Smart city classification identification management platform and method based on Internet of Things perception

Abstract

The invention relates to the technical field of smart city management, and particularly discloses a smart city classification and identification management platform and method based on Internet of Things perception. And an advanced image processing and deep learning technology is utilized to analyze the garbage surface state image so as to automatically identify and classify the garbage. Finally, by comparing the garbage type label with the garbage can type label, the system can intelligently judge whether garbage classification errors occur or not and send out a prompt signal in time, and therefore urban managers and citizens are helped to jointly maintain a good urban environment. Thus, the intelligent level of the garbage classification and recovery process can be improved, garbage classification and recovery management work is promoted, and the management requirement of a smart city and the city construction requirement are met.

Description

Technical Field

[0001] The present application relates to the field of smart city management technology, and more specifically, to a smart city classification and identification management platform and method based on Internet of Things perception. Background Art

[0002] Garbage sorting and recycling management are important components of modern urban management, aiming to reduce the total amount of garbage, improve resource utilization, and reduce pressure on the environment. In actual operation, garbage sorting and recycling management still face many challenges. For example, some citizens still lack sufficient knowledge and skills to correctly sort garbage, resulting in poor garbage sorting results. In addition, the existing garbage sorting and recycling management usually relies on manual work, which has a low level of intelligence and is difficult to meet the needs of smart city construction and modern urban management.

[0003] Therefore, an optimized smart city classification identification management platform is desired. Summary of the invention

[0004] The present application provides a smart city classification identification management platform and method based on Internet of Things perception, which can improve the intelligence level of garbage classification and recycling process, promote garbage classification and recycling management, and meet the management needs of smart cities and urban construction needs.

[0005] In a first aspect, a smart city classification identification management platform based on Internet of Things perception is provided, including: a garbage image acquisition module, used to obtain a garbage surface state image of the garbage to be classified collected by a camera; a garbage surface state feature extraction module, used to perform multi-scale feature extraction on the garbage surface state image to obtain a shallow feature of the garbage state and a deep semantic feature of the garbage state; a garbage multi-scale feature interactive fusion module, used to perform kernel correlation feature significant joint interactive coding on the shallow feature of the garbage state and the deep semantic feature of the garbage state to obtain a shallow-deep interactive fusion feature of the garbage state, including: a garbage state shallow-deep kernel feature association coding unit, used to extract the shallow feature of the garbage state and the deep semantic feature of the garbage state; The deep semantic features of the garbage status are associatedly encoded based on kernel semantics to obtain a global association representation between the shallow and deep kernel features of the garbage status; a shallow-deep feature interaction unit of the garbage status is used to interactively fusion encode and significantly sparse modulation analyze the shallow features of the garbage status and the deep semantic features of the garbage status based on the global association representation between the shallow-deep kernel features of the garbage status to obtain the shallow-deep interactive fusion features of the garbage status; a garbage classification management module is used to perform garbage identification based on the shallow-deep interactive fusion features of the garbage status to determine the identification result for representing the garbage type label, and perform garbage classification signal management based on the identification result to confirm whether to generate a garbage classification error prompt signal.

[0006] In a second aspect, a smart city classification identification management method based on Internet of Things perception is provided, including: obtaining a garbage surface state image of the garbage to be classified collected by a camera; performing multi-scale feature extraction on the garbage surface state image to obtain shallow features of the garbage state and deep semantic features of the garbage state; performing kernel association feature significant joint interactive coding on the shallow features of the garbage state and the deep semantic features of the garbage state to obtain shallow-deep interactive fusion features of the garbage state, including: performing kernel semantic-based association coding on the shallow features of the garbage state and the deep semantic features of the garbage state to obtain a global association representation between the shallow-deep kernel features of the garbage state; based on the global association representation between the shallow-deep kernel features of the garbage state, interactive fusion coding and significant sparse modulation analysis are performed on the shallow features of the garbage state and the deep semantic features of the garbage state to obtain the shallow-deep interactive fusion features of the garbage state; performing garbage identification based on the shallow-deep interactive fusion features of the garbage state to determine the identification result for representing the garbage type label, and performing garbage classification signal management based on the identification result to confirm whether to generate a garbage classification error prompt signal.

[0007] The present application provides a smart city classification identification management platform and method based on Internet of Things perception, which can collect garbage surface state images through smart cameras installed in garbage recycling stations, and use advanced image processing and deep learning technology to analyze the garbage surface state images, so as to automatically identify and classify garbage. Finally, by comparing the garbage type label with the garbage bin type label, the system can intelligently determine whether there is a garbage classification error, and send out prompt signals in time, thereby helping city managers and citizens to jointly maintain a good urban environment. In this way, the intelligence level of the garbage classification and recycling process can be improved, and garbage classification and recycling management can be promoted to meet the management needs and urban construction needs of smart cities. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application are briefly introduced below. Obviously, the drawings described below only relate to some embodiments of the present application, and are not intended to limit the present application.

[0009] Figure 1 This is a schematic block diagram of a smart city classification identification management platform based on Internet of Things perception according to an embodiment of the present application.

[0010] Figure 2 This is a data flow diagram of a smart city classification identification management platform based on Internet of Things perception in an embodiment of the present application.

[0011] Figure 3This is a schematic block diagram of a garbage surface state feature extraction module in a smart city classification identification management platform based on Internet of Things perception in an embodiment of the present application.

[0012] Figure 4 This is a schematic block diagram of a garbage multi-scale feature interactive fusion module in a smart city classification identification management platform based on Internet of Things perception in an embodiment of the present application.

[0013] Figure 5 This is a schematic flowchart of a smart city classification, identification and management method based on Internet of Things perception according to an embodiment of the present application. DETAILED DESCRIPTION

[0014] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.

[0015] In recent years, the development of Internet of Things (IoT) technology has provided new ideas for the construction and management of smart cities. By deploying a large number of sensors and smart devices in urban infrastructure, real-time monitoring and data collection of urban operation status can be achieved. Especially in the field of garbage disposal, the use of IoT technology can realize intelligent management of garbage sorting and recycling processes, improve garbage disposal efficiency, and ensure that garbage is correctly classified, which in turn helps reduce environmental pollution.

[0016] Based on this, in the technical solution of this application, a smart city classification identification management platform based on IoT perception is proposed, which can collect garbage surface state images through smart cameras installed in garbage recycling stations, and use advanced image processing and deep learning technology to analyze the garbage surface state images, so as to automatically identify and classify garbage. Finally, by comparing the garbage type label with the garbage bin type label, the system can intelligently determine whether there is a garbage classification error, and send out prompt signals in time, so as to help city managers and citizens jointly maintain a good urban environment. In this way, the intelligence level of the garbage classification and recycling process can be improved, and garbage classification and recycling management can be promoted to meet the management needs of smart cities and urban construction needs.

