug packaging and container shape modeling identification and counting method of claim 1, it is characterized in that, based on the recognition pattern, the packaging container shape image data of the N sheets of narcotic drugs are processed by the YOLOv5 machine learning algorithm. Learning to model, before acquiring a recognition model, also includes: Obtain the model parameters of the preset YOLOv4 machine learning model; Screen out model input terminal parameters, benchmark network parameters, Neck network parameters and Head output layer parameters in the model parameters; Optimizing the model input parameters, reference network parameters, Neck network parameters and Head output layer parameters; Obtain the YOLOv5 machine learning algorithm based on the optimized YOLOv4 machine learning model. 5. according to the described drug packaging and container shape modeling recognition and counting method of claim 1, it is characterized in that, the described N sheets of drug packaging container shape image data are processed by YOLOv5 machine learning algorithm based on the recognition pattern Carry out learning modeling and obtain recognition models, including: Divide N images of the packaging container shape of narcotic drugs into M training images and N-M verification images and generate a data training set and a data verification set respectively; Divide the M training images into a plurality of training data groups according to the preset data load quantity of each training group, and set the maximum number of single batch processing data, hyperparameters and resolution of the training data group; Writing the YOLOv5 machine learning algorithm and the recognition pattern into the preset neural network model, utilizing the preset neural network model to carry out six recognition trainings to the training images in a plurality of training data groups; Perform error analysis on the output data of the preset neural network model after six trainings, determine whether the training model is qualified according to the analysis results, and if so, confirm it as the recognition model. 6. The drug packaging and container shape modeling identification and counting method according to claim 1, characterized in that, loading the identification model into the terminal application program to perform batch identification, counting and recycling of narcotics comprises: Encoding the identification model to obtain an encoding result, and writing the encoding result into the terminal application program to generate a narcotics identification program; Collect the target image of the narcotics packaging container to be identified, and use the narcotics identification program to identify the target image to determine the number of containers, the type of the container, and the use status of the container; Classify and recycle the packaging containers of narcotic drugs in batches according to the number and type of containers and the state of use of the containers; Add the counting identification information of narcotics and narcotics packaging containers to the filing system to match the information of the person who takes the drug and the person who returns it; If it does not match, an information mismatch prompt is generated to inform the medication doctor and the drug administrator through the preset information platform. 7. According to the pharmaceutical packaging and container shape modeling identification and counting method according to claim 5, it is characterized in that the error analysis is performed on the output data of the preset neural network model after six trainings, and the training model is determined according to the analysis results Eligibility, and if so, identify it as an identification model, including: Inputting the verification image in the data verification set to the trained preset neural network model for recognition, and obtaining the recognition result; Determine the recognition error probability of the trained preset neural network model according to the recognition result, and analyze the recognition error of the trained preset neural network model according to the recognition error probability; Confirm whether the recognition error is within the preset error threshold range, if yes, determine that the training model is qualified, if not, determine that the training model is unqualified; When it is confirmed that the training model is qualified, it is confirmed as the recognition model. 8. The pharmaceutical packaging and container shape modeling identification and counting method according to claim 5, wherein the method further comprises: Obtaining the model parameters of the recognition model, and obtaining the mean value of the loss function, the mean value of the detection loss, the mean value of the classification loss, the precision value and the threshold value of the recall rate of the recognition model according to the model parameters; Determine the detection accuracy of the recognition model according to the loss function mean value and the detection loss mean value; determining the classification accuracy of the recognition model based on the classification loss mean value; The recognition accuracy of the recognition model is determined according to the precision value and the recall rate threshold, and whether the recognition model needs to continue training is determined based on the detection precision, classification precision and recognition precision. 9. The drug packaging and container shape modeling identification and counting method according to claim 6, characterized in that, according to the number of containers, the type of containers and the state of use of the containers, the packaging containers of narcotics and narcotic drugs are classified and recycled in batches, including : Divide the packaging containers of narcotics and narcotics to be identified according to the type of the container, and obtain the first division result; Carry out used empty bottles, unused half bottles, and unused full bottles for each of the poisonous and narcotic drug packaging containers in the first classification result according to the use state of the container, and obtain the second classification result; According to the first classification result and the second classification result, an adapted recovery mode is selected to perform batch recovery processing on the packaging containers of narcotics and narcotics. 10. A drug packaging and container shape modeling recognition and counting system, characterized in that the system includes: The collection module is used to collect the shape image data of N pieces of packaging containers of narcotics and narcotics through a preset image collection method; The setting module is used to set the recognition mode of the packaging container image of narcotic drugs; A modeling module, for learning and modeling the packaging container outline image data of the N pieces of narcotics and narcotics through the YOLOv5 machine learning algorithm based on the recognition pattern, to obtain a recognition model; The loading module is used to load the recognition model into the terminal application program to perform batch recognition, counting and recovery of narcotics and narcotics.

Images