Mine blasting monitoring image abnormity identification method and system

Abstract

The invention provides a mine blasting monitoring image abnormity identification method and system, and the method comprises the steps: obtaining the rock mass humidity distribution topology and the rock mass temperature distribution topology of a mine, carrying out the hardness partitioning through combining with the rock mass type distribution topology, and obtaining the rock mass hardness distribution topology; by taking the rock mass type distribution topology, the rock mass fracture distribution topology and the rock mass hardness distribution topology as constraints, performing blasting sample collection on the blasting position list and the blasting stage number list to obtain a blasting sample set, and executing high-frequency muck pile form excavation to obtain a muck pile reference form; and when the muck pile monitoring form is inconsistent with the muck pile reference form, uncontrollable identification is performed on the muck pile monitoring form, and the muck pile monitoring form is sent to the user side. According to the method and the device, the technical problem that the mine blasting monitoring image is difficult to accurately judge the abnormality in the prior art is solved, and the technical effects of accurately recognizing the abnormality of the mine blasting monitoring image and timely warning are achieved.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and particularly to a method and system for abnormally identifying mine blasting monitoring images. Background Art

[0002] With the continuous expansion of the scale of mine exploitation, mine blasting, as an efficient rock fragmentation method, is widely used in the process of mineral resource exploitation. Mine blasting monitoring data is of great significance for ensuring blasting safety and improving blasting efficiency. However, the dimensions of mine blasting monitoring data are numerous, including various parameters such as vibration, noise, temperature, and humidity, and there are complex mutual influence relationships among these parameters; at the same time, the factors affecting the mine blasting effect show strong non-linear characteristics. For example, rock mass types, fracture distributions, hardness distributions, etc. all have significant effects on the blasting effect. Existing technologies mostly analyze from a single dimension, such as only focusing on vibration parameters or only analyzing temperature changes, and cannot comprehensively consider the overall state of mine blasting. Therefore, abnormal states cannot be effectively identified in existing mine blasting monitoring technologies. Summary of the Invention

[0003] The present invention aims at the technical problem that it is difficult to accurately identify abnormalities in mine blasting monitoring images in the prior art, and provides a method and system for abnormally identifying mine blasting monitoring images to solve this problem.

[0004] The technical solutions of the present invention for solving the above technical problems are as follows: In a first aspect, the present invention provides a method for abnormally identifying mine blasting monitoring images, including: obtaining the topological distribution of rock mass humidity and the topological distribution of rock mass temperature in a mine, performing hardness zoning in combination with the topological distribution of rock mass types to obtain the topological distribution of rock mass hardness; taking the topological distribution of rock mass types, the topological distribution of rock mass fractures, and the topological distribution of rock mass hardness as constraints, collecting blasting samples from a blasting position list and a blasting stage list to obtain a blasting sample set, performing high-frequency muck pile morphology mining to obtain a reference muck pile morphology; collecting a mine muck pile monitoring image through an image acquisition device, performing image feature extraction to obtain a muck pile monitoring morphology; when the muck pile monitoring morphology is inconsistent with the reference muck pile morphology, sending an uncontrollable identification of the mine muck pile monitoring image to the user terminal.

[0005] In a second aspect, the present invention provides a system for abnormally identifying monitoring images of mine blasting, comprising: a rock mass characteristic analysis module, configured to obtain the topological distribution of the humidity of the rock mass in the mine and the topological distribution of the temperature of the rock mass, perform hardness zoning in combination with the topological distribution of the rock mass type, and obtain the topological distribution of the rock mass hardness; a reference shape acquisition module, configured to collect blasting samples from the blasting position list and the blasting level list with the topological distribution of the rock mass type, the topological distribution of the rock mass fissures, and the topological distribution of the rock mass hardness as constraints, obtain a blasting sample set, and perform high-frequency muck pile shape mining to obtain the reference shape of the muck pile; a monitoring shape acquisition module, configured to collect mine muck pile monitoring images through an image acquisition device, perform image feature extraction, and obtain the monitored shape of the muck pile; and an image identification module, configured to send an uncontrollable identification of the mine muck pile monitoring image to the user terminal when the monitored shape of the muck pile is inconsistent with the reference shape of the muck pile.

[0006] The beneficial effects of the present invention are as follows: The topological distribution of the humidity of the rock mass in the mine and the topological distribution of the temperature of the rock mass are obtained, hardness zoning is performed in combination with the topological distribution of the rock mass type, and the topological distribution of the rock mass hardness is obtained, establishing a multi-dimensional characteristic data set of the mine rock mass, providing a basic parameter framework for subsequent blasting monitoring, and thus being able to fully consider the influence of multi-dimensional factors such as rock mass humidity, temperature, and hardness on the blasting effect; with the topological distribution of the rock mass type, the topological distribution of the rock mass fissures, and the topological distribution of the rock mass hardness as constraints, blasting samples are collected from the blasting position list and the blasting level list, a blasting sample set is obtained, high-frequency muck pile shape mining is performed, and the reference shape of the muck pile is obtained. By comprehensively considering the multi-dimensional characteristics of the rock mass, a blasting sample database constrained by multiple dimensions is established, and the common characteristics in the sample set are analyzed through high-frequency muck pile shape mining technology to form the reference shape of the muck pile, providing a reference basis for judging whether the blasting is normal; mine muck pile monitoring images are collected through an image acquisition device, image feature extraction is performed, and the monitored shape of the muck pile is obtained, realizing the automatic acquisition and analysis of the monitored shape of the muck pile from the actual blasting site, efficiently and accurately obtaining the monitored shape of the muck pile, and providing support for judging whether the mine muck pile monitoring image is abnormal; when the monitored shape of the muck pile is inconsistent with the reference shape of the muck pile, an uncontrollable identification of the mine muck pile monitoring image is sent to the user terminal, realizing intelligent identification and early warning of the abnormal state of the mine blasting monitoring image, providing an intuitive basis and decision-making support for the safety assessment of blasting operations, and thus realizing automatic identification and early warning of blasting anomalies and avoiding the occurrence of safety accidents.

[0007] Through the above technical solutions, intelligent identification and automatic early warning of the abnormal state of mine blasting monitoring images are realized, the technical problem that it is difficult to accurately identify anomalies in existing mine blasting monitoring images is solved, and the accuracy of mine blasting monitoring image analysis and the safety of blasting operations are improved. Description of the Drawings

[0008] Figure 1 Schematic flow diagram of a method for abnormally identifying mine blasting monitoring images provided by the present invention; Figure 2 Schematic structural diagram of a system for abnormally identifying mine blasting monitoring images provided by the present invention.

[0009] In the accompanying drawings, the components represented by the reference numerals are as follows: Rock mass characteristic analysis module 11, reference shape acquisition module 12, monitored shape acquisition module 13, image identification module 14. Specific implementation manners

[0010] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts fall within the protection scope of the present invention.

[0011] In the description of the present invention, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present invention, "a plurality of" means two or more unless otherwise specifically defined.

