Green ball crack detection system and method based on interpretable space-time atlas

Abstract

The invention discloses a green pellet crack detection system and method based on an interpretable space-time atlas, and relates to the technical field of mineral product detection, and the system comprises a knowledge graph construction module which is used for constructing a knowledge graph of space-time perception green pellets and cracks thereof based on the image data of the preprocessed historical green pellets and cracks thereof; the green ball feature extraction module is used for obtaining a feature data set of screened green balls and cracks thereof by utilizing a knowledge graph of the green balls and the cracks thereof through space-time perception; the detection model construction module is used for training the green ball crack detection initial model to obtain a green ball crack detection model; and the green ball crack detection module is used for carrying out green ball crack detection on the collected real-time green balls and crack data thereof to obtain a green ball crack detection result. The method can solve the problems of low detection efficiency and insufficient reliability of current human subjective factors and a traditional detection method, and ensures the quality of green pellets and the stability of the production process.

Description

Technical Field

[0001] The present invention relates to the technical field of mineral detection, and in particular to a biosphere crack detection system and method based on an interpretable spatiotemporal graph. Background Art

[0002] In the field of mining industry, pellets are an important raw material for modern blast furnace ironmaking. Iron ore powder is mixed with water under the action of a rotating disc pelletizer to form spherical green balls. The green balls are dried, hardened, and cooled to become pellets. The quality of green balls largely affects the quality of finished pellets, so the quality monitoring of green balls during the manufacturing process is of great significance to the quality of finished pellets.

[0003] Green ball quality control is a key link in the pellet production process, among which crack detection is one of the important indicators for evaluating green ball quality. During the transportation and storage process, green balls will fall and collide many times. If the falling strength of the green balls does not meet the production standards, cracks will appear in the green balls before drying and roasting. A large number of cracks will appear in the green balls after roasting or a large amount of powder will be generated during the roasting process, which is very likely to cause the accumulation of powder in the green ball roasting equipment and cause dangerous hidden dangers, and even cause damage to the green ball roasting equipment, which will seriously affect the operation rate, product output and quality indicators.

[0004] At present, in the process of crack detection of green balls, it is necessary for inspectors to identify whether there are cracks on the surface of the green balls with their naked eyes. The manual method is inefficient, labor-intensive, and there are too many subjective factors in the detection process. Some companies use visible light photoelectric sensors or edge detection image processing technology through equipment development and mechanical transformation, but green balls are round particles formed by the capillary force and mechanical force of raw materials such as iron ore concentrate in water, through rotation and extrusion. The surface has not undergone secondary processing such as polishing and smoothing, so its surface is rough and uneven, which is easily mistaken for cracks during the detection process. Due to the characteristics of green balls, the existing crack detection methods cannot be directly used for green ball crack detection.

[0005] In order to eliminate the problems of low detection efficiency and insufficient reliability in the process of green ball crack detection caused by human subjective factors and the limitations of traditional detection methods, technical personnel in this field have been seeking effective detection methods for green ball cracks so as to achieve efficient and accurate automated green ball crack detection and ensure the quality of green balls and the stability of the production process. Summary of the invention

[0006] The purpose of the present invention is to provide a green ball crack detection system and method based on an interpretable spatiotemporal map, which can solve the problems of low detection efficiency and insufficient reliability existing in current human subjective factors and traditional detection methods, and ensure the quality of green balls and the stability of the production process.

[0007] The biosphere crack detection system based on interpretable spatiotemporal graph designed by the present invention to achieve one of the above purposes is special in that it includes:

[0008] A knowledge graph construction module is used to preprocess the image data of the historical biosphere and its cracks collected at different times and locations to obtain the preprocessed image data of the historical biosphere and its cracks, and to construct a knowledge graph of the spatiotemporal perception of the biosphere and its cracks based on the preprocessed image data of the historical biosphere and its cracks;

[0009] The biosphere feature extraction module is used to use the knowledge graph of spatiotemporal perception of biospheres and their cracks to guide the perception attention network to perform feature screening on the image data of historical biospheres and their cracks, and obtain the feature data set of the screened biospheres and their cracks;