[0017] Figure 1 This is a schematic block diagram of a smart city classification identification management platform based on Internet of Things perception according to an embodiment of the present application. Figure 2 This is a data flow diagram of the smart city classification identification management platform based on Internet of Things perception in the embodiment of this application. Figure 1 and Figure 2As shown, the smart city classification identification management platform 100 based on Internet of Things perception includes: a garbage image acquisition module 110, which is used to obtain a garbage surface state image of the garbage to be classified collected by a camera; a garbage surface state feature extraction module 120, which is used to perform multi-scale feature extraction on the garbage surface state image to obtain shallow features of the garbage state and deep semantic features of the garbage state; a garbage multi-scale feature interactive fusion module 130, which is used to perform kernel correlation feature significant joint interactive encoding on the shallow features of the garbage state and the deep semantic features of the garbage state to obtain shallow-deep interactive fusion features of the garbage state; a garbage classification management module 140, which is used to perform garbage identification based on the shallow-deep interactive fusion features of the garbage state to determine the identification result for representing the garbage type label, and perform garbage classification signal management based on the identification result to confirm whether to generate a garbage classification error prompt signal.

[0018] In the above-mentioned smart city classification and identification management platform based on IoT perception, the garbage image acquisition module 110 is used to obtain the garbage surface state image of the garbage to be classified collected by the camera. It should be understood that the traditional garbage classification method mainly relies on manual recognition, which is not only inefficient but also prone to errors, especially when faced with large amounts of or complex garbage. By collecting garbage surface state images through smart cameras installed at garbage recycling stations, and analyzing the garbage surface state images using advanced image processing and deep learning technologies, automatic identification and classification of garbage can be achieved. This method not only improves the accuracy and efficiency of garbage classification, but also reduces labor costs, helps to promote the intelligent level of garbage classification and recycling management, and thus meets the management needs of smart cities.

[0019] Optionally, in one embodiment of the present application, obtaining a garbage surface state image of the garbage to be sorted collected by a camera includes: first, installing a smart camera at a key position of the garbage recycling station to ensure that the camera's field of view covers all garbage delivery ports so that the thrown-in garbage can be clearly photographed. When the user throws the garbage into the trash can, the smart camera will be automatically triggered to capture the surface state image of the garbage. This process is usually triggered by sensors such as motion detection or weight change to ensure that image acquisition is only performed when the garbage is thrown in, thereby saving storage space and processing time. The captured garbage surface state image will be transmitted to the central processing system. This process can be completed through a wired or wireless network to ensure that the image data can reach the processing center quickly and safely.

[0020] In the above-mentioned smart city classification identification management platform based on IoT perception, the garbage surface state feature extraction module 120 is used to perform multi-scale feature extraction on the garbage surface state image to obtain shallow features of the garbage state and deep semantic features of the garbage state. It should be understood that shallow features usually include local detail information such as texture, color, edge and shape of the garbage surface. These features are very useful for identifying the specific physical properties of garbage, such as the smooth surface of plastic bottles and the rough texture of paper. Deep semantic features contain higher-level semantic information, such as the category (plastic, paper, metal, etc.) and structural features of garbage. These features help to understand the overall properties and categories of garbage. Through multi-scale feature extraction, the system can capture the feature information of garbage from different levels and scales, thereby more comprehensively describing the state of garbage. This multi-level feature representation helps to improve the accuracy and robustness of garbage identification.

[0021] Optionally, in one embodiment of the present application, Figure 3 This is a schematic block diagram of a garbage surface state feature extraction module in a smart city classification identification management platform based on Internet of Things perception according to an embodiment of the present application. Figure 3 As shown, the garbage surface state feature extraction module 120 includes: a garbage state multi-scale feature extraction unit 121, which is used to input the garbage surface state image into a garbage state multi-scale feature extractor based on a hollow pyramid network model to obtain a garbage state shallow feature map and a garbage state deep semantic feature map as the garbage state deep semantic feature; a spatial constraint feature enhancement processing unit 122, which is used to input the garbage state shallow feature map into a spatial constraint feature enhancement module to obtain a garbage state shallow enhancement feature map as the garbage state shallow feature.

[0022] In the above-mentioned smart city classification identification management platform based on Internet of Things perception, the garbage state multi-scale feature extraction unit 121 is used to input the garbage surface state image into the garbage state multi-scale feature extractor based on the hollow pyramid network model to obtain the garbage state shallow feature map and the garbage state deep semantic feature map as the garbage state deep semantic feature. It should be understood that after the garbage surface state image is processed by the garbage state multi-scale feature extractor based on the hollow pyramid network model, the obtained garbage state shallow feature map contains the shallow features and local fine-grained information such as the surface texture, color, edge and shape of the garbage in the garbage surface state image, while the garbage state deep semantic feature map contains the overall category of the garbage in the garbage surface state image, such as plastic, paper and metal. These features are helpful to identify and classify garbage and improve the robustness and generalization ability of the model.

[0023] In the above-mentioned smart city classification identification management platform based on Internet of Things perception, the spatial constraint feature enhancement processing unit 122 is used to input the garbage status shallow feature map into the spatial constraint feature enhancement module to obtain the garbage status shallow enhancement feature map as the garbage status shallow feature. It should be understood that since the garbage status shallow feature map contains shallow feature information of different local areas related to the garbage status, these feature information have a mutual correlation relationship, which plays an important role in the subsequent garbage identification and classification tasks. This is because different types of garbage may have similarities in the feature performance of certain local areas, and the same type of garbage may also have differences between different local areas. Based on this, in order to improve the quality and expressiveness of the garbage status shallow feature map and make it more suitable for subsequent garbage classification tasks, in the technical solution of the present application, the garbage status shallow feature map is further input into the spatial constraint feature enhancement module to obtain the garbage status shallow enhancement feature map as the garbage status shallow feature. It is worth mentioning that the spatial constraint feature enhancement module can better capture the shallow detail features of the garbage surface state by combining the shallow local features of the garbage state and the global context information, while retaining the overall shallow correlation feature information of the garbage surface state. This is crucial for more accurate garbage identification and classification tasks, and helps to improve the robustness and generalization ability of garbage classification and management systems, so that the model can better cope with various complex situations in practical applications.

