Intelligent kitchen garbage identification and classification method

Abstract

The invention relates to the technical field of big data and machine learning, in particular to an intelligent kitchen garbage recognition and classification method, which comprises the following steps of S1, constructing a kitchen garbage multi-label classification data set which comprises a plurality of images of different types of kitchen garbage and temperature and humidity data corresponding to the kitchen garbage; s2, a multi-label kitchen garbage classification model is constructed, an image feature extraction module extracts an image feature map, an environment feature extraction module extracts temperature and humidity feature values, and the extracted feature map is sent into a data convolution module to be processed so as to be adaptive to an image capture area and a sensor induction area; s3, using the kitchen garbage multi-label classification data set to train a multi-label kitchen garbage classification model; and S4, by collecting to-be-predicted kitchen garbage data at a garbage throwing place, a model loading module is used for performing multi-label classification on the to-be-predicted kitchen garbage data by using the trained multi-label kitchen garbage classification model, so that the garbage classification accuracy and classification efficiency are improved.

Description

Intelligent kitchen waste identification and classification method Technical Field

[0001] This invention relates to the fields of big data and machine learning technology, and in particular to an intelligent method for identifying and classifying kitchen waste. Background Technology

[0002] In the current field of environmental protection and resource recycling, the sorting and treatment of food waste is an important issue. However, most existing food waste sorting methods rely on manual sorting, which is not only inefficient but also easily affected by human factors, leading to inaccurate sorting. Furthermore, with the acceleration of urbanization, the amount of food waste has increased dramatically, and traditional manual sorting methods can no longer meet the needs of large-scale, high-efficiency sorting. In recent years, although some automated classification methods based on computer vision and machine learning have been proposed, these methods often only focus on image features, ignoring the environmental characteristics of the food waste, such as temperature and humidity. These features are equally important for accurate classification of food waste. Some food waste may change under specific temperature and humidity conditions, thus affecting the classification results. Existing multi-label classification methods also face some challenges in processing food waste. Because food waste is diverse, and an image may contain multiple types of waste, accurately identifying and classifying these types of waste is a difficult problem. Existing multi-label classification methods often perform poorly when dealing with images with complex backgrounds and occlusions. Data set construction is also a challenge in current technologies. Given the diverse types and forms of kitchen waste, constructing a comprehensive, accurate dataset that includes multiple categories and environmental features is a challenging task. The quality of the dataset directly impacts the performance and accuracy of multi-label kitchen waste classification models. Therefore, there is an urgent need for an intelligent kitchen waste identification and classification method that can comprehensively consider image and environmental features and effectively handle multi-label classification problems.

[0003] Therefore, an intelligent method for identifying and classifying kitchen waste is proposed. Summary of the Invention

[0004] The present invention aims to address the problems of current traditional kitchen waste sorting, which cannot judge kitchen waste according to actual seasonal changes, lacks flexibility and adaptability, and consumes a lot of human resources. Existing kitchen waste sorting methods cannot identify the changes that certain kitchen waste may undergo under specific temperature and humidity conditions, thus affecting the sorting results and resulting in low sorting accuracy. Furthermore, existing multi-label sorting methods suffer from slow processing speed and low sorting efficiency.

[0005] The technical solution adopted by this invention to solve its technical problem is: an intelligent kitchen waste identification and classification method, comprising the following steps: S1: Constructing a multi-label kitchen waste classification dataset, including several images of multiple different categories of kitchen waste and corresponding temperature and humidity data of the kitchen waste, with each image's label including one or more categories; S2: Constructing a multi-label kitchen waste classification model, including an image feature extraction module, a temperature and humidity feature extraction module, and a dynamic data convolution module; the image feature extraction module extracts feature maps from the images input to the multi-label kitchen waste classification model, and the environmental feature extraction module extracts feature values ​​from the temperature and humidity parameters input to the multi-label kitchen waste classification model; the extracted feature maps are then fed into the data convolution module for processing to adapt to the image capture area and the sensor sensing area; S3: Training the multi-label kitchen waste classification model using the constructed multi-label kitchen waste classification dataset, including a data preparation module, a loss function selection module, an optimizer setting module, and a model training module; S4: Collecting kitchen waste data to be predicted at the waste disposal site, and using the trained multi-label kitchen waste classification model to perform multi-label classification on the kitchen waste data to be predicted.