[0010] The detection model construction module is used to scan the selected feature data set of green balls and their cracks through a multi-granularity scanning method to obtain a multi-dimensional feature vector of green balls and their cracks; based on the multi-dimensional feature vector of green balls and their cracks, the cascade forest model is used to construct an initial model for green ball crack detection; based on the multi-particle swarm algorithm calculation, the optimal parameters of the initial model for green ball crack detection are obtained; according to the optimal parameters of the initial model for green ball crack detection, the initial model for green ball crack detection is trained to obtain a green ball crack detection model;

[0011] The green ball crack detection module is used to perform green ball crack detection on the collected real-time green ball and its crack data through the green ball crack detection model to obtain the green ball crack detection result.

[0012] Furthermore, the knowledge graph construction module preprocesses the image data of historical green balls and their cracks collected at different times and locations, including: deleting invalid image data and abnormal image data in the collected image data of historical green balls and their cracks or replacing them with smooth interpolation to obtain preprocessed image data of historical green balls and their cracks.

[0013] Furthermore, the knowledge graph construction module constructs a spatiotemporal perception knowledge graph G of the biosphere and its cracks based on the preprocessed image data of the historical biosphere and its cracks, as described in the following formula:

[0014] G={G(1),G(2),…,G(ST)}=(E,R,S,L)

[0015] Among them, E represents the image data of historical green balls and their cracks, R represents the relationship data between green balls and their cracks, S represents the time data of collecting green ball cracks, L represents the location data of collecting green ball cracks, and ST represents the number of image data of green balls and their cracks.

[0016] Furthermore, the biosphere feature extraction module uses the knowledge graph of spatiotemporal perception of biospheres and their cracks to guide the perception attention network to perform feature screening on the image data of historical biospheres and their cracks, as described in the following formula:

[0017]

[0018] Among them, X′ t represents the feature data set of the selected biosphere and its cracks, σ represents the activation function, N represents the image data of the historical biosphere and its cracks collected at different times and locations, and v i Represents the key features in the image data of the historical biosphere and its fractures, v j Represents the key features v in the image data of the historical biosphere and its fractures collected i The adjacent features of ij represents the perceptual attention weight coefficient, W represents the weight parameter matrix, X j Represents the key features v in the image data of the historical biosphere and its fractures collected j The adjacent features v i The characteristic vector of the selected green balls and their cracks includes: the size and shape of the green balls and the size, shape and distribution position of the cracks of the green balls;

[0019] The perceptual attention weight coefficient α ij , as described in the following formula

[0020]

[0021] Among them, Leaky ReLu represents the activation function, α represents the correlation function of the features in the image data of the historical raw balls and their cracks, W represents the weight parameter matrix, X i The feature vector representing the key features in the image data of the historical biosphere and its fractures, X j A feature vector representing neighboring features of key features in the image data of the historical biosphere and its fractures collected.

[0022] Furthermore, the detection model construction module scans the filtered feature data set of green balls and their cracks through a multi-granularity scanning method to obtain feature sub-vectors of green balls and their cracks; and splices the feature sub-vectors of green balls and their cracks to obtain multi-dimensional feature vectors of green balls and their cracks.

[0023] As a preferred solution, the detection model construction module calculates based on the multi-particle swarm algorithm to obtain the optimal parameters of the initial model for detecting biosphere cracks, including: defining a parameter space and an objective function, initializing a particle population based on the defined parameter space, and obtaining initial parameters of the initialized particle population; using the objective function to calculate the initial parameters of the initialized particle population, and obtaining a fitness value of the particle population, as described in the following formula:

[0024] fit′(x′ PSO )=-|δ PSO -d PSO (x′ PSO ,x PSO,m )|

[0025] Among them, fit′(x′ PSO ) represents the fitness value of the additional population, x′ PSO represents the individual particles of the additional population, δ PSO represents the distance between the main population and the additional population, x PSO,m represents the optimal solution of the main population, d PSO (x′ PSO ,x PSO,m ) represents the individual particle x′ of the additional population PSO and the optimal solution x of the main population PSO,m The distance between the main population; the fitness value of the main population is fit(x PSo ) is calculated using the following formula:

[0026] fit(x PSO )=-f PSO (x PSO )

[0027] Among them, x PSO represents the individual particles of the main population, f PSO represents the optimization target of the main population; the initialized particle population is updated according to the fitness value, and the position of the particle is iterated according to the following formula:

[0028] P i =Nx besti +(1-N)x gbest

[0029] Among them, P i represents the updated value of the i-th particle, Nx besti represents the historical optimal value of the i-th particle, x gbestrepresents the global optimal value of the entire particle population, N represents the balance factor of the optimal value of the individual particle and the optimal value of the entire particle population for particle update; repeatedly execute the particle position update until a preset number of iterations is reached or the fitness value meets the preset requirements to obtain the final position; the position of the final particle position corresponding to the global optimal value of the entire particle population is used as the parameter of the optimal cascade forest model, thereby improving the accuracy and reliability of different types of biosphere crack detection.

[0030] Furthermore, the detection model construction module trains the initial model for detecting cracks in the biosphere according to the parameters of the optimal cascade forest model obtained by the multi-particle swarm algorithm, including: each layer of the cascade forest is composed of multiple random forests, each layer of the cascade forest receives the multi-dimensional feature vector of the biosphere and its cracks processed by the previous layer of the cascade forest, and performs further feature transformation of the biosphere and its cracks through the random forest inside it to obtain a new multi-dimensional feature vector of the biosphere and its cracks, and outputs the processed new multi-dimensional feature vector of the biosphere and its cracks to the next layer of the cascade forest; each layer of the cascade forest after the second layer of the cascade forest concatenates the multi-dimensional feature vector of the biosphere and its cracks output by the previous layer with the multi-dimensional feature vector of the original biosphere and its cracks as the input of this layer; the last layer of the cascade forest averages and integrates the multi-dimensional feature vectors of the biosphere and its cracks generated to obtain a biosphere crack detection model.

[0031] The biosphere crack detection method based on interpretable spatiotemporal graph designed by the present invention to achieve the second purpose above is special in that it includes the following steps:

[0032] Preprocess the image data of historical biospheres and their cracks collected at different times and locations to obtain the preprocessed image data of historical biospheres and their cracks, and construct a knowledge graph of spatiotemporal perception of biospheres and their cracks based on the preprocessed image data of historical biospheres and their cracks;

[0033] Using the knowledge graph of spatiotemporal perception of the life sphere and its cracks, the perception attention network is guided to perform feature screening on the image data of the historical life sphere and its cracks, and the feature dataset of the screened life sphere and its cracks is obtained;

[0034] Through the multi-granularity scanning method, the characteristic data set of the screened green balls and their cracks is scanned to obtain the multi-dimensional characteristic vectors of the green balls and their cracks; based on the multi-dimensional characteristic vectors of the green balls and their cracks, the initial model of green ball crack detection is constructed using the cascade forest model; based on the multi-particle swarm algorithm calculation, the optimal parameters of the initial model of green ball crack detection are obtained; according to the optimal parameters of the initial model of green ball crack detection, the initial model of green ball crack detection is trained to obtain the green ball crack detection model;

[0035] Through the green ball crack detection model, the collected real-time green ball and its crack data are subjected to green ball crack detection to obtain the green ball crack detection result.

[0036] A computing device designed by the present invention to achieve the third objective above includes a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0037] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned biosphere crack detection method based on interpretable spatiotemporal graph.

[0038] In order to achieve the fourth purpose of the present invention, a computer storage medium is designed, in which executable instructions are stored, and the executable instructions enable the processor to execute operations corresponding to the above-mentioned biosphere crack detection method based on interpretable spatiotemporal graph.

[0039] The present invention has the following beneficial effects:

[0040] (1) The biosphere crack detection system and method based on interpretable spatiotemporal graphs achieves a comprehensive characterization of biosphere crack data in time and space dimensions by constructing a spatiotemporal-aware biosphere crack knowledge graph. It can not only capture the static characteristics of biosphere cracks, but also track their dynamic changes. The knowledge graph improves the interpretability of the model, helps to deeply understand the process of biosphere crack formation, and provides reliable decision support for the optimization of production processes. Furthermore, the construction of the spatiotemporal graph provides structured prior knowledge for subsequent biosphere feature extraction and detection model training, thereby improving the accuracy of biosphere crack detection.