[0024] Optionally, in one embodiment of the present application, the spatial constraint feature enhancement processing unit includes: a semantic association constraint score calculation subunit, which is used to perform feature dispersion on the garbage status shallow feature map to obtain a set of garbage status shallow pixel granularity feature vectors, and then perform semantic association constraint score calculation on the set of garbage status shallow pixel granularity feature vectors to obtain a garbage status shallow pixel granularity semantic association spatial constraint score topological representation; a garbage status shallow feature enhancement subunit, which is used to enhance the garbage status shallow feature map based on the garbage status shallow pixel granularity semantic association spatial constraint score topological representation to obtain the garbage status shallow enhanced feature map.

[0025] Optionally, in one embodiment of the present application, the semantic association constraint score calculation subunit is used to: perform feature dispersion on the garbage status shallow feature map along the channel dimension to obtain a set of the garbage status shallow pixel granularity feature vectors; calculate the semantic association score between any two garbage status shallow pixel granularity feature vectors in the set of the garbage status shallow pixel granularity feature vectors to obtain a garbage status shallow pixel granularity semantic association score topological matrix; perform spatial attention attenuation modulation on the garbage status shallow pixel granularity semantic association score topological matrix based on the spatial distance between any two garbage status shallow pixel granularity feature vectors in the set of the garbage status shallow pixel granularity feature vectors to obtain a garbage status shallow pixel granularity semantic association space soft constraint score topological matrix; perform hole convolution encoding on the garbage status shallow pixel granularity semantic association space soft constraint score topological matrix to obtain a garbage status shallow pixel granularity semantic association space soft constraint score topological feature matrix as the garbage status shallow pixel granularity semantic association space constraint score topological representation.

[0026] Optionally, in one embodiment of the present application, the junk status shallow feature enhancement subunit is used to: input the set of junk status shallow feature vectors and the junk status shallow pixel granularity semantic association space soft constraint score topology matrix into a graph convolutional coding module to obtain a set of junk status shallow context pixel granularity feature vectors; and perform feature reshaping on the set of junk status shallow context pixel granularity feature vectors to obtain the junk status shallow enhanced feature map.

[0027] Optionally, in one embodiment of the present application, the feature dispersion of the garbage status shallow feature map along the channel dimension to obtain a set of the garbage status shallow pixel granularity feature vectors, including: performing feature reshaping processing on the garbage status shallow feature map along the channel dimension to obtain a set of garbage status shallow feature vectors; mapping each garbage status shallow feature vector in the set of garbage status shallow feature vectors to the Poincare space to obtain the set of garbage status shallow pixel granularity feature vectors.

[0028] Optionally, in one embodiment of the present application, the semantic association score between any two garbage status shallow pixel granularity feature vectors in the set of garbage status shallow pixel granularity feature vectors is calculated to obtain a garbage status shallow pixel granularity semantic association score topological matrix, including: calculating the square of the norm of the position difference between the i-th garbage status shallow pixel granularity feature vector and the j-th garbage status shallow pixel granularity feature vector in the set of garbage status shallow pixel granularity feature vectors to obtain a garbage status shallow pixel granularity semantic difference one-norm representation; calculating the difference between a constant 1 and the square of the norm of the i-th garbage status shallow pixel granularity feature vector to obtain the i-th garbage status shallow pixel granularity semantic representation; calculating the difference between a constant 1 and the j-th garbage status shallow pixel granularity feature vector The method comprises the following steps: calculating the difference between the square of the one-norm quantity to obtain the j-th garbage status shallow pixel granularity semantic representation; calculating the product between the ith garbage status shallow pixel granularity semantic representation and the j-th garbage status shallow pixel granularity semantic representation, dividing the one-norm representation of the garbage status shallow pixel granularity semantic difference by the product value to obtain the garbage status shallow pixel granularity semantic association information representation; calculating the sum of the double of the garbage status shallow pixel granularity semantic association information representation and the constant 1, inputting the obtained value into the inverse hyperbolic cosine function to obtain the garbage status shallow pixel granularity semantic association score value; arranging multiple garbage status shallow pixel granularity semantic association score values ​​into a matrix form to obtain the garbage status shallow pixel granularity semantic association score topological matrix.

[0029] Optionally, in one embodiment of the present application, based on the spatial distance between any two junk state shallow pixel granularity feature vectors in the set of junk state shallow pixel granularity feature vectors, the junk state shallow pixel granularity semantic association score topology matrix is ​​subjected to spatial attention attenuation modulation to obtain a junk state shallow pixel granularity semantic association spatial soft constraint score topology matrix, including: calculating the spatial distance between the i-th junk state shallow pixel granularity feature vector and the j-th junk state shallow pixel granularity feature vector to obtain a pixel granularity semantic space distance representation; calculating the pixel granularity semantic space distance representation as an exponential power, and taking the self The natural exponential function value with constant e as base is used to obtain the pixel granularity semantic space distance class support representation; after multiplying the pixel granularity semantic space distance class support representation by the spatial attention attenuation modulation hyperparameter to obtain the spatial distance attention attenuation representation, the eigenvalue of the (i, j)th position in the junk state shallow pixel granularity semantic association score topological matrix is ​​divided by the spatial distance attention attenuation representation to obtain the junk state shallow pixel granularity semantic association space soft constraint score; multiple junk state shallow pixel granularity semantic association space soft constraint scores are arranged in a matrix form to obtain the junk state shallow pixel granularity semantic association space soft constraint score topological matrix.

[0030] In summary, in the embodiment of the present application, the garbage state shallow feature map is input into the spatial constraint feature enhancement module for processing using the following spatial constraint enhancement formula to obtain the garbage state shallow enhancement feature map as the garbage state shallow feature; wherein the spatial constraint enhancement formula is: ;in, is the shallow feature map of the garbage state, are the height, width and number of channels of the shallow feature map of the garbage state, respectively. For feature reshaping, and is the set of shallow feature vectors of garbage status, The value of is the number of feature vectors in the set of shallow feature vectors of the garbage state, and is the feature transformation weight matrix, and are the first and shallow pixel granularity feature vectors of garbage state, is the one-norm of the vector, is the inverse hyperbolic cosine function, is the semantic association score calculation operation, is the topological matrix of shallow pixel-level semantic association scores of garbage status. The eigenvalues ​​of the positions, It is the spatial distance calculation operation between vectors. is the spatial attention decay modulation hyperparameter, is the topological matrix of the soft constraint score in the shallow pixel-granular semantic association space of garbage status. The eigenvalues ​​of the positions, is the soft constraint score topology matrix of the shallow pixel granularity semantic association space of the garbage state, It is the hole convolution coding process. is the topological feature matrix of soft constraint score in shallow pixel-granular semantic association space of garbage state, is the graph convolutional encoding operation, is the set of shallow context pixel-granularity feature vectors of garbage state, It is a shallow enhanced feature map of the garbage state.