[0012] In the description of the present invention, the term "for example" is used to mean "serving as an example, illustration, or explanation". Any embodiment described as "for example" in the present invention is not necessarily construed as being more preferred or having more advantages than other embodiments. In order for any person skilled in the art to implement and use the present invention, the following description is given. In the following description, details are set forth for the purpose of explanation. It should be understood that those skilled in the art can recognize that the present invention can be implemented without using these specific details. In other instances, well-known structures and processes are not elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope that conforms to the principles and features disclosed in the present invention.

[0013] Embodiment 1, as Figure 1 shown, the embodiment of the present invention provides a method for abnormally identifying mine blasting monitoring images, including: S100: Obtain the topological distribution of the rock mass humidity and the topological distribution of the rock mass temperature of the mine, combine the topological distribution of the rock mass type for hardness zoning, and obtain the topological distribution of the rock mass hardness.

[0014] Specifically, first, obtain the topological data of the rock mass humidity distribution and the topological data of the rock mass temperature distribution in the mine through sensor networks or on-site survey means. Among them, the topological rock mass humidity distribution refers to the spatial distribution structure of the water content of the rock mass in each area of the mine, which directly affects the physical properties of the rock mass; the topological rock mass temperature distribution refers to the spatial distribution structure of the rock mass temperature in each area of the mine, and temperature changes will cause changes in the internal stress state of the rock mass.

[0015] Then, combine the topological rock mass humidity distribution and the topological rock mass temperature distribution with the known topological rock mass type distribution, and through comprehensively considering the influence of these three factors on the hardness of the rock mass, realize the hardness zoning of the rock mass in the mine. Among them, the topological rock mass type distribution refers to the spatial distribution structure of different rock types (such as granite, limestone, shale, etc.) in the mine, and different rock mass types have different basic hardness characteristics. Through the comprehensive analysis of multi-dimensional factors, the problem of deviation in hardness evaluation caused by only considering the rock mass type in the traditional method is overcome, so as to obtain an accurate topological rock mass hardness distribution, laying a foundation for the subsequent identification of abnormal blasting monitoring data.

[0016] The hardness of the rock mass directly affects the blasting effect, such as the stacking shape after blasting, the particle size distribution of the crushed stones, etc. Therefore, the obtained topological rock mass hardness distribution provides a parameter basis for determining the reference shape of the blasted muck pile in the subsequent stage and provides data support for identifying abnormalities in the blasting monitoring data.

[0017] S200: Taking the topological rock mass type distribution, the topological rock mass fracture distribution, and the topological rock mass hardness distribution as constraints, conduct blasting sample collection on the blasting position list and the blasting level list to obtain a blasting sample set, and perform high-frequency blasted muck pile shape mining to obtain the reference shape of the blasted muck pile.

[0018] Specifically, first, taking the topological rock mass type distribution, the topological rock mass fracture distribution, and the topological rock mass hardness distribution as constraint conditions, conduct targeted blasting sample collection on the predetermined blasting position list and the blasting level list. Among them, the topological rock mass fracture distribution refers to the spatial distribution structure of discontinuous surfaces such as fractures, faults, and joints in the mine rock mass, which directly affects the transmission efficiency of blasting energy and the fragmentation of the rock mass after blasting. By taking the three topological rock mass characteristics as constraint conditions, the precise screening of blasting samples is realized, ensuring the representativeness and effectiveness of the collected samples.

[0019] The blasting location list contains the coordinate information of multiple blasting points in the mine; the blasting level list contains the blasting intensity level data corresponding to each blasting point. By collecting blasting samples for the blasting location list and blasting level list of the current mine blasting, a blasting sample set containing a variety of blasting conditions and results is obtained. After obtaining the blasting sample set, high-frequency blast pile morphology mining is further performed to extract the most representative blast pile characteristic parameter combination from a large number of blasting samples to form a blast pile benchmark morphology. The blast pile benchmark morphology is a description of the blast pile morphology characteristics under specific rock mass conditions and blasting parameters, including but not limited to key parameters such as blast pile height, width, length, slope angle, etc.

[0020] Through high-frequency morphological mining based on multi-dimensional constraints, the problem that traditional single-dimensional analysis cannot comprehensively consider the overall status of mine blasting is effectively solved, providing a reliable reference standard for subsequent blasting abnormality monitoring.

[0021] S300: using an image acquisition device to collect a monitoring image of a mine explosion pile, performing image feature extraction, and obtaining an explosion pile monitoring morphology; Specifically, after the construction of the benchmark morphology of the blast pile is completed, the blast pile formed after the mine blasting is monitored in real time through the image acquisition device deployed on site. The image acquisition device includes but is not limited to high-definition cameras, optical sensors carried by drones, infrared thermal imagers and other multi-source image acquisition devices. By arranging these devices, multi-angle and all-round monitoring of the mine blast pile is achieved to obtain the mine blast pile monitoring image.

[0022] After preprocessing, the collected mine blast pile monitoring images enter the image feature extraction stage. The feature extraction process first performs image segmentation to accurately separate the blast pile area from the background environment; then the geometric contour of the blast pile is identified through the edge detection algorithm; then the key morphological features of the blast pile are extracted using a deep learning model or a traditional computer vision algorithm, including parameters such as blast pile height, blast pile width, blast pile length, blast pile slope angle, and accumulation morphological symmetry. In addition, physical state parameters such as looseness and particle size distribution of the blast pile surface can also be extracted through texture analysis. The image feature extraction process adopts a multi-scale analysis method, which not only focuses on the overall morphological characteristics of the blast pile, but also captures local detail changes. By constructing a feature vector space, a comprehensive and accurate description of the blast pile monitoring morphology is achieved. The extracted blast pile monitoring morphology is a digital expression of the geometric characteristics and physical state of the blast pile, laying the foundation for subsequent comparative analysis with the blast pile benchmark morphology.

[0023] By combining image acquisition with feature extraction, the subjectivity and inefficiency of traditional manual observation methods are overcome, and objective, accurate and efficient monitoring of mine blast pile morphology is achieved, providing reliable data support for the abnormal identification of blasting monitoring images.

[0024] S400: When the monitored muck pile shape is inconsistent with the reference muck pile shape, an uncontrollable identifier is sent to the user terminal for the monitored mine muck pile image.

[0025] Specifically, for the phased characteristics of the mine blasting process, corresponding reference muck pile shapes are established for each stage of the blasting operation. After the blasting operation of a certain stage is completed, the monitored muck pile shape formed after blasting is obtained in real time through the sensor array or image acquisition device deployed on site, and is compared and analyzed with the reference muck pile shape of the corresponding stage. Among them, the monitored muck pile shape includes geometric characteristics and physical state parameters of the rock mass accumulation after blasting, such as indicators like muck pile height, muck pile width, muck pile length, muck pile slope angle, looseness coefficient, large block rate, etc.