[0041] (2) The ship-shore data lightweight communication system and method based on the hybrid compression algorithm improves the ability to capture the multi-dimensional features of green balls through multi-granularity scanning, and can simultaneously identify small cracks and large cracks in green balls. Furthermore, the design of the multi-layer cascade forest improves the processing speed of detection and meets the needs of real-time detection in the production process. The particle swarm algorithm realizes the automatic optimization of model parameters and reduces the workload of manual parameter adjustment. This intelligent and adaptive green ball feature extraction can adapt to green balls produced from different batches and different raw materials, significantly improving the versatility and stability of detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A module schematic diagram of a specific embodiment of a biosphere crack detection system based on an interpretable spatiotemporal graph of the present invention is shown.

[0043] Figure 2A specific embodiment of a biosphere crack detection system based on an interpretable spatiotemporal graph of the present invention is shown, including a schematic diagram of a detection model construction module.

[0044] Figure 3 A flow chart showing a specific embodiment of a biosphere crack detection method based on an interpretable spatiotemporal graph of the present invention. DETAILED DESCRIPTION

[0045] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0046] like Figures 1 to 3 As shown, an embodiment of the present invention discloses a green ball crack detection system and method based on an interpretable spatiotemporal map, which can meet the needs of high-efficiency and high-accuracy automated green ball crack detection and ensure the quality of green balls and the stability of the production process.

[0047] Example 1

[0048] This embodiment discloses a biosphere crack detection system based on an interpretable spatiotemporal graph, the system comprising:

[0049] A knowledge graph construction module is used to preprocess the image data of the historical biosphere and its cracks collected at different times and locations to obtain the preprocessed image data of the historical biosphere and its cracks, and to construct a knowledge graph of the spatiotemporal perception of the biosphere and its cracks based on the preprocessed image data of the historical biosphere and its cracks;

[0050] The biosphere feature extraction module is used to use the knowledge graph of spatiotemporal perception of biospheres and their cracks to guide the perception attention network to perform feature screening on the image data of historical biospheres and their cracks, and obtain the feature data set of the screened biospheres and their cracks;

[0051] The detection model construction module is used to scan the selected feature data set of green balls and their cracks through a multi-granularity scanning method to obtain a multi-dimensional feature vector of green balls and their cracks; based on the multi-dimensional feature vector of green balls and their cracks, the cascade forest model is used to construct an initial model for green ball crack detection; based on the multi-particle swarm algorithm calculation, the optimal parameters of the initial model for green ball crack detection are obtained; according to the optimal parameters of the initial model for green ball crack detection, the initial model for green ball crack detection is trained to obtain a green ball crack detection model;

[0052] The green ball crack detection module is used to perform green ball crack detection on the collected real-time green ball and its crack data through the green ball crack detection model to obtain the green ball crack detection result.

[0053] Specifically, the knowledge graph construction module preprocesses the image data of historical biospheres and their cracks collected at different times and locations, including: deleting invalid image data and abnormal image data in the collected image data of historical biospheres and their cracks, or replacing them with smooth interpolation to obtain preprocessed image data of historical biospheres and their cracks.

[0054] In this embodiment, the knowledge graph construction module constructs a knowledge graph G of spatiotemporal perception of the biosphere and its cracks based on the preprocessed image data of the historical biosphere and its cracks, as described in the following formula:

[0055] G={G(1),G(2),…,G(ST)}=(E,R,S,L)

[0056] Among them, E represents the image data of historical green balls and their cracks, R represents the relationship data between green balls and their cracks, S represents the time data of collecting green ball cracks, L represents the location data of collecting green ball cracks, and ST represents the number of image data of green balls and their cracks.