[0031] In the above-mentioned smart city classification identification management platform based on IoT perception, the garbage multi-scale feature interactive fusion module 130 is used to perform kernel correlation feature significant joint interactive encoding on the shallow features of the garbage status and the deep semantic features of the garbage status to obtain the shallow-deep interactive fusion features of the garbage status. It should be understood that since the shallow enhanced feature map of the garbage status and the deep semantic feature map of the garbage status respectively contain shallow features and deep semantic features related to the garbage status, specifically, the shallow features of the garbage status usually contain rich local detail information, such as feature information such as texture, edge, color and shape of the garbage surface, while the deep semantic features of the garbage status contain higher-level semantic information, such as the category and structural feature information of the garbage. These two types of representation information on different aspects of the garbage status are of great significance to the task of identifying and classifying garbage. Therefore, in order to comprehensively utilize the interaction between the surface state of garbage (shallow features) and the category semantics (deep semantic features), and extract the most representative and discriminative features therefrom, to provide a basis for subsequent garbage classification tasks, in the technical solution of the present application, the shallow features of the garbage state and the deep semantic features of the garbage state are further subjected to significant joint interaction coding of kernel-related features to obtain shallow-deep interactive fusion features of the garbage state. The processing of significant joint interaction coding of kernel-related features can extract the most representative and discriminative feature information from the original shallow enhanced feature map of the garbage state and the deep semantic feature map of the garbage state through multi-level processing steps, and on this basis, effectively fuse these two different levels and multi-scale features of the garbage state to obtain a new feature representation of the garbage state, which includes both the shallow information of the surface state of the garbage and its inherent type semantic features and structural information, thereby providing a more comprehensive description of the garbage state, providing a basis for subsequent garbage identification and classification tasks.

[0032] Optionally, in one embodiment of the present application, Figure 4 This is a schematic block diagram of the garbage multi-scale feature interactive fusion module in the smart city classification identification management platform based on Internet of Things perception in the embodiment of the present application. Figure 4 As shown, the garbage multi-scale feature interactive fusion module 130 includes: a garbage status shallow-deep kernel feature association coding unit 131, which is used to perform kernel semantic-based association coding on the garbage status shallow feature and the garbage status deep semantic feature to obtain a global association representation between the garbage status shallow-deep kernel features; a garbage status shallow-deep feature interaction unit 132, which is used to perform interactive fusion coding and significant sparse modulation analysis on the garbage status shallow feature and the garbage status deep semantic feature based on the global association representation between the garbage status shallow-deep kernel features to obtain the garbage status shallow-deep interactive fusion feature.

[0033] Optionally, in one embodiment of the present application, the garbage status shallow-deep kernel feature association encoding unit includes: a garbage status shallow-deep kernel feature distillation subunit, used to input the garbage status shallow enhancement feature map and the garbage status deep semantic feature map into the kernel feature distillation network to obtain a garbage status shallow kernel feature vector and a garbage status deep semantic kernel feature vector; a garbage status shallow-deep kernel feature global encoding subunit, used to perform global association encoding on the garbage status shallow kernel feature vector and the garbage status deep semantic kernel feature vector to obtain a garbage status shallow-deep kernel feature global encoding matrix as a global association representation between the garbage status shallow-deep kernel features.

[0034] Optionally, in one embodiment of the present application, the garbage status shallow-deep kernel feature distillation subunit is used to: perform feature decomposition on the garbage status shallow enhancement feature map to obtain a sequence of garbage status shallow decomposition feature vectors; calculate the difference features of each garbage status shallow decomposition feature vector in the sequence of garbage status shallow decomposition feature vectors relative to other garbage status shallow decomposition feature vectors to obtain a garbage status shallow decomposition feature difference representation vector; normalize each eigenvalue in the garbage status shallow decomposition feature difference representation vector to obtain a normalized garbage status shallow decomposition feature difference representation vector; use each eigenvalue in the normalized garbage status shallow decomposition feature difference representation vector as a weight, calculate the weighted sum of the sequence of garbage status shallow decomposition feature vectors to obtain the garbage status shallow kernel feature vector.

[0035] Optionally, in one embodiment of the present application, the garbage status shallow-deep kernel feature global coding subunit is used to: calculate the multiplication between the transposed vector of the garbage status shallow kernel feature vector and the garbage status deep semantic kernel feature vector to obtain the garbage status shallow-deep kernel feature global coding matrix.

[0036] Optionally, in one embodiment of the present application, the garbage status shallow-deep feature interaction unit is used to: input the global coding matrix between the garbage status shallow-deep kernel features into a sparse mask unit to obtain a sparse weight matrix associated with the garbage status shallow-deep kernel features; input the garbage status shallow enhancement feature map and the garbage status deep semantic feature map into a position-by-position associated interaction network to obtain a garbage status shallow-deep position-by-position interactive fusion feature map; based on the sparse weight matrix associated with the garbage status shallow-deep kernel features, perform feature significant sparse modulation on the garbage status shallow-deep position-by-position interactive fusion feature map to obtain a garbage status shallow-deep significant sparse interactive fusion feature map as the garbage status shallow-deep interactive fusion feature.

[0037] Optionally, in one embodiment of the present application, the difference characteristics of each garbage status shallow decomposition feature vector in the sequence of garbage status shallow decomposition feature vectors relative to other garbage status shallow decomposition feature vectors are calculated to obtain a garbage status shallow decomposition feature difference representation vector, including: calculating the norm of the position difference between each garbage status shallow decomposition feature vector in the sequence of garbage status shallow decomposition feature vectors and the other garbage status shallow decomposition feature vectors respectively to obtain a sequence of garbage status shallow decomposition feature difference one-norm representation values; calculating the mean value of the sequence of garbage status shallow decomposition feature difference one-norm representation values ​​to obtain a garbage status shallow decomposition feature difference one-norm average representation value; vectorizing and arranging a plurality of the garbage status shallow decomposition feature difference one-norm average representation values ​​to obtain the garbage status shallow decomposition feature difference representation vector.

[0038] Optionally, in one embodiment of the present application, each eigenvalue in the garbage status shallow decomposition feature difference representation vector is normalized to obtain a normalized garbage status shallow decomposition feature difference representation vector, including: calculating the positional exponential function value with a natural constant as the base for each eigenvalue in the garbage status shallow decomposition feature difference representation vector to obtain a garbage status shallow decomposition feature difference class support representation vector; calculating the sum of each eigenvalue in the garbage status shallow decomposition feature difference class support representation vector to obtain a garbage status shallow decomposition feature difference class support sum; calculating the positional division between each eigenvalue in the garbage status shallow decomposition feature difference class support representation vector and the garbage status shallow decomposition feature difference class support sum to obtain the normalized garbage status shallow decomposition feature difference representation vector.