[0026] When it is detected that the monitored muck pile shape of a certain stage is inconsistent with the reference muck pile shape, it is determined that there is an abnormal situation in the blasting of the current stage, and there may be interference from uncontrollable factors, such as abnormal rock mass structure, improper blasting parameter setting, uneven distribution of blasting charge, etc. After identifying the blasting anomaly, an uncontrollable identifier is immediately generated for the obtained monitored mine muck pile image, and this identifier is sent to the user terminal in real time through the network communication module. Among them, the uncontrollable identifier can include information such as anomaly type, anomaly degree, anomaly location, etc., providing timely and accurate blasting anomaly warnings for mine management personnel.

[0027] By performing uncontrollable identification on the monitored mine muck pile image that does not conform to the reference muck pile shape and sending it to the user terminal, mine workers can effectively identify anomalies in the mine blasting process, thereby effectively avoiding disasters caused by blasting anomalies and handling potential risks in the blasting process, improving the safety and controllability of mine blasting operations, and achieving precise identification of anomalies in the monitored mine blasting images.

[0028] Furthermore, with the rock mass type distribution topology, rock mass fracture distribution topology, and rock mass hardness distribution topology as constraints, blasting sample collection is performed on the blasting position list and the blasting stage list to obtain a blasting sample set, including: S210: With the rock mass type distribution topology, the rock mass fracture distribution topology, and the rock mass hardness distribution topology as constraints, blasting sample collection is performed on the blasting position list and the blasting stage list to obtain a first-level blasting sample set, where the first-level blasting sample set has a first-level blasting position label list and a first-level blasting stage label list; S220: With the rock mass type distribution topology, the rock mass fracture distribution topology, and the rock mass hardness distribution topology as constraints, blasting sample collection is performed on the first-level blasting position label list and the first-level blasting stage label list to obtain a second-level blasting sample set; S230: Until the blasting sample collection is carried out on the N-1 level blasting position label list and the N-1 level blasting stage label list under the constraints of the rock mass type distribution topology, the rock mass fracture distribution topology, and the rock mass hardness distribution topology, to obtain the N level blasting sample set, where N is an integer and 5≥N≥2; S240: Add the first-level blasting sample set, the second-level blasting sample set until the N level blasting sample set into the blasting sample set.

[0029] In an optional implementation manner, a statistically significant blasting sample set is constructed through a progressive acquisition strategy, providing sufficient data support for the subsequent construction of the benchmark morphology of the muck pile.

[0030] First, taking the rock mass type distribution topology, the rock mass fracture distribution topology, and the rock mass hardness distribution topology as constraint conditions, the blasting sample collection is carried out on the predetermined blasting position list and the blasting stage list to obtain the first-level blasting sample set. Among them, the first-level blasting sample set has multiple first-level blasting samples, and each first-level blasting sample has a corresponding first-level blasting position label list and a first-level blasting stage label list; the first-level blasting position label list records the blasting position information in the first-level blasting sample, and the first-level blasting stage label list records the corresponding blasting intensity level information.

[0031] Then, using the progressive acquisition method, the first-level blasting position label list and the first-level blasting stage label list generated by the first-level blasting sample set are used as new parameters, and the blasting sample collection is continued under the constraints of the rock mass type distribution topology, the rock mass fracture distribution topology, and the rock mass hardness distribution topology to obtain the second-level blasting sample set. Through the strategy of finding samples with samples, the quantity and diversity of the blasting samples can be effectively expanded, making the blasting sample set more statistically representative, thereby improving the accuracy and reliability of the subsequent muck pile morphology analysis. The progressive acquisition process can flexibly adjust the acquisition depth according to actual needs, support the continuous execution of multi-level acquisition, that is, the blasting sample collection can be carried out on the N-1 level blasting position label list and the N-1 level blasting stage label list to obtain the N level blasting sample set. Where N is an integer and satisfies the condition limit of 5≥N≥2. When N = 2, it means that only one round of expansion is carried out on the basis of the first-level blasting sample set to form a second-level sample structure; when the sample quantity requirement is large, deeper acquisition can be selected, such as the five-level sample expansion of N = 5, so as to ensure that the sample quantity meets the significance requirements of statistical analysis. After that, the first-level blasting sample set, the second-level blasting sample set until the N level blasting sample set are all added into the overall blasting sample set.

[0032] Through multi-level sample collection and integration, the problem of insignificant statistics caused by insufficient sample size in traditional blasting monitoring is effectively solved, providing reliable data guarantee for high-frequency muck pile shape mining, thereby improving the accuracy and reliability of mine blasting monitoring.

[0033] Furthermore, constrained by the rock mass type distribution topology, the rock mass fracture distribution topology, and the rock mass hardness distribution topology, blasting sample collection is carried out on the blasting position list and the blasting stage list to obtain a first-level blasting sample set, including: S211: Extract the sample rock mass type distribution topology, the sample rock mass fracture distribution topology, the sample rock mass hardness distribution topology, the sample blasting position list, and the sample blasting stage list of the blasting sample to be analyzed; S212: Statistically calculate the ratio of the number of rock mass types with the same position in the sample rock mass type distribution topology and the rock mass type distribution topology to the number of the union of rock mass types, and set it as the rock mass type topology similarity; S213: Statistically calculate the fracture structure similarity between the sample rock mass fracture distribution topology and the rock mass fracture distribution topology, and set it as the rock mass fracture topology similarity; S214: Statistically calculate the ratio of the number of regions with the same hardness in the sample rock mass hardness distribution topology and the rock mass hardness distribution topology to the total number of hardness partitions, and set it as the rock mass hardness topology similarity; S215: Based on the blasting position list and the blasting stage list, combined with the sample blasting position list and the sample blasting stage list, statistically calculate the ratio of the number of positions with the same blasting stage to the number of the union of blasting positions, and set it as the blasting parameter similarity; S216: When the rock mass type topology similarity is greater than or equal to the first similarity threshold, and the rock mass fracture topology similarity is greater than or equal to the second similarity threshold, and the rock mass hardness topology similarity is greater than or equal to the third similarity threshold, and the blasting parameter similarity is greater than or equal to the fourth similarity threshold, add the blasting sample to be analyzed to the first-level blasting sample set.

[0034] In a preferred implementation manner, through multi-dimensional similarity calculation and threshold judgment of the blasting sample to be analyzed and the reference data, the screening of blasting samples based on multi-dimensional similarity is realized, providing a basis for constructing the first-level blasting sample set and ensuring the accuracy and representativeness of blasting sample collection.

[0035] When collecting blasting samples, each blasting sample in the blasting sample database should be regarded as a blasting sample to be analyzed, and analyzed in conjunction with the current rock mass type distribution topology, rock mass fracture distribution topology, rock mass hardness distribution topology, blasting position list, and blasting level list to determine whether the blasting sample to be analyzed can be included in the first-level blasting sample set. First, extract the key feature data of the blasting sample to be analyzed, including the sample rock mass type distribution topology, sample rock mass fracture distribution topology, sample rock mass hardness distribution topology, sample blasting position list, and sample blasting level list. These feature data comprehensively reflect the rock mass conditions and blasting parameter information of the blasting sample to be analyzed, providing a data basis for subsequent similarity calculations.