[0057] In this embodiment, the biosphere feature extraction module uses the knowledge graph of spatiotemporal perception of biospheres and their cracks to guide the perception attention network to perform feature screening on the image data of historical biospheres and their cracks, as described in the following formula:

[0058]

[0059] Among them, X′ t represents the feature data set of the selected biosphere and its cracks, σ represents the activation function, N represents the image data of the historical biosphere and its cracks collected at different times and locations, and v i Represents the key features in the image data of the historical biosphere and its fractures, v j Represents the key features v in the image data of the historical biosphere and its fractures collected i The adjacent features of ij represents the perceptual attention weight coefficient, W represents the weight parameter matrix, X j Represents the key features v in the image data of the historical biosphere and its fractures collected j The adjacent features v i The characteristic vector of the green ball and its cracks after screening includes: the size and shape of the green ball and the size, shape and distribution position of the cracks of the green ball;

[0060] Perceptual attention weight coefficient α ij , as described in the following formula

[0061]

[0062] Among them, Leaky ReLu represents the activation function, α represents the correlation function of the features in the image data of the historical raw balls and their cracks, W represents the weight parameter matrix, X i The feature vector representing the key features in the image data of the historical biosphere and its fractures, X j A feature vector representing neighboring features of key features in the image data of the historical biosphere and its fractures collected.

[0063] The detection model construction module uses a multi-granularity scanning method to scan the filtered feature data set of the green ball and its cracks to obtain the feature sub-vectors of the green ball and its cracks; the feature sub-vectors of the green ball and its cracks are spliced ​​to obtain the multi-dimensional feature vectors of the green ball and its cracks.

[0064] In this embodiment, the detection model construction module is based on the multi-particle swarm algorithm to calculate and obtain the optimal parameters of the initial model for ball crack detection, including: defining a parameter space and an objective function, initializing a particle population based on the defined parameter space, and obtaining the initial parameters of the initialized particle population; using the objective function to calculate the initial parameters of the initialized particle population, and obtaining the fitness value of the particle population, which is specifically described in the following formula:

[0065] fit′(x′ PSO )=-|δ PSO -d PSO (x′ PSO ,x PSO,m )|

[0066] Among them, fit′(x′ PSO ) represents the fitness value of the additional population, x′ PSO represents the individual particles of the additional population, δ PSO represents the distance between the main population and the additional population, x PSO,m represents the optimal solution of the main population, d PSO (x′ PSO ,x PSO,m ) represents the individual particle x′ of the additional population PSO and the optimal solution x of the main population PSO,m The distance between the main population; the fitness value of the main population is fit(x PSO ) is calculated using the following formula:

[0067] fit(x PSO )=-f PSO (x PSO )

[0068] Among them, x PSOrepresents the individual particles of the main population, f PSO Represents the optimization goal of the main population; the initialized particle population is updated according to the fitness value, and the position of the particle is iterated according to the following formula:

[0069] P i =Nx besti +(1-N)x gbest

[0070] Among them, P i represents the updated value of the i-th particle, Nx besti represents the historical optimal value of the i-th particle, x gbest represents the global optimal value of the entire particle population, N represents the balance factor of the optimal value of the individual particle and the optimal value of the entire particle population for particle update; the particle position update is repeated until the preset number of iterations is reached or the fitness value meets the preset requirements to obtain the final position; the position of the final particle position corresponding to the global optimal value of the entire particle population is used as the parameter of the optimal cascade forest model, thereby improving the accuracy and reliability of different types of biosphere crack detection.

[0071] In this embodiment, the detection model construction module trains the initial model for detecting cracks in the biosphere according to the parameters of the optimal cascade forest model obtained by the multi-particle swarm algorithm, including: each layer of the cascade forest is composed of multiple random forests, each layer of the cascade forest receives the multi-dimensional feature vector of the biosphere and its cracks processed by the previous layer of the cascade forest, and performs further feature transformation of the biosphere and its cracks through the random forest inside it to obtain a new multi-dimensional feature vector of the biosphere and its cracks, and outputs the processed new multi-dimensional feature vector of the biosphere and its cracks to the next layer of the cascade forest; each layer of the cascade forest after the second layer of the cascade forest concatenates the multi-dimensional feature vector of the biosphere and its cracks output by the previous layer with the multi-dimensional feature vector of the original biosphere and its cracks as the input of this layer; the last layer of the cascade forest averages and integrates the multi-dimensional feature vectors of the biosphere and its cracks generated to obtain a biosphere crack detection model.