[0039] Optionally, in one embodiment of the present application, the global coding matrix between shallow and deep kernel features of the garbage status is input into a sparse mask unit to obtain a sparse weight matrix associated with shallow and deep kernel features of the garbage status, including: performing a sparse mask operation on the eigenvalues ​​of each position in the global coding matrix between shallow and deep kernel features of the garbage status to obtain a sparse weight matrix associated with shallow and deep kernel features of the garbage status; wherein, in response to the eigenvalues ​​in the global coding matrix between shallow and deep kernel features of the garbage status being greater than twice a preset threshold, the eigenvalues ​​greater than twice the preset threshold are multiplied by a constant 3; in response to the eigenvalues ​​in the global coding matrix between shallow and deep kernel features of the garbage status being less than or equal to the preset threshold, the eigenvalues ​​less than or equal to the preset threshold are divided by a constant 3; in response to the eigenvalues ​​in the global coding matrix between shallow and deep kernel features of the garbage status being greater than the preset threshold and less than or equal to twice the preset threshold, the eigenvalues ​​are taken as original values ​​to obtain the sparse weight matrix associated with shallow and deep kernel features of the garbage status.

[0040] Optionally, in one embodiment of the present application, the garbage status shallow enhancement feature map and the garbage status deep semantic feature map are input into a position-by-position association interaction network to obtain a garbage status shallow-deep position-by-position interactive fusion feature map, including: performing position point multiplication on the garbage status shallow enhancement feature map and the garbage status deep semantic feature map to obtain the garbage status shallow-deep position-by-position interactive fusion feature map.

[0041] Optionally, in one embodiment of the present application, based on the correlation sparse weight matrix between the shallow-deep kernel features of the garbage status, the shallow-deep position-by-position interactive fusion feature map of the garbage status is subjected to feature significant sparse modulation to obtain the shallow-deep significant sparse interactive fusion feature map of the garbage status as the shallow-deep interactive fusion feature of the garbage status, including: multiplying each feature matrix along the channel dimension in the shallow-deep position-by-position interactive fusion feature map of the garbage status by the correlation sparse weight matrix between the shallow-deep kernel features of the garbage status to obtain the shallow-deep significant sparse interactive fusion feature map of the garbage status.

[0042] In summary, in the embodiment of the present application, the shallow features of the garbage status and the deep semantic features of the garbage status are significantly jointly and interactively encoded with kernel-related features using the following feature joint interaction formula to obtain the shallow-deep interactive fusion features of the garbage status; wherein the feature joint interaction formula is: ;in, and are respectively the shallow enhanced feature map of the garbage state and the deep semantic feature map of the garbage state, For kernel feature distillation, and are the shallow kernel feature vector of garbage state and the deep semantic kernel feature vector of garbage state, respectively. is a decomposition operation, are the first decomposition eigenvector, the second decomposition eigenvector, The decomposed eigenvectors, The decomposition eigenvectors and decomposition eigenvectors, is the one-norm of the vector, is the sequence of decomposed eigenvectors, is the number of vectors in the sequence of decomposed eigenvectors minus one, To decompose the eigenvalues ​​of each position in the feature difference representation vector, is the length of the decomposed feature difference representation vector, is the exponential operation, is the kernel eigenvector, is the global encoding matrix between shallow and deep kernel features of garbage state, is the feature value of each position in the global encoding matrix between the shallow and deep kernel features of the garbage state, is the preset threshold, is a sparse mask operation, is the sparse weight matrix of the correlation between shallow and deep kernel features in the garbage state, It is the garbage state shallow-deep interactive fusion feature map by position, It is the point multiplication by position. is matrix multiplication, It is the shallow-deep significant sparse interactive fusion feature map of the garbage state.

[0043] In the above-mentioned smart city classification identification management platform based on IoT perception, the garbage classification management module 140 is used to identify garbage based on the shallow-deep interactive fusion features of the garbage state to determine the identification results used to represent the garbage type label, and to manage the garbage classification signal based on the identification results to confirm whether to generate a garbage classification error prompt signal. Optionally, in one embodiment of the present application, the garbage classification management module is used to: input the shallow-deep significant sparse interactive fusion feature map of the garbage state into a classifier-based garbage identifier to obtain a recognition result, and the recognition result is used to represent the garbage type label; based on the garbage type label and the trash can type label, confirm whether to generate a garbage classification error prompt signal. That is to say, the classification process is performed by using the significant interactive fusion semantics between the shallow features of the surface state of the garbage to be classified and the deep semantic features, so as to automatically identify and classify the garbage. Furthermore, based on the garbage type label and the trash can type label, confirm whether to generate a garbage classification error prompt signal. In this way, by comparing and distinguishing the garbage type labels from the garbage bin type labels, it is possible to intelligently determine whether there is a garbage classification error and issue a garbage classification error prompt signal in a timely manner, thereby helping city managers and citizens to jointly maintain a good urban environment. In this way, the intelligence level of the garbage classification and recycling process can be improved, and garbage classification and recycling management can be promoted to meet the management needs of smart cities and urban construction needs.

[0044] Here, in the case where the garbage state shallow enhancement feature map and the garbage state deep semantic feature map respectively represent image semantic features of different scales and different depths of the garbage surface state image, when performing feature significant joint interactive encoding based on kernel-associated sparse masks, the cross-scale and cross-depth correspondence gaps of the image semantic features will also lead to local distribution gaps of the kernel-associated sparse masks, resulting in large-scale simple repetition of the feature significant joint of the encoded features, affecting the interactive image semantic logical dependency and reducing the accuracy of the classification results based on the interactive image semantic distribution.

[0045] Therefore, the applicant of the present application hopes to improve the overall interactive image semantic consistent aggregation expression effect of the shallow-deep significantly sparse interactive fusion feature map of the garbage state when the shallow-deep significantly sparse interactive fusion feature map of the garbage state is input into a classifier-based garbage identifier to obtain a recognition result, thereby optimizing it.

[0046] In a preferred example, the garbage state shallow-deep significant sparse interactive fusion feature map is input into a classifier-based garbage identifier to obtain a recognition result, including: performing feature clustering on all feature values ​​of the garbage state shallow-deep significant sparse interactive fusion feature map to obtain a garbage state shallow-deep significant sparse interactive fusion intra-class feature set Shallow-deep significant sparse interactive fusion of out-of-class feature sets and garbage status ;in, Indicates the garbage state shallow-deep significant sparse interactive fusion intra-class feature set, The shallow-deep significant sparse interaction fusion intra-class feature set represents the garbage state eigenvalues, The shallow-deep significant sparse interactive fusion out-of-class feature set represents the garbage state feature values.