[0006] Furthermore, S1 specifically includes: S11: By installing an image acquisition unit, a temperature sensor, a humidity sensor, and a Zigbee module at the food waste disposal point, the image acquisition unit, temperature sensor, and humidity sensor are all connected to the data processing module via a wireless communication protocol; S12: The Zigbee module is used to periodically send the collected data to the data processing module and perform data integration, filtering, and noise reduction processing; S13: Based on the collected food waste categories according to different seasons, the data processing module selects multiple combinations of food waste images, temperature data, and humidity data from different time periods throughout the year to obtain a multi-label classification dataset of food waste, and then transmits the multi-label classification dataset of food waste to the image feature extraction module and the temperature and humidity feature extraction module.

[0007] Furthermore, S2 specifically includes: S21: The image feature extraction module, upon receiving the multi-label classification dataset of kitchen waste, uses a pre-trained convolutional neural network as an image feature extractor to extract feature maps of the images in the multi-label classification dataset of kitchen waste based on the data convolution operation, ReLU activation function formula, and pooling operation formula in the convolutional neural network. Then, the extracted feature maps of the images in the multi-label classification dataset of kitchen waste are input to the data convolution module; S22: The environmental feature extraction module, upon receiving the multi-label classification dataset of kitchen waste, extracts feature maps of the images based on the multilayer perceptron in the preset neural network layer and the multi-label classification dataset of kitchen waste. Temperature and humidity feature vectors are extracted from the temperature and humidity data, and then the extracted temperature and humidity feature vectors from the multi-label classification dataset of kitchen waste are transmitted to the data convolution module; S23: After receiving the feature map of the image and the temperature and humidity feature vectors from the multi-label classification dataset of kitchen waste, the data convolution module performs global average pooling on the image feature map; and adjusts the parameters of the convolution kernel according to the preset multilayer perceptron input image feature map and temperature and humidity feature vectors; and performs convolution operation on the image feature map using the convolution kernel with adjusted parameters, and outputs the output feature map after convolution operation to adapt to the image capture area and the sensor sensing area.

[0008] Furthermore, the method for extracting feature maps from images in the multi-label classification dataset of kitchen waste based on the data convolution operation, ReLU activation function, and pooling operation in the convolutional neural network is as follows: Convolution operation: ReLU activation function: Pooling operations: ;in, It is the feature map of the extracted image. It is the output after the convolutional layer operation. It is the output of the ReLU activation function layer. These are images from a multi-label classification dataset for kitchen waste. It is a convolution kernel; These are the coordinates of the output feature map. , It is the size of the convolution kernel. and It is the size of the pooling window.

[0009] Furthermore, the formula for extracting the feature vectors of temperature and humidity based on the multilayer perceptron in the preset neural network layers and the temperature and humidity data of the kitchen waste multi-label classification dataset is as follows: ;in, It extracts the feature vectors from the temperature and humidity data of the multi-label classification dataset for kitchen waste. It is the ReLU activation function. The weights of the Lth layer, and (0, 1), It is a bias term. It is the first The output of the layer, for the output layer, , It is a temperature vector. , This is the temperature data for day n, in degrees Celsius. It is a humidity vector , This is the humidity data for day n, and its unit is RH%.

[0010] Furthermore, S23 specifically includes: S231: The process of performing global average pooling on the image feature map is as follows: ;in, , is a statistical representation of the image feature map. , Let be the feature map of the image, where It is the number of channels. It's about height. Where R is the width, and R represents the set of real numbers; S232: Adjust the parameters of the convolution kernel according to the preset multilayer perceptron input image feature map and temperature and humidity feature vector: ;in, These are the adjusted convolution kernel parameters. , is a statistical representation of the image feature map. , is the temperature and humidity feature vector, R is the set of real numbers, and C and D are the number of channels; S233: Use the convolution kernel with adjusted parameters to perform a convolution operation on the image feature map, and output the convolutional feature map to adapt to the image capture area and the sensor sensing area: ;in, It is the output feature map after the convolution operation. It is a statistical representation of image feature maps. These are the adjusted convolution kernel parameters.