[0072] Example 2

[0073] This embodiment discloses a method for detecting biosphere cracks based on an interpretable spatiotemporal graph, the method comprising the following steps:

[0074] Step 1: preprocess the image data of the historical biosphere and its cracks collected at different times and locations to obtain the preprocessed image data of the historical biosphere and its cracks, and construct a knowledge graph of the spatiotemporal perception of the biosphere and its cracks based on the preprocessed image data of the historical biosphere and its cracks;

[0075] Step 2: Using the knowledge graph of spatiotemporal perception of the biosphere and its cracks, guide the perception attention network to perform feature screening on the image data of the historical biosphere and its cracks, and obtain the feature data set of the screened biosphere and its cracks;

[0076] Step 3, by using a multi-granularity scanning method, the characteristic data set of the screened green balls and their cracks is scanned to obtain a multi-dimensional characteristic vector of the green balls and their cracks; based on the multi-dimensional characteristic vector of the green balls and their cracks, an initial model for green ball crack detection is constructed using a cascade forest model; based on the multi-particle swarm algorithm calculation, the optimal parameters of the initial model for green ball crack detection are obtained; according to the optimal parameters of the initial model for green ball crack detection, the initial model for green ball crack detection is trained to obtain a green ball crack detection model;

[0077] Step 4: Using the green ball crack detection model, the collected real-time green ball and its crack data are subjected to green ball crack detection to obtain the green ball crack detection result;

[0078] Example 3

[0079] This embodiment discloses a schematic diagram of the structure of a computing device, and the specific embodiment of the present invention does not limit the specific implementation of the computing device.

[0080] The computing device may include: a processor, a communication interface, a memory, and a communication bus.

[0081] Wherein: the processor, the communication interface, and the memory communicate with each other via a communication bus.

[0082] Communication interface, used to communicate with other devices such as client or other server network elements.

[0083] The processor is used to execute the program, and specifically can execute the relevant steps in the above-mentioned embodiment of the biosphere crack detection method based on the interpretable spatiotemporal map.

[0084] Specifically, the program may include program codes including computer operation instructions.

[0085] The processor may be a central processing unit (CPU), or a specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. The one or more processors included in the computing device may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0086] The memory is used to store programs. The memory may include a high-speed RAM memory, and may also include a non-volatile memory, such as at least one disk memory.

[0087] The program can be specifically used to enable the processor to execute the biosphere crack detection method based on the interpretable spatiotemporal map in any of the above-mentioned method embodiments. The specific implementation of each step in the program can refer to the corresponding descriptions in the corresponding steps and units in the above-mentioned ship-shore data lightweight communication system and method embodiment based on the hybrid compression algorithm, which will not be repeated here. The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, the specific working process of the above-described equipment and modules can refer to the corresponding process description in the above-mentioned method embodiment, which will not be repeated here.

[0088] The algorithm and display provided herein are not inherently related to any particular computer, virtual system or other device. Various general purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious that the structure required for constructing such systems. In addition, the present invention is not directed to any specific programming language either. It should be understood that various programming languages ​​can be utilized to realize the content of the present invention described herein, and the description of the above specific languages ​​is for disclosing the best mode of the present invention.

[0089] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0090] Similarly, it should be understood that in order to streamline the present disclosure and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting the intention that the claimed invention requires more features than those expressly recited in each claim. Rather, as reflected in the claims, inventive aspects lie in less than all of the features of the individual embodiments previously disclosed. Therefore, the claims that follow the detailed description are hereby expressly incorporated into the detailed description, with each claim itself serving as a separate embodiment of the present invention.