[0047] Calculate the ratio of the number of eigenvalues ​​in the shallow-deep significant sparse interactive fusion feature set of the garbage state to the number of eigenvalues ​​of all features of the shallow-deep significant sparse interactive fusion feature map of the garbage state ,in, represents the number of eigenvalues ​​in the shallow-deep significant sparse interactive fusion feature set of the garbage state, The number of eigenvalues ​​of all features of the shallow-deep significant sparse interactive fusion feature map representing the garbage state, Indicates a ratio.

[0048] Calculate the absolute value sum of all feature values ​​in the shallow-deep significant sparse interactive fusion class feature set of the garbage state The power of the sum of the absolute values ​​of all eigenvalues ​​in the shallow-deep significant sparse interactive fusion feature map of the garbage state The ratio of the powers is used to obtain the shallow-deep significant sparse interaction fusion weights of the first garbage state.

[0049] Calculate the sum of squares of all eigenvalues ​​in the shallow-deep significant sparse interactive fusion class feature set of the garbage state The power of the garbage state shallow-deep significant sparse interaction fusion feature map of the sum of squares of all eigenvalues The ratio of the powers is used to obtain the shallow-deep significant sparse interaction fusion weights of the second garbage state.

[0050] That is, the first garbage state shallow-deep significant sparse interaction fusion weight And the second garbage state shallow-deep significant sparse interaction fusion weights It is expressed as: ;in, Represents each eigenvalue of the shallow-deep significant sparse interactive fusion feature map of the garbage state, represents the shallow-deep significant sparse interaction fusion weight of the first garbage state, Represents the shallow-deep significant sparse interaction fusion weight of the second garbage state.

[0051] Then, for each feature value in the shallow-deep significant sparse interactive fusion intra-class feature set of the garbage state, the product of the feature value and the second garbage state shallow-deep significant sparse interactive fusion weight is calculated, and then the product of the feature value and the first garbage state shallow-deep significant sparse interactive fusion weight is added to obtain the optimized feature value. ,in, Represents the first intra-class feature set based on the shallow-deep significant sparse interaction fusion of the garbage state Optimize the eigenvalues.

[0052] For each feature value in the garbage state shallow-deep significant sparse interactive fusion out-of-class feature set, calculate the product of the feature value and the first garbage state shallow-deep significant sparse interactive fusion weight to obtain the optimized feature value ;in, represents the first out-of-class feature set based on the shallow-deep significant sparse interaction fusion of the garbage state Optimize the eigenvalues.

[0053] Finally, the optimized feature values ​​of the shallow-deep significant sparse interactive fusion in-class feature set of the garbage status and the shallow-deep significant sparse interactive fusion out-class feature set of the garbage status are used to form an optimized shallow-deep significant sparse interactive fusion feature map of the garbage status, and the optimized shallow-deep significant sparse interactive fusion feature map of the garbage status is input into the classifier-based garbage identifier to obtain the recognition result.

[0054] That is, while clustering the shallow-deep significantly sparse interactive fusion feature map of the garbage state, the interactive description of the key feature information of the shallow-deep significantly sparse interactive fusion feature map of the garbage state in the clustering process is carried out, and the geometric equivariant topology of the features is constructed by low-rank harmonic modulation of the clustering features of the shallow-deep significantly sparse interactive fusion feature map of the garbage state and the equivariant basis of the features as a whole, so as to obtain the schematic distribution translation and rotation symmetry of the clustering features of the shallow-deep significantly sparse interactive fusion feature map of the garbage state relative to the features as a whole, thereby On the basis of introducing geometric message passing into the feature expression of the shallow-deep significantly sparse interactive fusion feature map of the garbage status, the clustering mapping symmetry of the shallow-deep significantly sparse interactive fusion feature map of the garbage status is realized by manipulating the irreducible low-rank order coefficients, thereby improving the clustering-based feature representation consistency of the shallow-deep significantly sparse interactive fusion feature map of the garbage status, thereby improving the accuracy of the recognition result obtained by inputting the shallow-deep significantly sparse interactive fusion feature map of the garbage status into a classifier-based garbage identifier, providing support for accurate garbage classification and recycling management.

[0055] In summary, according to the embodiment of the present application, the smart city classification identification management platform based on IoT perception is explained, which can collect garbage surface state images through smart cameras installed in garbage recycling stations, and use advanced image processing and deep learning technology to analyze the garbage surface state images, so as to automatically identify and classify garbage. Finally, by comparing the garbage type label with the garbage bin type label, the system can intelligently determine whether there is a garbage classification error, and send out prompt signals in time, thereby helping city managers and citizens to jointly maintain a good urban environment. In this way, the intelligence level of the garbage classification and recycling process can be improved, and garbage classification and recycling management can be promoted to meet the management needs of smart cities and urban construction needs.

[0056] Figure 5 This is a schematic flow chart of a smart city classification identification management method based on Internet of Things perception according to an embodiment of the present application. Figure 5As shown, the smart city classification identification management method based on Internet of Things perception includes: S1, obtaining a garbage surface state image of the garbage to be classified collected by a camera; S2, performing multi-scale feature extraction on the garbage surface state image to obtain shallow features of the garbage state and deep semantic features of the garbage state; S3, performing kernel correlation feature significant joint interactive encoding on the shallow features of the garbage state and the deep semantic features of the garbage state to obtain shallow-deep interactive fusion features of the garbage state; S4, performing garbage identification based on the shallow-deep interactive fusion features of the garbage state to determine the identification result for representing the garbage type label, and performing garbage classification signal management based on the identification result to confirm whether to generate a garbage classification error prompt signal.

[0057] Optionally, in one embodiment of the present application, the shallow features of the garbage status and the deep semantic features of the garbage status are subjected to significant joint interactive coding of kernel association features to obtain shallow-deep interactive fusion features of the garbage status, including: performing kernel semantic-based association coding on the shallow features of the garbage status and the deep semantic features of the garbage status to obtain a global correlation representation between the shallow and deep kernel features of the garbage status; based on the global correlation representation between the shallow and deep kernel features of the garbage status, interactive fusion coding and significant sparse modulation analysis are performed on the shallow features of the garbage status and the deep semantic features of the garbage status to obtain the shallow-deep interactive fusion features of the garbage status.