[0011] Furthermore, S3 specifically includes: S31: After receiving the multi-label classification dataset for kitchen waste, the data preparation module loads the multi-label classification dataset for kitchen waste; it also preprocesses the image data in the multi-label classification dataset for kitchen waste; and it divides the preprocessed multi-label classification dataset for kitchen waste into training set, validation set, and test set, and then transmits the training set, validation set, and test set data to the model training module; S32: The loss function selection module is used to select the binary cross-entropy loss function to handle the independence of each category label; S33: The optimizer setting module is used to select the Adam algorithm... The method is used to update the weights of the multi-label kitchen waste classification model; S34: The model training module is used to iteratively train the multi-label kitchen waste classification model based on the received preprocessed training set, validation set, and test set data after receiving them, calculates the loss through forward propagation, and updates the weights through backpropagation; and is used to divide the data of each type of waste in the test set after being classified by the multi-label kitchen waste classification model into positive sample data and negative sample data; and is used to calculate the classification accuracy of each type of waste based on the positive sample data and negative sample data of each type of waste.

[0012] Furthermore, the formula for calculating the classification accuracy of each waste type based on the positive and negative sample data is as follows: ;in, It is the classification accuracy of type c. These are positive sample data that were correctly predicted as category c by the multi-label kitchen waste classification model. Negative sample data that were incorrectly predicted as category c by the multi-label kitchen waste classification model.

[0013] Furthermore, S4 specifically includes: S41: acquiring image data and temperature and humidity data of the kitchen waste to be predicted at the disposal site by installing an image acquisition unit, temperature sensor, humidity sensor, and Zigbee module; S42: the data preparation module, upon receiving the image data and temperature and humidity data of the kitchen waste to be predicted, preprocesses the image data and temperature and humidity data of the kitchen waste to be predicted, and then transmits the preprocessed image data and temperature and humidity data of the kitchen waste to be predicted to the model loading module; S43: the model loading module, upon receiving the preprocessed image data and temperature and humidity data of the kitchen waste to be predicted, loads the pre-trained multi-label kitchen waste classification model for classification and recognition; and compares the category probability output by the multi-label kitchen waste classification model, the accuracy of each category, and the classification accuracy of each type of waste with a preset probability threshold to determine the kitchen waste classification result.

[0014] Furthermore, in S43, the method for comparing the category probabilities identified by the multi-label kitchen waste classification model with the classification accuracy and preset probability thresholds for each waste type is as follows: ;in, The categories output by the multi-label kitchen waste sorting model The predicted probability, It is a preset category. The predicted probability threshold and 0,1), It is the classification accuracy of category c, when The output result is At that time, determine the category Types of waste.

[0015] The beneficial effects of this invention are: 1. It can comprehensively collect images and temperature and humidity data of kitchen waste to form a dataset containing multi-label classification information. This helps the multi-label kitchen waste classification model learn the characteristics of kitchen waste in different seasons and environments, improving the accuracy and generalization ability of classification.

[0016] 2. The image feature extraction module efficiently extracts key features from images, providing strong support for classification. The temperature and humidity feature extraction module enables the multi-label food waste classification model to comprehensively consider the impact of environmental factors on food waste classification. The data convolution module improves the adaptability of the multi-label food waste classification model to complex scenarios.

[0017] 3. By calculating channel descriptors, learning attention weights, and applying these weights to feature maps, we can focus on more important regions and channels in the image, thereby improving classification accuracy.

[0018] 4. By collecting and preprocessing data in real time, a pre-trained multi-label kitchen waste classification model is loaded, providing reliable input for the multi-label kitchen waste classification model. By comparing the output category probability identified by the multi-label kitchen waste classification model with the preset probability threshold, the efficiency of kitchen waste identification can be improved. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments will be briefly described below. The drawings described below only relate to some embodiments of the present invention and are not intended to limit the present invention.

[0020] In the accompanying drawings: Figure 1 is a schematic diagram of an intelligent kitchen waste identification and classification method according to an embodiment of the present invention. Detailed Implementation

[0021] The following will describe the concept and technical effects of the present invention clearly and completely with reference to the embodiments, so as to fully understand the purpose, features and effects of the present invention.

[0022] It should be noted that the use of terms such as "an embodiment," "an embodiment," "an exemplary embodiment," and "some embodiments" in the specification indicates that the described embodiment may include a specific feature, structure, or characteristic, but not every embodiment necessarily includes that specific feature, structure, or characteristic. Furthermore, when describing a specific feature, structure, or characteristic in conjunction with embodiments, implementing such a feature, structure, or characteristic in conjunction with other embodiments should be within the knowledge of those skilled in the art.

