Intelligent recognition system and method for blasting block size based on deep learning
By using a deep learning-based intelligent recognition system, rock image features are automatically extracted and block size segmented and classified, solving the problems of low efficiency and poor accuracy in traditional methods. This enables high-precision block size recognition and visualization in complex scenarios, and supports blasting effect evaluation and parameter optimization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HENAN UNIVERSITY
- Filing Date
- 2026-03-24
- Publication Date
- 2026-06-26
AI Technical Summary
Traditional methods for identifying the size of blasted blocks are inefficient and inaccurate, making it difficult to achieve high-precision identification under harsh conditions such as complex lighting, shadows, and dust interference.
A deep learning-based intelligent recognition system is adopted. Through image preprocessing, feature extraction and block size analysis, convolutional neural network models such as ResNet and EfficientNet are used to automatically extract rock image features. The YOLOv8 Seg model is used for block size segmentation and classification, and the OpenCV algorithm is combined for size measurement and visualization.
It achieves accurate identification of rock block size in complex scenarios, improves identification accuracy and robustness, and generates intuitive block size distribution curves, providing a scientific basis for blasting effect evaluation and parameter optimization.
Smart Images
Figure CN122289715A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of image data processing technology, and more specifically to a deep learning-based intelligent identification system and method for explosive block density. Background Technology
[0002] Blasting technology is a commonly used rock breaking method in mining and geotechnical engineering. The quality of blasting directly affects subsequent mining efficiency, safe production, and resource utilization. The distribution of rock block size after blasting is one of the important indicators for evaluating the quality of blasting. An ideal block size distribution can reduce the cost of subsequent crushing and transportation, increase ore recovery rate, and reduce safety hazards.
[0003] Traditional methods for identifying the size of blasted blocks mainly include manual visual inspection, sampling and sieving, and image processing. Manual visual inspection is inefficient, highly subjective, and difficult to guarantee accuracy. While sampling and sieving offers higher precision, it is cumbersome, time-consuming, labor-intensive, costly, and destructive, making it impossible to measure all rock blocks. Traditional image processing methods, although capable of some automation, are limited by the robustness and applicability of their algorithms; under harsh conditions such as complex lighting, shadows, and dust interference, their accuracy often falls short of practical requirements. Summary of the Invention
[0004] To address the technical problem of poor accuracy in blasting block size recognition, this invention proposes a deep learning-based intelligent blasting block size recognition system and method. The system includes: acquiring rock images after blasting using image acquisition equipment; an image preprocessing module performing specialized processing on the rock images acquired under complex blasting conditions: enhancing image clarity and contrast through histogram equalization and sharpening techniques; eliminating dust and noise interference using algorithms such as Gaussian filtering and median filtering; and achieving accurate separation of rock areas from the background through threshold segmentation and edge detection, laying a high-quality data foundation for subsequent feature extraction and block size recognition. Subsequently, convolutional neural network models such as ResNet and EfficientNet are used to automatically mine deep features such as texture and contour of the rocks in the preprocessed images, eliminating the need for manual feature parameter design and significantly improving the efficiency and accuracy of feature extraction. Based on the extracted feature vectors, the block size recognition module completes accurate segmentation and category labeling of rock blocks using the YOLOv8 Seg model, and accurately measures key dimensions such as the equivalent diameter and area of individual rock blocks using image processing algorithms such as OpenCV. Next, the block size analysis module performs systematic statistical analysis on the aforementioned identification and measurement data. On one hand, it calculates the proportion of rocks in different grain size ranges and plots block size distribution curves such as frequency distribution histograms and cumulative distribution curves to quantify the distribution characteristics of the blasted block size. Then, the results visualization module presents the block size identification and distribution analysis results intuitively through various forms such as image overlay annotation (merging the identification results with the original image) and chart visualization (line charts, bar charts, etc.), facilitating engineers' quick understanding of the blasting effect.