[0036] Then, calculate the similarity between the sample rock mass type distribution topology and the rock mass type distribution topology. Specifically, use the ratio of the number of positions with the same rock mass type to the number of the union of rock mass types as the evaluation index, denoted as the rock mass type topology similarity, so as to reflect the matching degree of the two topological structures in the spatial distribution of rock mass types. At the same time, calculate the similarity between the sample rock mass fracture distribution topology and the rock mass fracture distribution topology, denoted as the rock mass fracture topology similarity. For example, when statistically analyzing the fracture structure similarity between the sample rock mass fracture distribution topology and the rock mass fracture distribution topology, consider multi-dimensional features such as fracture density distribution similarity, fracture strike consistency, and fracture connectivity characteristic matching degree, and obtain a comprehensive fracture structure similarity value through comprehensive evaluation as the rock mass fracture topology similarity. Meanwhile, calculate the similarity between the sample rock mass hardness distribution topology and the rock mass hardness distribution topology. Specifically, use the ratio of the number of positions with the same hardness in the same region to the total number of hardness zones as the evaluation index, denoted as the rock mass hardness topology similarity, so as to effectively reflect the consistency degree of the two topological structures in the rock mass hardness distribution characteristics. In addition, according to the blasting position list, blasting level list, sample blasting position list, and sample blasting level list, calculate the similarity between the blasting sample to be analyzed and the current blasting parameters. Specifically, according to the blasting position list and blasting level list, as well as the sample blasting position list and sample blasting level list, calculate the ratio of the number of positions with the same blasting level in the same position to the number of the union of blasting positions to obtain the blasting parameter similarity.

[0037] Afterwards, a multi-threshold judgment mechanism is set. When the topological similarity of the rock mass type is greater than or equal to the first similarity threshold, the topological similarity of the rock mass fissures is greater than or equal to the second similarity threshold, the topological similarity of the rock mass hardness is greater than or equal to the third similarity threshold, and the similarity of the blasting parameters is greater than or equal to the fourth similarity threshold, it is determined that the blasting sample to be analyzed has sufficient similarity and is added to the first-level blasting sample set. Among them, the first similarity threshold, the second similarity threshold, the third similarity threshold, and the fourth similarity threshold can be flexibly set by the mine blasting expert group according to the actual situation of the mine, taking into account both the strict requirements for blasting sample screening and the data fluctuation characteristics in the actual mine environment to ensure the high-quality construction of the first-level blasting sample set.

[0038] Through the screening of blasting samples based on multi-dimensional similarity, the accurate screening of blasting samples is realized, ensuring that the samples in the first-level blasting sample set have sufficient representativeness and laying a solid data foundation for the subsequent identification of abnormal blasting monitoring data.

[0039] Furthermore, a blasting sample set is obtained, and high-frequency muck pile shape mining is performed to obtain the reference muck pile shape, including: S250: Extract the first-level blasting sample set, the second-level blasting sample set until the N-level blasting sample set from the blasting sample set; S260: Perform high-frequency muck pile shape mining on the N-level blasting sample set to obtain the N-level high-frequency muck pile shape; S270: Perform high-frequency muck pile shape mining on the N-level high-frequency muck pile shape and the N-1 level blasting sample set to obtain the N-1 level high-frequency muck pile shape; S280: Until high-frequency muck pile shape mining is performed on the second-level high-frequency muck pile shape and the first-level blasting sample set to obtain the reference muck pile shape.

[0040] In a preferred embodiment, by performing high-frequency muck pile shape mining according to multi-level samples, recursively analyzing the characteristic distributions of each level of blasting sample sets, and gradually constructing a muck pile shape model, representative ones are extracted from the multi-level blasting sample sets.

[0041] First, separate the blast sample subsets at all levels from the overall blast sample set, and extract the first-level blast sample set, the second-level blast sample set until the N-level blast sample set, laying a data foundation for subsequent hierarchical analysis. Then, perform high-frequency muckpile morphology mining on the N-level blast sample set. The so-called high-frequency muckpile morphology mining refers to extracting the parameter combination with the highest occurrence frequency by statistically analyzing the distribution characteristics of each muckpile morphology parameter in the blast sample set to form a muckpile morphology description with statistical significance. Through the analysis of the N-level blast sample set, the high-frequency morphology of the N-level muckpile is obtained, which represents the most representative muckpile characteristics in the N-level blast sample set. Subsequently, using the high-frequency morphology of the N-level muckpile as prior knowledge, combine it with the (N - 1)-level blast sample set for high-frequency muckpile morphology mining to obtain the high-frequency morphology of the (N - 1)-level muckpile. This way of gradually fusing from the extended samples to the basic samples can effectively utilize the diversity and statistical significance of the extended samples, improving the accuracy and reliability of muckpile morphology feature extraction.

[0042] Continue to execute the above recursive analysis process until finally perform high-frequency muckpile morphology mining on the high-frequency morphology of the second-level muckpile and the first-level blast sample set to obtain the benchmark muckpile morphology. The obtained benchmark muckpile morphology synthesizes the characteristic distribution laws of blast samples at all levels, reflecting both the common characteristics of a large number of samples and taking into account the particularity of samples at different levels, making the obtained benchmark muckpile morphology highly representative and applicable.

[0043] Through the high-frequency muckpile morphology mining method of gradual fusion, an accurate benchmark muckpile morphology is obtained, providing a reliable reference standard for subsequent blast anomaly monitoring, thereby improving the accuracy and reliability of mine blast monitoring.

[0044] Furthermore, perform high-frequency muckpile morphology mining on the N-level blast sample set to obtain the high-frequency morphology of the N-level muckpile, including: S261: Group the N-level blast sample set according to the (N - 1)-level blast sample set to obtain multiple groups of N-level blast samples; S262: For the first group of N-level blast samples in the multiple groups of N-level blast samples, extract the muckpile morphology parameters and perform mode statistics on the same-attribute parameters to obtain the high-frequency morphology of the first group of muckpiles; S263: Until for the Mth group of N-level blast samples in the multiple groups of N-level blast samples, extract the muckpile morphology parameters and perform mode statistics on the same-attribute parameters to obtain the high-frequency morphology of the Mth group of muckpiles; S264: Add the high-frequency morphology of the first group of muckpiles until the high-frequency morphology of the Mth group of muckpiles to the high-frequency morphology of the N-level muckpile.

[0045] In a preferred embodiment, by grouping and parameter statistics on the N-level blasting sample set, the accurate extraction of the muck pile morphological characteristics is realized, providing a basis for the construction of the benchmark muck pile morphology.

[0046] First, according to the characteristic distribution of the (N - 1)-level blasting sample set, the N-level blasting sample set is grouped to obtain multiple groups of N-level blasting samples. This grouping strategy based on the characteristics of the previous-level samples can classify blasting samples with similar attributes into one category, improving the accuracy and pertinence of subsequent morphology mining.