[0023] Generally, terms can be understood at least partly from their use in context. For example, depending at least partly on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in a singular sense, or a combination of features, structures, or characteristics in a plural sense. Additionally, the term "based on" can be understood not necessarily to convey an exclusive set of factors, but rather, alternatively, depending at least partly on the context, to allow for the presence of other factors that are not necessarily explicitly described.

[0024] Please refer to Figure 1. An intelligent kitchen waste identification and classification method includes the following steps: S1: Construct a multi-label kitchen waste classification dataset, including several images of different categories of kitchen waste and corresponding temperature and humidity data of the kitchen waste. Each image is labeled with one or more categories; S2: Construct a multi-label kitchen waste classification model, including an image feature extraction module, a temperature and humidity feature extraction module, and a dynamic data convolution module. The image feature extraction module extracts feature maps from the images input to the multi-label kitchen waste classification model, and the environmental feature extraction module extracts feature values ​​from the temperature and humidity parameters input to the multi-label kitchen waste classification model. The extracted feature maps are then fed into the data convolution module for processing to adapt to the image capture area and the sensor sensing area; S3: Train the multi-label kitchen waste classification model using the constructed multi-label kitchen waste classification dataset, including a data preparation module, a loss function selection module, an optimizer setting module, and a model training module; S4: Collect kitchen waste data to be predicted at the waste disposal site. The model loading module is used to perform multi-label classification of the kitchen waste data to be predicted using the trained multi-label kitchen waste classification model.

[0025] In this embodiment, S1 specifically includes: S11: By installing an image acquisition unit, a temperature sensor, a humidity sensor, and a Zigbee module at the food waste disposal site, the image acquisition unit, temperature sensor, and humidity sensor are all connected to the data processing module via a wireless communication protocol; S12: The Zigbee module is used to periodically send the collected data to the data processing module and perform data integration, filtering, and noise reduction processing; S13: The data processing module selects multiple combinations of food waste images, temperature data, and humidity data from different time periods throughout the year based on the collected categories of food waste in different seasons to obtain a multi-label classification dataset of food waste, and then transmits the multi-label classification dataset of food waste to the image feature extraction module and the temperature and humidity feature extraction module.

[0026] Through the combination of the above steps, the sensors and image acquisition units are connected to the data processing module via a wireless communication protocol, ensuring real-time data transmission and efficient processing. Through data integration, data from different sensors and image acquisition units can be uniformly managed and analyzed. Based on the collected categories of kitchen waste in different seasons, the data processing module selects multiple combinations of kitchen waste images and temperature and humidity data from different time periods throughout the year. This ensures that the dataset contains rich category information and the impact of seasonal changes on kitchen waste, providing strong data support for the subsequent training of multi-label kitchen waste classification models. The multi-label kitchen waste classification dataset is transmitted to the image feature extraction module and the temperature and humidity feature extraction module, providing the necessary input for the subsequent training of multi-label kitchen waste classification models.

[0027] In this embodiment, S2 specifically includes: S21: The image feature extraction module, after receiving the multi-label classification dataset of kitchen waste, uses a pre-trained convolutional neural network as an image feature extractor to extract feature maps of the images in the multi-label classification dataset of kitchen waste based on the data convolution operation, ReLU activation function formula, and pooling operation formula in the convolutional neural network, and then outputs the extracted feature maps of the images in the multi-label classification dataset of kitchen waste to the data convolution module; S22: The environmental feature extraction module, after receiving the multi-label classification dataset of kitchen waste, extracts feature maps of the images in the multi-layer perceptron in the preset neural network layer and the multi-label classification dataset of kitchen waste based on the data convolution operation, ReLU activation function formula, and pooling operation formula, and then outputs the extracted feature maps of the images in the multi-label classification dataset of kitchen waste to the data convolution module; The temperature and humidity data are used to extract temperature and humidity feature vectors, and then the extracted temperature and humidity feature vectors of the kitchen waste multi-label classification dataset are transmitted to the data convolution module; S23: The data convolution module is used to perform global average pooling on the image feature map after receiving the feature map of the image and the temperature and humidity feature vectors in the kitchen waste multi-label classification dataset; and is used to adjust the parameters of the convolution kernel according to the preset multilayer perceptron input image feature map and temperature and humidity feature vectors; and is used to perform convolution operation on the image feature map using the convolution kernel with adjusted parameters, and output the output feature map after convolution operation to adapt to the image capture area and the sensor sensing area.