Claims

1. A biosphere crack detection system based on interpretable spatiotemporal graph, characterized in that: include: A knowledge graph construction module is used to preprocess the image data of the historical biosphere and its cracks collected at different times and locations to obtain the preprocessed image data of the historical biosphere and its cracks, and to construct a knowledge graph of the spatiotemporal perception of the biosphere and its cracks based on the preprocessed image data of the historical biosphere and its cracks; The biosphere feature extraction module is used to use the knowledge graph of spatiotemporal perception of biospheres and their cracks to guide the perception attention network to perform feature screening on the image data of historical biospheres and their cracks, and obtain the feature data set of the screened biospheres and their cracks; The detection model construction module is used to scan the selected feature data set of green balls and their cracks through a multi-granularity scanning method to obtain a multi-dimensional feature vector of green balls and their cracks; based on the multi-dimensional feature vector of green balls and their cracks, the cascade forest model is used to construct an initial model for green ball crack detection; based on the multi-particle swarm algorithm calculation, the optimal parameters of the initial model for green ball crack detection are obtained; according to the optimal parameters of the initial model for green ball crack detection, the initial model for green ball crack detection is trained to obtain a green ball crack detection model; The green ball crack detection module is used to perform green ball crack detection on the collected real-time green ball and its crack data through the green ball crack detection model to obtain the green ball crack detection result.

2. The biosphere crack detection system based on interpretable spatiotemporal graph according to claim 1 is characterized in that: The knowledge graph construction module preprocesses the image data of historical green balls and their cracks collected at different times and locations, including: deleting invalid image data and abnormal image data in the collected image data of historical green balls and their cracks or replacing them with smooth interpolation to obtain preprocessed image data of historical green balls and their cracks.

3. The biosphere crack detection system based on interpretable spatiotemporal graph according to claim 1 is characterized in that: The knowledge graph construction module constructs a knowledge graph G of spatiotemporal perception of the biosphere and its cracks based on the preprocessed image data of the historical biosphere and its cracks, as described in the following formula: G={G(1),G(2),…,G(ST)}=(E,R,S,L) Among them, E represents the image data of historical green balls and their cracks, R represents the relationship data between green balls and their cracks, S represents the time data of collecting green ball cracks, L represents the location data of collecting green ball cracks, and ST represents the number of image data of green balls and their cracks.

4. The biosphere crack detection system based on interpretable spatiotemporal graph according to claim 1 is characterized in that: The biosphere feature extraction module uses the knowledge graph of spatiotemporal perception of biospheres and their cracks to guide the perception attention network to perform feature screening on the image data of historical biospheres and their cracks, as described in the following formula: Among them, X t ′ represents the feature data set of the selected green balls and their cracks, σ represents the activation function, N represents the image data of the historical green balls and their cracks collected at different times and locations, and v i Represents the key features in the image data of the historical biosphere and its fractures, v j Represents the key features v in the image data of the historical biosphere and its fractures collected i The adjacent features of ij represents the perceptual attention weight coefficient, W represents the weight parameter matrix, X j Represents the key features v in the image data of the historical biosphere and its fractures collected j The adjacent features v i The characteristic vector of the selected green balls and their cracks includes: the size and shape of the green balls and the size, shape and distribution position of the green ball cracks; The perceptual attention weight coefficient α ij , as described in the following formula Among them, Leaky ReLu represents the activation function, α represents the correlation function of the features in the image data of the historical raw balls and their cracks, W represents the weight parameter matrix, X i The feature vector representing the key features in the image data of the historical biosphere and its fractures, X j A feature vector representing neighboring features of key features in the image data of the historical biosphere and its fractures collected.

5. The biosphere crack detection system based on interpretable spatiotemporal graph according to claim 1 is characterized in that: The detection model building module scans the selected feature data set of green balls and their cracks through a multi-granularity scanning method to obtain feature sub-vectors of green balls and their cracks; and splices the feature sub-vectors of green balls and their cracks to obtain multi-dimensional feature vectors of green balls and their cracks.