[0005] In a first aspect, the present invention provides a method for intelligent identification of burst block density based on deep learning, the method comprising: Step 1: Use an image acquisition device to acquire images of the rock after blasting, and use an image preprocessing module to enhance, reduce noise, and segment the images of the rock after blasting to obtain a preprocessed image; Step 2: Based on the preprocessed image, automatically extract image features using the convolutional neural network model in the feature extraction module; Step 3: Based on the extracted image features, the rock block size is segmented, classified, and its size is measured using the block size recognition module; Step 4: Based on the segmentation, classification, and size measurement results, use the block size distribution analysis module to perform statistical analysis on the identification results, generate a block size distribution curve, and obtain the analysis results; Step 5: Use the results visualization module to display the identification and analysis results in the form of images and charts.
[0006] Optionally, the image preprocessing module includes: an image enhancement unit, a noise reduction unit, and an image segmentation unit; the image enhancement unit uses histogram equalization, sharpening, and contrast adjustment techniques to enhance the image; the noise reduction unit uses Gaussian filtering, median filtering, or bilateral filtering algorithms to reduce noise; and the image segmentation unit uses threshold segmentation, edge detection, and region growing methods to separate the rock region from the background.
[0007] Optionally, the feature extraction module uses ResNet or EfficientNet convolutional neural network models and is pre-trained through transfer learning to automatically extract deep features such as texture and contour of rock images.
[0008] Optionally, the block size recognition module uses the YOLOv8 Seg model for block size segmentation and classification, and combines image processing algorithms to complete the measurement of rock block size. The size measurement covers key parameters such as equivalent diameter, and the classification includes categories such as small blocks, medium blocks, large blocks, and extra-large blocks.
[0009] Optionally, the block size distribution analysis module includes: a block size statistics unit, a block size distribution curve plotting unit, and a block size visualization unit.
[0010] Optionally, the steps implemented by the block degree statistics unit include: The particle size was divided into 13 intervals: 0-10cm, 10-20cm, 20-30cm, 30-40cm, 40-50cm, 50-60cm, 60-70cm, 70-80cm, 80-90cm, 90-100cm, 100-110cm, 110-120cm, and above 120cm. The proportion of rocks in each interval was counted.
[0011] Optionally, the steps implemented by the block size distribution curve drawing unit include: The cumulative distribution curve was generated using Python's Matplotlib library. The x-axis of the cumulative distribution curve represents the particle size, and the y-axis represents the cumulative quantity percentage. The curve was then smoothed and fitted.
[0012] Optionally, the steps implemented by the block-level visualization unit include: The core identification information of a single rock block, such as its category, equivalent diameter, and confidence level, is overlaid on the original image as a bounding box, and the block size distribution curve and interval statistics are displayed simultaneously. The system also provides a data interaction query function, allowing users to retrieve the size data and statistical proportion of the corresponding rock block by clicking on the bounding box or the chart interval.
[0013] Optionally, the result visualization module supports the display of the original image, segmented labeled image, and analysis charts, and overlays the recognition results onto the original image in the form of labeled boxes, supporting user interaction.
[0014] Secondly, this invention provides a deep learning-based intelligent identification system for blasted rock block size, comprising: an image acquisition device for acquiring images of rocks after blasting, a data processing and computing server for running a deep learning model and processing data, a data storage device for storing various types of data, and a software system running on the data processing and computing server; the software system includes an image preprocessing module for optimizing the original image, a feature extraction module for automatically extracting rock features from the image, a block size identification module for realizing rock block size segmentation, classification, and size measurement, a block size distribution analysis module for statistically analyzing the identification results, and a result visualization module for presenting the identification and analysis results; each hardware device and software module achieves bidirectional data interaction through a data transmission interface, forming a closed-loop processing system.