[0028] By combining the above steps, the environmental feature extraction module extracts feature vectors of temperature and humidity through a neural network layer, taking into account the impact of environmental factors on kitchen waste classification and improving the generalization ability of the multi-label kitchen waste classification model. The data convolution module receives feature maps and feature values ​​from the image feature extraction module and the environmental feature extraction module, realizing feature fusion. By introducing an attention mechanism or adaptive convolutional layer, the data convolution module can dynamically adjust the parameters of the convolution kernel according to the image feature map, adaptively focusing on key regions in the image and improving classification accuracy. The data convolution module can also dynamically adjust according to temperature and humidity feature values ​​to adapt to changes in the sensor sensing area, further enhancing the adaptability of the multi-label kitchen waste classification model.

[0029] In this embodiment, the method for extracting feature maps from images in the multi-label classification dataset of kitchen waste based on data convolution operations, ReLU activation function, and pooling operations in the convolutional neural network is as follows: Convolution operation: ReLU activation function: Pooling operations: ;in, It is the feature map of the extracted image. It is the output after the convolutional layer operation. It is the output of the ReLU activation function layer. These are images from a multi-label classification dataset for kitchen waste. It is a convolution kernel; These are the coordinates of the output feature map. , It is the size of the convolution kernel. and It is the size of the pooling window.

[0030] In this embodiment, the formula for extracting the feature vector of temperature and humidity based on the temperature data and humidity data of the multilayer perceptron in the preset neural network layer and the multi-label classification dataset of kitchen waste is as follows: ;in, It extracts the feature vectors from the temperature and humidity data of the multi-label classification dataset for kitchen waste. It is the ReLU activation function. The weights of the Lth layer, and (0, 1), It is a bias term. It is the first The output of the layer, for the output layer, , It is a temperature vector. , This is the temperature data for day n, in degrees Celsius. It is a humidity vector , This is the humidity data for day n, and its unit is RH%.

[0031] In this embodiment, S23 specifically includes: S231: The process of performing global average pooling on the image feature map is as follows: ;in, , is a statistical representation of the image feature map. , Let be the feature map of the image, where It is the number of channels. It's about height. Where R is the width, and R represents the set of real numbers; S232: Adjust the parameters of the convolution kernel according to the preset multilayer perceptron input image feature map and temperature and humidity feature vector: ;in, These are the adjusted convolution kernel parameters. , is a statistical representation of the image feature map. , is the temperature and humidity feature vector, R is the set of real numbers, and C and D are the number of channels; S233: Use the convolution kernel with adjusted parameters to perform a convolution operation on the image feature map, and output the convolutional feature map to adapt to the image capture area and the sensor sensing area: ;in, It is the output feature map after the convolution operation. It is a statistical representation of image feature maps. These are the adjusted convolution kernel parameters.

[0032] In this embodiment, S3 specifically includes: S31: After receiving the multi-label classification dataset for kitchen waste, the data preparation module loads the multi-label classification dataset for kitchen waste; and preprocesses the image data in the multi-label classification dataset for kitchen waste; and divides the preprocessed multi-label classification dataset for kitchen waste into training set, validation set and test set, and then transmits the training set, validation set and test set data to the model training module; S32: The loss function selection module is used to select the binary cross-entropy loss function to handle the independence of each category label; S33: The optimizer setting module is used to select Adam... The algorithm is used to update the weights of the multi-label kitchen waste classification model; S34: The model training module is used to iteratively train the multi-label kitchen waste classification model based on the received preprocessed training set, validation set, and test set data after receiving them. It calculates the loss through forward propagation and updates the weights through backpropagation. It is also used to divide the data of each type of waste in the test set into positive sample data and negative sample data after being classified by the multi-label kitchen waste classification model. It is also used to calculate the classification accuracy of each type of waste based on the positive sample data and negative sample data of each type of waste.