[0058] Here, those skilled in the art can understand that the specific operations of each module in the above-mentioned smart city classification identification management method based on Internet of Things perception have been referred to above. Figures 1 to 4 It has been introduced in detail in the description of the smart city classification identification management platform based on IoT perception, and therefore, its repeated description will be omitted.

[0059] In another embodiment of the present application, a smart city Internet of Everything base integrated management platform is also provided, and the smart city Internet of Everything base integrated management platform supports the smart city classification identification management platform based on IoT perception to achieve indiscriminate software and hardware integration. Specifically, it is achieved by: facing infinite and dynamic scenarios, supporting the access and integration of multiple heterogeneous devices and systems, and outputting standardized and structured data; establishing a technical system for application development and deployment distribution by zero code; freely decoupling and coupling applications and data according to usage needs; fully realizing the interconnection and ubiquitous use of various smart hardware device sensors, fully connecting the previously built systems, fully integrating multiple heterogeneous data, quickly building scenario applications, and continuously and seamlessly upgrading the smart city system and subsystems.

[0060] In the above-mentioned smart city Internet of Everything base integrated management platform, the problem of structuring and standardizing logical relationships of multi-source heterogeneous data is solved, and various devices and systems with different communication protocols, functions, application scenarios, and manufacturers can be certified through the BCCA protocol pipeline by simply filling in a form without changing the device's software and hardware, so as to achieve rapid integration and achieve the Internet of Everything. At the same time, the decoupling of software and hardware is achieved, that is, users no longer need complex program development to build applications, and can develop applications through simple operations such as checking, dragging, pulling, and dragging in the IDE development environment; the hardware side only needs to consider whether the smart hardware can meet the usage requirements. Any smart hardware with an open communication protocol is directly coupled with the application system developed in the IDE environment and can be freely deployed. A meshed decentralized collaborative computing power layout is achieved, that is, the system can still operate normally when it is disconnected from the network. In extreme cases, such as when the application system is attacked, it can operate independently and normally without the application system.

[0061] The smart city Internet of Everything base integrated management platform includes: a communication protocol quantization coding platform, a gateway plug-in adaptive management and distribution platform, an application zero-code synthesis platform, an application deployment distribution and operation and maintenance platform, and a logic calculation system for communication protocol quantization data. Among them, the communication protocol quantization coding platform is a secondary coding platform for the communication protocols of various devices and systems according to the minimum function. By indiscriminately encoding the chaotic data, the chaotic data is structured to achieve the purpose of structured data, which is the basis for realizing the Internet of Everything. A large number of gateways will be used in the Internet of Everything to complete the conversion from one communication structure to another. It is based on computers and related customized programs. In the construction of smart city projects, it is also faced with frequent changes, which makes the development of gateways extremely heavy. Based on the modular gateway design concept of the APCOS operating system, a gateway plug-in adaptive management and distribution platform is formed. The plug-ins in the gateway plug-in adaptive management and distribution platform are divided into three parts, namely, protocol analysis plug-ins, basic application plug-ins, and plug-in logic processing. These three types of plug-ins can be added by themselves according to project construction. With the accumulation of time, the protocol analysis plug-in has basically been tooled, most of which have been configurable, and the basic plug-in and plug-in logic have basically been directly checked and configurable. Smart city construction faces massive dynamic application development. As the construction time goes by, various applications are isolated from each other, forming difficult problems. The modular synthesis platform of the application system built based on the modular development idea of ​​the APCOS operating system not only realizes zero-code development, but also realizes the rapid development of new application systems with any functional combination (about 10 minutes), in which application modules can be added by dragging and dropping in an ecological way. The application zero-code synthesis platform not only greatly improves efficiency, but also greatly reduces BUGs, thereby greatly reducing the development cost of smart city projects. As the content of smart city project construction increases, it will bring huge pressure on operation and maintenance. How to ensure the healthy operation of hundreds of applications, from the allocation of server resources, server operation status, application operation status, various data source status, various communication operation status to real-time data monitoring of the command and dispatch center, is completed through application deployment distribution and operation and maintenance platforms. The smart city data base based on the Internet of Everything operating system has two key points. One is the structuring of multi-heterogeneous data, and the other is the quantum logical mesh computing system composed of universal kernels. Among them, the quantum logical mesh computing system composed of universal kernels is composed of universal kernels distributed between devices and systems and super universal kernels deployed in data centers, realizing the logical computing system of quantum data and achieving load balancing according to certain rules.

[0062] The basic principle of the present invention is described above in conjunction with specific embodiments. However, it should be pointed out that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and facilitation of understanding, rather than limitation, and the above details do not limit the present invention to being implemented by adopting the above specific details.

[0063] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the module division is only a logical function division, and there may be other division methods in actual implementation. The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0064] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0065] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference to a figure in a claim should not be considered as limiting the claim to which it relates.

[0066] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units stated in the device claim can also be implemented by one unit through software or hardware.

[0067] Finally, it should be noted that the above description has been given for the purpose of illustration and description. In addition, the above embodiments are only used to illustrate the technical solution of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A smart city classification identification management platform based on Internet of Things perception, characterized by: include: The garbage image acquisition module is used to obtain the garbage surface state image of the garbage to be classified collected by the camera; The garbage surface state feature extraction module is used to perform multi-scale feature extraction on the garbage surface state image to obtain shallow features of the garbage state and deep semantic features of the garbage state; the garbage multi-scale feature interactive fusion module is used to perform kernel association feature significant joint interactive coding on the shallow features of the garbage state and the deep semantic features of the garbage state to obtain shallow-deep interactive fusion features of the garbage state, including: a garbage state shallow-deep kernel feature association coding unit, which is used to perform kernel semantic-based association coding on the shallow features of the garbage state and the deep semantic features of the garbage state to obtain shallow-deep kernel features of the garbage state. A global correlation representation between core features; a garbage status shallow-deep feature interaction unit, which is used to interactively fusion encode the garbage status shallow features and the garbage status deep semantic features based on the global correlation representation between the garbage status shallow-deep core features to obtain the garbage status shallow-deep interactive fusion features; a garbage classification management module, which is used to perform garbage identification based on the garbage status shallow-deep interactive fusion features to determine the identification result for representing the garbage type label, and perform garbage classification signal management based on the identification result to confirm whether to generate a garbage classification error prompt signal.