6. The biosphere crack detection system based on interpretable spatiotemporal graph according to claim 1 is characterized in that: The detection model building module calculates based on the multi-particle swarm algorithm to obtain the optimal parameters of the initial model for ball crack detection, including: defining a parameter space and an objective function, initializing a particle population based on the defined parameter space, and obtaining initial parameters of the initialized particle population; using the objective function to calculate the initial parameters of the initialized particle population, and obtaining a fitness value of the particle population, as described in the following formula: fit′(x′ PSO )=-|δ PSO -d PSO (x′ PSO ,x PSO,m )| Among them, fit′(x′ PSO ) represents the fitness value of the additional population, x′ PSO represents the individual particles of the additional population, δ PSO represents the distance between the main population and the additional population, x PSO,m represents the optimal solution of the main population, d PSO (x′ PSO ,x PSO,m ) represents the individual particle x′ of the additional population PSO and the optimal solution x of the main population PSO,m The distance between the main population; the fitness value of the main population is fit(x PSO ) is calculated using the following formula: fit(x PSO )=-f PSO (x PSO ) Among them, x PSO represents the individual particles of the main population, f PSO represents the optimization target of the main population; the initialized particle population is updated according to the fitness value, and the position of the particle is iterated according to the following formula: P i =Nx besti +(1-N)x gbest Among them, P i represents the updated value of the i-th particle, Nx besti represents the historical optimal value of the i-th particle, x gbest represents the global optimal value of the entire particle population, N represents the balance factor of the optimal value of the individual particle and the optimal value of the entire particle population for particle update; repeatedly execute the particle position update until the preset number of iterations is reached or the fitness value meets the preset requirements to obtain the final position; the position of the final particle position corresponding to the global optimal value of the entire particle population is used as the parameter of the optimal cascade forest model.

7. The biosphere crack detection system based on interpretable spatiotemporal graph according to claim 1 is characterized in that: The detection model construction module trains the initial model for detecting cracks in the ball according to the parameters of the optimal cascade forest model obtained by the multi-particle swarm algorithm, including: each layer of cascade forest is composed of multiple random forests, each layer of cascade forest receives the multi-dimensional feature vector of the ball and its cracks processed by the previous layer of cascade forest, and further converts the features of the ball and its cracks through the random forest inside it to obtain a new multi-dimensional feature vector of the ball and its cracks, and outputs the processed new multi-dimensional feature vector of the ball and its cracks to the next layer of cascade forest; each layer of cascade forest after the second layer of cascade forest splices the multi-dimensional feature vector of the ball and its cracks output by the previous layer with the multi-dimensional feature vector of the original ball and its cracks as the input of this layer; the last layer of cascade forest averages and integrates the multi-dimensional feature vectors of the ball and its cracks generated to obtain a detection model for cracks in the ball.

8. A method for detecting biosphere cracks based on interpretable spatiotemporal graphs, characterized in that: The steps include: Preprocess the image data of historical biospheres and their cracks collected at different times and locations to obtain preprocessed image data of historical biospheres and their cracks, and construct a knowledge graph of spatiotemporal perception of biospheres and their cracks based on the preprocessed image data of historical biospheres and their cracks; Using the knowledge graph of spatiotemporal perception of the biosphere and its cracks, the perception attention network is guided to perform feature screening on the image data of the historical biosphere and its cracks, and the feature dataset of the screened biosphere and its cracks is obtained; Through the multi-granularity scanning method, the characteristic data set of the screened green balls and their cracks is scanned to obtain the multi-dimensional characteristic vectors of the green balls and their cracks; based on the multi-dimensional characteristic vectors of the green balls and their cracks, the initial model of green ball crack detection is constructed using the cascade forest model; based on the multi-particle swarm algorithm calculation, the optimal parameters of the initial model of green ball crack detection are obtained; according to the optimal parameters of the initial model of green ball crack detection, the initial model of green ball crack detection is trained to obtain the green ball crack detection model; Through the green ball crack detection model, the collected real-time green ball and its crack data are subjected to green ball crack detection to obtain the green ball crack detection result.

9. A computing device comprising a processor, a memory, a communication interface and a communication bus, characterized in that: The processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the biosphere crack detection method based on interpretable spatiotemporal graph as described in claim 9.

10. A computer storage medium, wherein at least one executable instruction is stored in the computer storage medium, wherein the executable instruction enables a processor to perform operations corresponding to the biosphere crack detection method based on interpretable spatiotemporal graph as described in claim 9.

Images