[0047] Then, the first group of multiple groups of N-level blasting samples is processed. The muck pile morphological parameters are extracted and the mode statistics of the same-attribute parameters are carried out to obtain the high-frequency muck pile morphology of the first group. Among them, the muck pile morphological parameters include, but are not limited to, multi-dimensional characteristic quantities such as muck pile height, muck pile width, muck pile length, muck pile slope angle, looseness coefficient, settlement rate, large block rate, block size uniformity, forward rush distance, symmetry of the accumulation morphology, toe and overhanging rock. The mode statistics of the same-attribute parameters means that for each type of parameter, the value with the highest occurrence frequency is respectively counted to form the most representative parameter combination of this group of samples. The above parameter statistics process is sequentially applied to multiple groups of N-level blasting samples, from the first group to the Mth group, and the high-frequency muck pile morphology of each group is obtained respectively. Through the grouping process, the differences in the muck pile morphology under different conditions can be fully considered, avoiding the problem of blurred features caused by mixing all samples for statistics.

[0048] After that, the high-frequency muck pile morphology of the first group to the high-frequency muck pile morphology of the Mth group are all added to the N-level high-frequency muck pile morphology to form a complete description system including high-frequency morphologies under various conditions. Through the integration of multi-group high-frequency morphologies, both the characteristic differences of each group of samples are retained, and a unified morphology system is constructed, providing a multi-dimensional reference standard for subsequent morphology matching and anomaly recognition.

[0049] Through the mining of the high-frequency muck pile morphology based on grouping statistics, the representative muck pile morphological characteristics under various conditions can be accurately extracted from the N-level blasting sample set, effectively improving the accuracy and applicability of the benchmark muck pile morphology, and providing reliable technical support for the anomaly recognition of mine blasting monitoring data.

[0050] Furthermore, for the N-level high-frequency muck pile morphology and the (N - 1)-level blasting sample set, the high-frequency muck pile morphology mining is performed to obtain the (N - 1)-level high-frequency muck pile morphology, including: S271: Based on the high-frequency muck pile morphology of the first group, the muck pile morphological parameters of the first associated blasting sample of the (N - 1)-level blasting sample set are expanded until the muck pile morphological parameters of the Mth associated blasting sample of the (N - 1)-level blasting sample set are expanded based on the high-frequency muck pile morphology of the Mth group, obtaining the updated result of the muck pile morphological parameters of the (N - 1)-level blasting sample; S272: Group the N-1 level blasting sample set according to the N-2 level blasting sample set to obtain multiple groups of N-1 level blasting samples; S273: Based on the multiple groups of N-1 level blasting samples, perform high-frequency muck pile morphology mining on the updated results of the muck pile morphology parameters of the N-1 level blasting samples to obtain the high-frequency morphology of the N-1 level muck pile.

[0051] In a preferred embodiment, through a combination of parameter expansion and high-frequency morphology mining, an effective fusion of the high-frequency morphology of the N-level muck pile and the N-1 level blasting sample set is carried out to achieve the transfer and fusion of morphological features between blasting samples of different levels, providing support for constructing the benchmark morphology of the muck pile.

[0052] First, expand the muck pile morphology parameters of the N-1 level blasting sample set based on the high-frequency morphology of the N-level muck pile. The so-called expansion of muck pile morphology parameters means adding the characteristic parameters extracted from the high-frequency morphology of the N-level muck pile as supplementary information to the morphological description of the N-1 level blasting samples to enrich their characteristics. Specifically, add the first group of high-frequency muck pile morphology in the high-frequency morphology of the N-level muck pile to the blasting sample (i.e., the first associated blasting sample) in the N-1 level blasting sample set that matches its characteristics, and so on, until the Mth group of high-frequency muck pile morphology parameters is added to the corresponding Mth associated blasting sample, finally obtaining the updated results of the muck pile morphology parameters of the N-1 level blasting samples.

[0053] Then, group the N-1 level blasting sample set according to the characteristic distribution of the N-2 level blasting sample set to obtain multiple groups of N-1 level blasting samples. This grouping strategy based on the characteristics of the upper-level samples can classify blasting samples with similar attributes into one category, improving the accuracy and pertinence of subsequent morphology mining. Subsequently, based on multiple groups of N-1 level blasting samples, perform high-frequency muck pile morphology mining on the updated results of the muck pile morphology parameters of the N-1 level blasting samples after parameter expansion to obtain the high-frequency morphology of the N-1 level muck pile. Through the method of grouped mining, on the basis of retaining the original characteristics of the N-1 level blasting samples, the characteristic information of the high-frequency morphology of the N-level muck pile can be fused to form a more comprehensive and accurate description of the high-frequency morphology of the N-1 level muck pile.

[0054] By combining parameter expansion and high-frequency morphology mining, the transfer and fusion of morphological features between blasting samples of different levels are effectively realized, and the effective transfer of morphological features at different levels is achieved, providing support for constructing a complete benchmark morphology of the muck pile, thereby improving the accuracy and reliability of abnormal recognition of mine blasting monitoring data.

[0055] Furthermore, obtain the topological distribution of the rock mass humidity and the topological distribution of the rock mass temperature of the mine, and perform hardness zoning in combination with the topological distribution of the rock mass type to obtain the topological distribution of the rock mass hardness, including: S110: Configure the rock mass humidity record data, rock mass temperature record data, and rock mass type record data; S120: Collect the central value of the rock mass hardness record data set that meets the above-mentioned rock mass humidity record data, rock mass temperature record data, and rock mass type record data, and set it as the true value of the rock mass hardness; S130: Using the true value of the rock mass hardness as the supervision, and using the rock mass humidity record data, rock mass temperature record data, and rock mass type record data to train a random forest to obtain a rock mass hardness predictor; S140: Process the rock mass humidity distribution topology, rock mass temperature distribution topology, and rock mass type distribution topology through the rock mass hardness predictor to obtain the rock mass hardness distribution information; S150: Based on the rock mass hardness threshold, partition the rock mass hardness distribution information to obtain the rock mass hardness distribution topology.

[0056] In a preferred embodiment, by collecting the multi-dimensional characteristic data of the mine rock mass and using the random forest technology to construct a prediction relationship, an accurate mapping from the basic characteristics of the rock mass to the hardness distribution is realized.

[0057] First, configure the rock mass humidity record data, rock mass temperature record data, and rock mass type record data. Among them, the rock mass humidity record data refers to the rock mass water content data of each point in the mine obtained by drilling sampling or on-site detection; the rock mass temperature record data refers to the actual temperature measurement values of the rock mass at each point in the mine; the rock mass type record data refers to the type classification information of the rocks at each point in the mine, such as granite, limestone, shale, etc. Then, collect the rock mass hardness record data set that meets the combination of specific rock mass humidity record data, rock mass temperature record data, and rock mass type record data, and extract its central value as the true value of the rock mass hardness. Specifically, for each specific combination of rock mass humidity, temperature, and type, collect multiple hardness measurement values, and extract its central tendency value (such as the mean or median) through statistical methods to eliminate the influence of measurement errors. This central value is the true value of the rock mass hardness. Then, using the true value of the rock mass hardness as the supervision and the rock mass humidity record data, rock mass temperature record data, and rock mass type record data as features, train a random forest to obtain a rock mass hardness predictor. Among them, the random forest is an ensemble learning method. By constructing multiple decision trees and taking their average prediction results, it can effectively handle the non-linear relationship in the rock mass hardness prediction. During the training process, the cross-validation method is used to optimize the hyperparameters of the random forest, such as the number of trees, the maximum depth, etc., to improve the prediction accuracy.