[0033] Through the combination of the above steps, the data preparation module preprocesses the image data, the loss function selection module chooses the binary cross-entropy loss function to handle the independence of each category label, and calculates the prediction error for each label independently. The Adam algorithm is a gradient descent-based optimization algorithm with adaptive learning rate adjustment, which can efficiently find the global optimum. The model training module receives the preprocessed training, validation, and test sets and optimizes the parameters of the multi-label food waste classification model through iterative training. During training, the model training module calculates the loss using forward propagation and updates the weights using backpropagation. The model training module calculates the accuracy for each category of data, which is one of the important indicators for evaluating the performance of the multi-label food waste classification model and helps to understand the performance of the multi-label food waste classification model in each category.

[0034] In this embodiment, the formula for calculating the classification accuracy of each waste type based on the positive and negative sample data is as follows: ;in, It is the classification accuracy of type c. These are positive sample data that were correctly predicted as category c by the multi-label kitchen waste classification model. Negative sample data that were incorrectly predicted as category c by the multi-label kitchen waste classification model.

[0035] In this embodiment, S4 specifically includes: S41: acquiring image data and temperature and humidity data of the kitchen waste to be predicted at the disposal site by installing an image acquisition unit, a temperature sensor, a humidity sensor, and a Zigbee module; S42: the data preparation module, upon receiving the image data and temperature and humidity data of the kitchen waste to be predicted, preprocesses the image data and temperature and humidity data of the kitchen waste to be predicted, and then transmits the preprocessed image data and temperature and humidity data of the kitchen waste to be predicted to the model loading module; S43: the model loading module, upon receiving the preprocessed image data and temperature and humidity data of the kitchen waste to be predicted, loads the pre-trained multi-label kitchen waste classification model for classification and recognition; and compares the category probability output by the multi-label kitchen waste classification model, the accuracy of each category, and the classification accuracy of each type of waste with a preset probability threshold to determine the kitchen waste classification result.

[0036] By combining the above steps, the model loading module can quickly load the pre-trained multi-label kitchen waste classification model and use it to classify and identify the pre-processed kitchen waste data. By directly inputting the pre-processed data into the multi-label kitchen waste classification model, repetitive data processing is avoided, and the classification and identification speed is improved. The model loading module classifies the data based on the category probabilities output by the multi-label kitchen waste classification model, calculates the accuracy of each category, and compares it with the preset probability threshold to determine the final kitchen waste classification result. By considering the output probability of the multi-label kitchen waste classification model and the characteristics of the data itself, the accuracy and reliability of the classification results are improved.

[0037] In this embodiment, the method for comparing the category probability identified by the multi-label kitchen waste classification model, the classification accuracy of each waste type, and the preset probability threshold in S43 is as follows: ;in, The categories output by the multi-label kitchen waste sorting model The predicted probability, It is a preset category. The predicted probability threshold and 0,1), It is the classification accuracy of category c, when The output result is At that time, determine the category Types of waste.

[0038] In summary, the entire method achieves intelligent processing of food waste sorting. By installing image acquisition units, temperature sensors, humidity sensors, and Zigbee modules at the food waste disposal points, a reliable data source is provided for subsequent data processing and classification. The data preparation module receives raw data from the sensors and performs preprocessing operations to improve data quality and the recognition effect of the multi-label food waste classification model. This module also divides the dataset into training, validation, and test sets, facilitating subsequent training and performance evaluation of the multi-label food waste classification model. The multi-label food waste classification model is iteratively trained using the preprocessed training set data. Appropriate loss functions and optimizers are selected to optimize the performance of the multi-label food waste classification model. This ensures that the multi-label food waste classification model can accurately identify the categories of food waste. The model loading module determines the final food waste classification result by combining the category probabilities output by the multi-label food waste classification model, the precision of each category, and preset probability thresholds. This step enables rapid and accurate sorting of kitchen waste. The entire method, from data collection and preprocessing to multi-label kitchen waste sorting model training and classification recognition, forms a complete data processing chain with high practicality and scalability. Furthermore, this method can be combined with other intelligent technologies, such as the Internet of Things and big data, to further enhance the intelligence level of waste sorting.

[0039] The above embodiments are only some embodiments of the present invention, and not all embodiments. Other embodiments obtained by users of the art based on the embodiments of the present invention without creative effort are all within the scope of protection of the present invention.