2. According to the smart city classification identification management platform based on Internet of Things perception according to claim 1, it is characterized in that: The garbage surface state feature extraction module includes: a garbage state multi-scale feature extraction unit, which is used to input the garbage surface state image into a garbage state multi-scale feature extractor based on a hollow pyramid network model to obtain a garbage state shallow feature map and a garbage state deep semantic feature map as the garbage state deep semantic feature; a spatial constraint feature enhancement processing unit, which is used to input the garbage state shallow feature map into the spatial constraint feature enhancement module to obtain a garbage state shallow enhanced feature map as the garbage state shallow feature, including: a semantic association constraint score calculation subunit, which is used to perform feature dispersion on the garbage state shallow feature map to obtain a set of garbage state shallow pixel granularity feature vectors, and then perform semantic association constraint score calculation on the set of garbage state shallow pixel granularity feature vectors to obtain a garbage state shallow pixel granularity semantic association space constraint score topological representation; a garbage state shallow feature enhancement subunit, which is used to enhance the garbage state shallow feature map based on the garbage state shallow pixel granularity semantic association space constraint score topological representation to obtain the garbage state shallow enhanced feature map.

3. According to the smart city classification identification management platform based on Internet of Things perception according to claim 2, it is characterized in that: The semantic association constraint score calculation subunit is used to: perform feature dispersion on the garbage status shallow feature map along the channel dimension to obtain a set of garbage status shallow pixel granularity feature vectors; Calculate the semantic association score between any two junk state shallow pixel granularity feature vectors in the set of junk state shallow pixel granularity feature vectors to obtain a junk state shallow pixel granularity semantic association score topological matrix; perform spatial attention attenuation modulation on the junk state shallow pixel granularity semantic association score topological matrix based on the spatial distance between any two junk state shallow pixel granularity feature vectors in the set of junk state shallow pixel granularity feature vectors to obtain a junk state shallow pixel granularity semantic association space soft constraint score topological matrix; perform dilated convolution encoding on the junk state shallow pixel granularity semantic association space soft constraint score topological matrix to obtain a junk state shallow pixel granularity semantic association space soft constraint score topological feature matrix as the junk state shallow pixel granularity semantic association space constraint score topological representation.

4. The smart city classification identification management platform based on Internet of Things perception according to claim 3 is characterized in that: The garbage status shallow feature enhancement subunit is used to: input the set of garbage status shallow feature vectors and the garbage status shallow pixel granularity semantic association space soft constraint score topology matrix into a graph convolutional coding module to obtain a set of garbage status shallow context pixel granularity feature vectors; The set of the junk state shallow context pixel granularity feature vectors is reshaped to obtain the junk state shallow enhanced feature map.

5. The smart city classification identification management platform based on Internet of Things perception according to claim 2 is characterized in that: The garbage status shallow-deep kernel feature association encoding unit includes: a garbage status shallow-deep kernel feature distillation subunit, which is used to input the garbage status shallow enhancement feature map and the garbage status deep semantic feature map into the kernel feature distillation network to obtain a garbage status shallow kernel feature vector and a garbage status deep semantic kernel feature vector; a garbage status shallow-deep kernel feature global encoding subunit, which is used to perform global association encoding on the garbage status shallow kernel feature vector and the garbage status deep semantic kernel feature vector to obtain a garbage status shallow-deep kernel feature global encoding matrix as a garbage status shallow-deep kernel feature global association representation.

6. The smart city classification identification management platform based on Internet of Things perception according to claim 5 is characterized in that: The garbage state shallow-deep kernel feature distillation subunit is used to: perform feature decomposition on the garbage state shallow enhancement feature map to obtain a sequence of garbage state shallow decomposition feature vectors; Calculating the difference feature of each garbage status shallow decomposition feature vector in the sequence of garbage status shallow decomposition feature vectors relative to other garbage status shallow decomposition feature vectors to obtain a garbage status shallow decomposition feature difference representation vector; Normalizing each eigenvalue in the garbage state shallow decomposition feature difference representation vector to obtain a normalized garbage state shallow decomposition feature difference representation vector; Taking each eigenvalue in the normalized garbage status shallow decomposition feature difference representation vector as a weight, a weighted sum of a sequence of the garbage status shallow decomposition feature vectors is calculated to obtain the garbage status shallow kernel feature vector.

7. The smart city classification identification management platform based on Internet of Things perception according to claim 6 is characterized in that: The garbage status shallow-deep kernel feature global encoding subunit is used to calculate the multiplication between the junk status shallow kernel feature vector and the transposed vector of the junk status deep semantic kernel feature vector to obtain the garbage status shallow-deep kernel feature global encoding matrix.

8. The smart city classification identification management platform based on Internet of Things perception according to claim 7 is characterized in that: The garbage status shallow-deep feature interaction unit is used to: input the global encoding matrix between the garbage status shallow-deep kernel features into the sparse mask unit to obtain the associated sparse weight matrix between the garbage status shallow-deep kernel features; Inputting the garbage status shallow enhanced feature map and the garbage status deep semantic feature map into a position-by-position association interaction network to obtain a garbage status shallow-deep position-by-position interaction fusion feature map; Based on the correlation sparse weight matrix between the shallow-deep kernel features of the garbage status, the shallow-deep position-by-position interactive fusion feature map of the garbage status is subjected to feature significant sparse modulation to obtain the shallow-deep significant sparse interactive fusion feature map of the garbage status as the shallow-deep interactive fusion feature of the garbage status.

9. The smart city classification identification management platform based on Internet of Things perception according to claim 8 is characterized in that: The garbage classification management module is used to: input the shallow-deep significant sparse interactive fusion feature map of the garbage state into a classifier-based garbage identifier to obtain a recognition result, and the recognition result is used to represent a garbage type label; Based on the garbage type label and the garbage bin type label, it is determined whether a garbage classification error prompt signal is generated.

10. A smart city classification identification management method based on Internet of Things perception, characterized in that: include: Obtaining the surface state image of the garbage to be classified collected by the camera; Performing multi-scale feature extraction on the garbage surface state image to obtain shallow features of the garbage state and deep semantic features of the garbage state; The shallow features of the garbage status and the deep semantic features of the garbage status are significantly jointly interactively encoded with kernel association features to obtain shallow-deep interactive fusion features of the garbage status, including: performing kernel semantic-based association encoding on the shallow features of the garbage status and the deep semantic features of the garbage status to obtain a global association representation between the shallow-deep kernel features of the garbage status; based on the global association representation between the shallow-deep kernel features of the garbage status, interactive fusion encoding and significant sparse modulation analysis are performed on the shallow features of the garbage status and the deep semantic features of the garbage status to obtain the shallow-deep interactive fusion features of the garbage status; based on the shallow-deep interactive fusion features of the garbage status, garbage identification is performed to determine the identification result for representing the garbage type label, and based on the identification result, garbage classification signal management is performed to confirm whether a garbage classification error prompt signal is generated.

Images