[0058] Subsequently, the trained rock mass hardness predictor is applied to the topology of the rock mass humidity distribution, temperature distribution, and rock mass type distribution in the entire mining area to predict continuous rock mass hardness distribution information. Among them, the rock mass hardness distribution information refers to the predicted hardness values at each spatial position in the mine, forming a three-dimensional hardness distribution field. Then, based on the rock mass hardness threshold, the rock mass hardness distribution information is partitioned to obtain the rock mass hardness distribution topology. The rock mass hardness threshold is a set of hardness demarcation values preset according to engineering experience and blasting requirements. Through these thresholds, the continuous hardness distribution information is divided into several discrete hardness intervals, such as extremely soft area, soft area, medium-hard area, hard area, and extremely hard area, etc., thus forming a clear rock mass hardness partition structure, which is convenient for the targeted setting of subsequent blasting parameters and abnormal monitoring.

[0059] Furthermore, through an image acquisition device, mine muckpile monitoring images are collected, and image feature extraction is performed to obtain the muckpile monitoring morphology, including: S310: Collect multiple groups of data, where any one of the multiple groups of data includes muckpile image data and labels identifying muckpile morphology parameters; S320: Based on the multiple groups of data, a muckpile morphology parameter extraction model is trained through a convolutional neural network; S330: Process the mine muckpile monitoring images according to the muckpile morphology parameter extraction model to obtain the muckpile monitoring morphology.

[0060] In a preferred implementation manner, a deep learning-based method is adopted to realize the automatic mapping from the muckpile monitoring image to the morphology parameters by constructing a muckpile morphology parameter extraction model.

[0061] First, multiple groups of data are collected as the basic data set for model training. Each group of data contains two core parts: muckpile image data and the corresponding labels identifying muckpile morphology parameters. Among them, the muckpile image data refers to the high-resolution images of the rock mass accumulation after blasting obtained through an image acquisition device; the labels identifying muckpile morphology parameters refer to the key parameters of the muckpile morphology marked by blasting experts or an automatic measurement system, such as quantitative indicators such as muckpile height, width, length, slope angle, looseness coefficient, and large block rate. By constructing a large-scale labeled data set containing various blasting conditions and various rock mass environments, that is, multiple groups of data, a rich and diverse sample basis is provided for the subsequent deep learning model training.

[0062] Then, based on the collected multiple sets of data, a model for extracting muckpile shape parameters is constructed and trained. This model adopts the structure of a convolutional neural network (CNN). Such networks are good at processing image data with spatial correlation and can automatically learn and extract hierarchical features in muckpile images. The input end of the model receives muckpile image data, and the output end generates predicted values of muckpile shape parameters. During the training process, the network parameters are continuously optimized through the backpropagation algorithm to minimize the error between the predicted values of the shape parameters output by the model and the labels. To improve the generalization ability and robustness of the model, data augmentation techniques such as rotation, scaling, and brightness adjustment are adopted during the training process, and regularization methods are applied to prevent overfitting.

[0063] After that, the trained model for extracting muckpile shape parameters is applied to the actually collected mine muckpile monitoring images. The model can quickly and accurately extract key shape parameters from the newly collected mine muckpile monitoring images and form a complete description of the muckpile monitoring shape. Specifically, the mine muckpile monitoring images are input into the model for extracting muckpile shape parameters. After multi-level feature extraction and transformation through convolutional layers, pooling layers, fully connected layers, etc., the finally output muckpile monitoring shape includes a comprehensive description of the geometric features and physical state parameters of the muckpile.

[0064] The method for extracting muckpile shape parameters based on deep learning overcomes the limitations of traditional manual extraction or rule-based algorithms in complex environments and improves the accuracy and efficiency of obtaining muckpile monitoring shapes. This method can adapt to the analysis of muckpile images under different lighting conditions and different perspectives, has strong environmental adaptability and parameter extraction stability, and lays a data foundation for subsequent blasting anomaly monitoring.

[0065] Embodiment 2, as Figure 2 shown, based on the same inventive concept as the method for abnormally identifying mine blasting monitoring images provided in Embodiment 1, the present invention embodiment also provides a system for abnormally identifying mine blasting monitoring images, including: A rock mass property analysis module 11, configured to obtain the topological distribution of rock mass humidity and the topological distribution of rock mass temperature in a mine, combine the topological distribution of rock mass types for hardness zoning, and obtain the topological distribution of rock mass hardness; A reference shape acquisition module 12, configured to collect blasting samples from a blasting position list and a blasting stage list with the topological distribution of rock mass types, the topological distribution of rock mass fractures, and the topological distribution of rock mass hardness as constraints, obtain a blasting sample set, and perform high-frequency muckpile shape mining to obtain a muckpile reference shape; A monitoring shape acquisition module 13, configured to collect mine muckpile monitoring images through an image acquisition device, perform image feature extraction, and obtain a muckpile monitoring shape; The image identification module 14 is used to send an uncontrollable identification of the mine bench monitoring image to the user terminal when the bench monitoring form is inconsistent with the bench reference form.

[0066] Further, the reference form acquisition module 12 includes the following execution steps: Taking the rock mass type distribution topology, the rock mass fracture distribution topology, and the rock mass hardness distribution topology as constraints, blasting sample collection is performed on the blasting position list and the blasting stage list to obtain a first-level blasting sample set, where the first-level blasting sample set has a first-level blasting position tag list and a first-level blasting stage tag list; Taking the rock mass type distribution topology, the rock mass fracture distribution topology, and the rock mass hardness distribution topology as constraints, blasting sample collection is performed on the first-level blasting position tag list and the first-level blasting stage tag list to obtain a second-level blasting sample set; Until, taking the rock mass type distribution topology, the rock mass fracture distribution topology, and the rock mass hardness distribution topology as constraints, blasting sample collection is performed on the (N - 1)-level blasting position tag list and the (N - 1)-level blasting stage tag list to obtain an N-level blasting sample set, where N is an integer and 5 ≥ N ≥ 2; Adding the first-level blasting sample set, the second-level blasting sample set until the N-level blasting sample set into the blasting sample set.