Claims

1. A method for intelligent identification and classification of kitchen waste, characterized in that: Includes the following steps: S1: Construct a multi-label classification dataset for kitchen waste, including several images of different categories of kitchen waste and corresponding temperature and humidity data. Each image is labeled with one or more categories. S2: Construct a multi-label kitchen waste classification model, including an image feature extraction module, a temperature and humidity feature extraction module, and a dynamic data convolution module. The image feature extraction module extracts feature maps from the images input to the multi-label kitchen waste classification model, and the environmental feature extraction module extracts feature values ​​from the temperature and humidity parameters input to the model. The extracted feature maps are then processed by the data convolution module to adapt to the image capture area and the sensor sensing area. S3: Train the multi-label kitchen waste classification model using the constructed dataset, including a data preparation module, a loss function selection module, an optimizer setting module, and a model training module. S31: Upon receiving the multi-label kitchen waste classification dataset, the data preparation module loads the dataset and preprocesses the image data within it. Furthermore, it is used to divide the preprocessed multi-label kitchen waste classification dataset into training, validation, and test sets, and then transmit the training, validation, and test set data to the model training module; S32: The loss function selection module is used to select the binary cross-entropy loss function to handle the independence of each category label; S33: The optimizer setting module is used to select the Adam algorithm to update the weights of the multi-label kitchen waste classification model; S34: The model training module is used to iteratively train the multi-label kitchen waste classification model based on the received preprocessed training, validation, and test set data after receiving them, calculating the loss through forward propagation and updating the weights through backpropagation; It is also used to divide the data of each type of waste in the test set into positive sample data and negative sample data after being classified by the multi-label kitchen waste classification model; It is used to calculate the classification accuracy of each type of waste based on the positive and negative sample data of each type of waste; S4: By collecting the kitchen waste data to be predicted at the waste disposal point, the model loading module is used to perform multi-label classification of the kitchen waste data to be predicted using the trained multi-label kitchen waste classification model.

2. The intelligent kitchen waste identification and classification method according to claim 1, characterized in that: S1 specifically includes: S11: Installing an image acquisition unit, a temperature sensor, a humidity sensor, and a Zigbee module at the food waste disposal point. The image acquisition unit, the temperature sensor, and the humidity sensor are all connected to the data processing module via a wireless communication protocol; S12: The Zigbee module is used to periodically send collected data to the data processing module and perform data integration, filtering, and noise reduction; S13: The data processing module selects multiple combinations of food waste images, temperature data, and humidity data from different time periods throughout the year based on the collected food waste categories for different seasons to obtain the food waste multi-label classification dataset, and then transmits the food waste multi-label classification dataset to the image feature extraction module and the temperature and humidity feature extraction module.

3. The intelligent kitchen waste identification and classification method according to claim 1, characterized in that: The formula for calculating the classification accuracy of each waste type based on positive and negative sample data is as follows: ;in, It is the classification accuracy of type c. These are positive sample data that were correctly predicted as category c by the multi-label kitchen waste classification model. Negative sample data that were incorrectly predicted as category c by the multi-label kitchen waste classification model.

4. The intelligent kitchen waste identification and classification method according to claim 1, characterized in that: S4 specifically includes: S41: acquiring image data and temperature and humidity data of the kitchen waste to be predicted at the disposal site by installing an image acquisition unit, a temperature sensor, a humidity sensor, and a Zigbee module; S42: the data preparation module, upon receiving the image data and temperature and humidity data of the kitchen waste to be predicted, preprocesses the image data and temperature and humidity data of the kitchen waste to be predicted, and then transmits the preprocessed image data and temperature and humidity data of the kitchen waste to be predicted to the model loading module; S43: the model loading module, upon receiving the preprocessed image data and temperature and humidity data of the kitchen waste to be predicted, loads a pre-trained multi-label kitchen waste classification model for classification and recognition; and compares the category probability output by the multi-label kitchen waste classification model, the accuracy of each category, and the classification accuracy of each type of waste with a preset probability threshold to determine the kitchen waste classification result.

5. The intelligent kitchen waste identification and classification method according to claim 4, characterized in that: The method for comparing the category probabilities identified by the multi-label kitchen waste classification model, the classification accuracy of each waste type, and the preset probability threshold in S43 is as follows: ;in, The categories output by the multi-label kitchen waste sorting model The predicted probability, It is a preset category. The predicted probability threshold and 0,1)， It is the classification accuracy of category c, when The output result is At that time, determine the category Types of waste.

Images