[0067] Further, the reference form acquisition module 12 further includes the following execution steps: Extracting the sample rock mass type distribution topology, the sample rock mass fracture distribution topology, the sample rock mass hardness distribution topology, the sample blasting position list, and the sample blasting stage list of the blasting sample to be analyzed; Statistically calculating the ratio of the number of the same-position rock mass types that are consistent between the sample rock mass type distribution topology and the rock mass type distribution topology to the number of the union of rock mass types, and setting it as the rock mass type topology similarity; Statistically calculating the fracture structure similarity between the sample rock mass fracture distribution topology and the rock mass fracture distribution topology, and setting it as the rock mass fracture topology similarity; Statistically calculating the ratio of the number of the same-region hardnesses that are consistent between the sample rock mass hardness distribution topology and the rock mass hardness distribution topology to the total number of hardness partitions, and setting it as the rock mass hardness topology similarity; Based on the blasting position list and the blasting stage list, combining the sample blasting position list and the sample blasting stage list, statistically calculating the ratio of the number of positions where the blasting stages are consistent at the same position to the number of the union of blasting positions, and setting it as the blasting parameter similarity; When the topological similarity of the rock mass type is greater than or equal to the first similarity threshold, and the topological similarity of the rock mass fissure is greater than or equal to the second similarity threshold, and the topological similarity of the rock mass hardness is greater than or equal to the third similarity threshold, and the similarity of the blasting parameters is greater than or equal to the fourth similarity threshold, add the blasting sample to be analyzed into the first-level blasting sample set.

[0068] Further, the reference form acquisition module 12 further includes the following execution steps: Extract the first-level blasting sample set, the second-level blasting sample set, and up to the N-level blasting sample set from the blasting sample set; Perform high-frequency muck pile form mining on the N-level blasting sample set to obtain the N-level muck pile high-frequency form; Perform high-frequency muck pile form mining on the N-level muck pile high-frequency form and the N-1-level blasting sample set to obtain the N-1-level muck pile high-frequency form; Until high-frequency muck pile form mining is performed on the second-level muck pile high-frequency form and the first-level blasting sample set to obtain the reference muck pile form.

[0069] Further, the reference form acquisition module 12 further includes the following execution steps: Group the N-level blasting sample set according to the N-1-level blasting sample set to obtain multiple groups of N-level blasting samples; Extract the muck pile form parameters of the first group of N-level blasting samples in the multiple groups of N-level blasting samples, and perform the mode statistics of the same-attribute parameters to obtain the first group of muck pile high-frequency forms; Until the muck pile form parameters of the Mth group of N-level blasting samples in the multiple groups of N-level blasting samples are extracted, and the mode statistics of the same-attribute parameters are performed to obtain the Mth group of muck pile high-frequency forms; Add the first group of muck pile high-frequency forms to the Mth group of muck pile high-frequency forms into the N-level muck pile high-frequency form.

[0070] Further, the reference form acquisition module 12 further includes the following execution steps: Based on the first group of muck pile high-frequency forms, expand the muck pile form parameters of the first associated blasting sample in the N-1-level blasting sample set until the muck pile form parameters of the Mth associated blasting sample in the N-1-level blasting sample set are expanded based on the Mth group of muck pile high-frequency forms to obtain the updated result of the muck pile form parameters of the N-1-level blasting samples; Group the N-1-level blasting sample set according to the N-2-level blasting sample set to obtain multiple groups of N-1-level blasting samples; Based on the multiple groups of N-1 level blasting samples, perform high-frequency muck pile morphology mining on the updated results of the muck pile morphology parameters of the N-1 level blasting samples to obtain the high-frequency morphology of the N-1 level muck pile.

[0071] Further, the rock mass property analysis module 11 includes the following implementation steps: Configure the rock mass humidity record data, rock mass temperature record data, and rock mass type record data; Collect the central value of the rock mass hardness record data set that meets the rock mass humidity record data, the rock mass temperature record data, and the rock mass type record data, and set it as the true value of the rock mass hardness; Using the true value of the rock mass hardness as the supervision, and using the rock mass humidity record data, the rock mass temperature record data, and the rock mass type record data, train a random forest to obtain a rock mass hardness predictor; Through the rock mass hardness predictor, process the rock mass humidity distribution topology, the rock mass temperature distribution topology, and the rock mass type distribution topology to obtain the rock mass hardness distribution information; Based on the rock mass hardness threshold, partition the rock mass hardness distribution information to obtain the rock mass hardness distribution topology.

[0072] Further, the monitoring morphology acquisition module includes the following implementation steps: Collect multiple groups of data, where any one of the multiple groups of data includes muck pile image data and a label identifying the muck pile morphology parameters; According to the multiple groups of data, train a muck pile morphology parameter extraction model through a convolutional neural network; Process the mine muck pile monitoring image according to the muck pile morphology parameter extraction model to obtain the muck pile monitoring morphology.

[0073] It should be noted that in the above embodiments, the descriptions of each embodiment have their own emphases. For the parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0074] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0075] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general purpose computers, special purpose computers, embedded computers, or other programmable data processing devices to produce a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices produce means for implementing the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 one or more blocks

[0076] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including instruction means that implement the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 one or more blocks

[0077] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in the Figure 1 one or more flows and / or blocks Figure 1 one or more blocks

[0078] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made to these embodiments by those skilled in the art once they learn of the basic inventive concept.

[0079] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalent technologies, the present invention is also intended to include these changes and modifications.

Claims

1. A method for abnormally identifying mine blasting monitoring images, characterized in that, Including: Obtain the topological distribution of the rock mass humidity and the topological distribution of the rock mass temperature in the mine, combine the topological distribution of the rock mass type for hardness zoning, and obtain the topological distribution of the rock mass hardness; Taking the topological distribution of the rock mass type, the topological distribution of the rock mass fissure, and the topological distribution of the rock mass hardness as constraints, conduct blasting sample collection on the blasting position list and the blasting level list to obtain a blasting sample set, and perform high-frequency muck pile morphology mining to obtain the benchmark muck pile morphology; Through an image acquisition device, collect the mine muck pile monitoring image, and perform image feature extraction to obtain the muck pile monitoring morphology; When the muck pile monitoring morphology is inconsistent with the benchmark muck pile morphology, send an uncontrollable identifier for the mine muck pile monitoring image to the user terminal.

2. The method according to claim 1, characterized in that, Taking the topological distribution of the rock mass type, the topological distribution of the rock mass fissure, and the topological distribution of the rock mass hardness as constraints, conduct blasting sample collection on the blasting position list and the blasting level list to obtain a blasting sample set, including: Taking the topological distribution of the rock mass type, the topological distribution of the rock mass fissure, and the topological distribution of the rock mass hardness as constraints, conduct blasting sample collection on the blasting position list and the blasting level list to obtain a primary blasting sample set, where the primary blasting sample set has a primary blasting position tag list and a primary blasting level tag list; Taking the topological distribution of the rock mass type, the topological distribution of the rock mass fissure, and the topological distribution of the rock mass hardness as constraints, conduct blasting sample collection on the primary blasting position tag list and the primary blasting level tag list to obtain a secondary blasting sample set; Until, taking the topological distribution of the rock mass type, the topological distribution of the rock mass fissure, and the topological distribution of the rock mass hardness as constraints, conduct blasting sample collection on the (N - 1)-level blasting position tag list and the (N - 1)-level blasting level tag list to obtain an N-level blasting sample set, where N is an integer and 5 ≥ N ≥ 2; Add the primary blasting sample set, the secondary blasting sample set until the N-level blasting sample set into the blasting sample set.

3. The method according to claim 2, characterized in that, Taking the topological distribution of the rock mass type, the topological distribution of the rock mass fissure, and the topological distribution of the rock mass hardness as constraints, conduct blasting sample collection on the blasting position list and the blasting level list to obtain a primary blasting sample set, including: Extract the sample rock mass type topological distribution, sample rock mass fissure topological distribution, sample rock mass hardness topological distribution, sample blasting position list, and sample blasting level list of the blasting sample to be analyzed; Statistically calculate the ratio of the number of consistent rock mass types at the same position between the sample rock mass type topological distribution and the rock mass type topological distribution to the number of the union of rock mass types, and set it as the rock mass type topological similarity; Statistically calculate the fissure structure similarity between the sample rock mass fissure topological distribution and the rock mass fissure topological distribution, and set it as the rock mass fissure topological similarity; Statistically calculate the ratio of the number of consistent hardness in the same area between the sample rock mass hardness topological distribution and the rock mass hardness topological distribution to the total number of hardness zones, and set it as the rock mass hardness topological similarity; Based on the list of blasting positions and the list of blasting levels, combined with the sample blasting position list and the sample blasting level list, calculate the ratio of the number of positions with the same blasting level to the number of the union of blasting positions, and set it as the blasting parameter similarity; When the topological similarity of the rock mass type is greater than or equal to the first similarity threshold, the topological similarity of the rock mass fissures is greater than or equal to the second similarity threshold, the topological similarity of the rock mass hardness is greater than or equal to the third similarity threshold, and the blasting parameter similarity is greater than or equal to the fourth similarity threshold, add the blasting sample to be analyzed into the first-level blasting sample set.

4. The method according to claim 2, wherein Obtain the blasting sample set, perform high-frequency muck pile morphology mining, and obtain the benchmark muck pile morphology, including: Extract the first-level blasting sample set, the second-level blasting sample set until the N-level blasting sample set from the blasting sample set; Perform high-frequency muck pile morphology mining on the N-level blasting sample set to obtain the N-level muck pile high-frequency morphology; Perform high-frequency muck pile morphology mining on the N-level muck pile high-frequency morphology and the N-1 level blasting sample set to obtain the N-1 level muck pile high-frequency morphology; Until high-frequency muck pile morphology mining is performed on the second-level muck pile high-frequency morphology and the first-level blasting sample set to obtain the benchmark muck pile morphology.

5. The method according to claim 4, characterized in that Perform high-frequency muck pile morphology mining on the N-level blasting sample set to obtain the N-level muck pile high-frequency morphology, including: Group the N-level blasting sample set according to the N-1 level blasting sample set to obtain multiple groups of N-level blasting samples; Extract the muck pile morphology parameters of the first group of N-level blasting samples in the multiple groups of N-level blasting samples, perform the mode statistics of the same-attribute parameters, and obtain the first group of muck pile high-frequency morphology; Until the muck pile morphology parameters of the Mth group of N-level blasting samples in the multiple groups of N-level blasting samples are extracted, perform the mode statistics of the same-attribute parameters, and obtain the Mth group of muck pile high-frequency morphology; Add the first group of muck pile high-frequency morphology until the Mth group of muck pile high-frequency morphology into the N-level muck pile high-frequency morphology.

6. The method according to claim 5, wherein Perform high-frequency muck pile morphology mining on the N-level muck pile high-frequency morphology and the N-1 level blasting sample set to obtain the N-1 level muck pile high-frequency morphology, including: Based on the first group of muck pile high-frequency morphology, expand the muck pile morphology parameters of the first associated blasting sample in the N-1 level blasting sample set until the muck pile morphology parameters of the Mth associated blasting sample in the N-1 level blasting sample set are expanded based on the Mth group of muck pile high-frequency morphology, and obtain the updated result of the muck pile morphology parameters of the N-1 level blasting samples; Group the N-1 level blasting sample set according to the N-2 level blasting sample set to obtain multiple groups of N-1 level blasting samples; Based on the multiple groups of N-1 level blasting samples, perform high-frequency muck pile morphology mining on the updated result of the muck pile morphology parameters of the N-1 level blasting samples to obtain the N-1 level muck pile high-frequency morphology.

7. The method according to claim 1, characterized in that, Obtain the topological distribution of the rock mass humidity and the topological distribution of the rock mass temperature of the mine, and combine the topological distribution of the rock mass type to perform hardness zoning to obtain the topological distribution of the rock mass hardness, including: Configure the rock mass humidity record data, the rock mass temperature record data, and the rock mass type record data; Collect the central value of the rock mass hardness record data set that satisfies the rock mass humidity record data, the rock mass temperature record data, and the rock mass type record data, and set it as the true value of the rock mass hardness; Using the true value of the rock mass hardness as the supervision, and using the rock mass humidity record data, the rock mass temperature record data, and the rock mass type record data, train a random forest to obtain a rock mass hardness predictor; Process the rock mass humidity distribution topology, the rock mass temperature distribution topology, and the rock mass type distribution topology through the rock mass hardness predictor to obtain the rock mass hardness distribution information; Based on the rock mass hardness threshold, partition the rock mass hardness distribution information to obtain the rock mass hardness distribution topology.

8. The method according to claim 1, wherein Through an image acquisition device, collect mine muckpile monitoring images and perform image feature extraction to obtain the muckpile monitoring morphology, including: Collect multiple groups of data, where any one of the multiple groups of data includes muckpile image data and a label indicating the muckpile morphology parameters; According to the multiple groups of data, train a muckpile morphology parameter extraction model through a convolutional neural network; Process the mine muckpile monitoring images according to the muckpile morphology parameter extraction model to obtain the muckpile monitoring morphology.

9. A system for abnormally identifying mine blasting monitoring images, characterized in that, For implementing the method according to any one of claims 1 to 8, the system includes: A rock mass property analysis module, configured to obtain the rock mass humidity distribution topology and the rock mass temperature distribution topology of the mine, and perform hardness zoning in combination with the rock mass type distribution topology to obtain the rock mass hardness distribution topology; A reference morphology acquisition module, configured to collect blasting samples from the blasting position list and the blasting level list with the rock mass type distribution topology, the rock mass fracture distribution topology, and the rock mass hardness distribution topology as constraints to obtain a blasting sample set, and perform high-frequency muckpile morphology mining to obtain the muckpile reference morphology; A monitoring morphology acquisition module, configured to collect mine muckpile monitoring images through an image acquisition device and perform image feature extraction to obtain the muckpile monitoring morphology; An image marking module, configured to send an uncontrollable mark of the mine muckpile monitoring image to the user terminal when the muckpile monitoring morphology is inconsistent with the muckpile reference morphology.

Images