[0015] The present invention has the following beneficial effects: This system automatically extracts image features using deep learning technology, eliminating the need for manual feature design and improving efficiency and accuracy. Advanced image segmentation and classification algorithms enable precise identification of rock block sizes in complex scenes, enhancing recognition accuracy and robustness. The block size analysis module quickly generates block size distribution curves, providing a scientific basis for blasting effect evaluation and blasting parameter optimization. The results visualization module allows users to intuitively understand the blasting effects. Attached Figure Description
[0016] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0017] Figure 1 This is a schematic diagram of the overall architecture of a deep learning-based intelligent identification system for explosive block density according to the present invention. Figure 2 This is a flowchart of a deep learning-based intelligent identification method for blasting block density according to the present invention; Figure 3 This is a schematic diagram of the block size recognition module processing flow of the present invention; Figure 4 This is a line graph showing the block size distribution of the present invention; Figure 5 This is a histogram showing the block size distribution of the present invention; Figure 6 This is a schematic diagram of the explosive block size detection interface of the present invention. Detailed Implementation
[0018] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the specific implementation methods, structures, features, and effects of the technical solution proposed according to the present invention are described in detail below with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.
[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0020] The project background is as follows: A large open-pit iron mine located in North China has an annual mining scale of 80 million tons. The main ore is granite, with a Protodyakonov hardness of f=12-15 and well-developed joints and fissures. The mine has long used bottom-hole single-point blasting for blasting operations, which has resulted in problems such as uneven distribution of blasted blocks, a high percentage of large blocks (approximately 18%), and high costs for secondary crushing. To optimize blasting parameters and improve blasting effectiveness, the mine introduced the intelligent blasting block size recognition system based on deep learning, as described in this invention. Effect evaluation and parameter optimization experiments were conducted on blasting operations at the EL.320-305 platform in the mining area. The single blasting area of this platform is approximately 2000㎡. The designed borehole diameter is 165mm, the hole depth is 15m (including 1.5m ultra-deep), and the borehole network parameters are 7.5m × 4.0m. On-site mixed emulsion explosives (detonation velocity 4600-5100m / s) were used. The test group used double-point initiation within the borehole (6m distance between the upper and lower initiation points), while the control group used traditional bottom-hole initiation. Each group had 24 boreholes, and the initiation network used electronic detonators with a 9ms delay between boreholes and a 25ms delay between rows.
[0021] refer to Figure 1This invention provides a deep learning-based intelligent identification system for blasted rock block size. The system includes an image acquisition device for collecting images of rocks after blasting, a data processing and computing server for running a deep learning model and processing data, a data storage device for storing various types of data, and a software system running on the data processing and computing server. The software system includes an image preprocessing module for optimizing the original images, a feature extraction module for automatically extracting rock features from the images, a block size identification module for segmenting, classifying, and measuring the size of rock blocks, a block size distribution analysis module for statistically analyzing the identification results, and a result visualization module for intuitively presenting the identification and analysis results. The hardware devices and software modules interact bidirectionally through a data transmission interface, forming a closed-loop processing system encompassing "acquisition—processing—analysis—display—output."
[0022] The image acquisition equipment includes high-resolution digital cameras, industrial cameras, or cameras mounted on drones. The acquired images must meet the clarity and coverage requirements of the rock piles at the blasting site, without significant blurring or distortion; that is, the image acquisition equipment is mainly used to acquire images of the rock piles after blasting. The data processing and computing server is equipped with a multi-core CPU, a high-performance GPU, and large-capacity memory, possessing the capability to support the efficient operation of deep learning models and the rapid processing of massive amounts of image data, and is used to run deep learning models. Data storage devices include local hard drives, network storage servers, or cloud storage services, used to securely store raw image data, model data, intermediate processing results, and final recognition reports, supporting data backup and traceability.
[0023] refer to Figure 2 This paper illustrates the flowchart of a deep learning-based intelligent identification method for blasted block size according to the present invention. The deep learning-based intelligent identification method for blasted block size includes the following steps: Step 1: Use an image acquisition device to acquire images of the rocks after blasting, and use an image preprocessing module to enhance, reduce noise, and segment the images to obtain preprocessed images.
[0024] In some embodiments, image acquisition and preprocessing are performed as follows: A full-coverage image of the blasted rock deposit is acquired using an image acquisition device (high-resolution digital camera, industrial camera, or camera mounted on a drone) to obtain original images without significant blurring or distortion. The original images are then imported into the image preprocessing module, where histogram equalization, sharpening, and contrast adjustment techniques of the image enhancement unit are sequentially used to improve image clarity and rock-background contrast. Noise from dust and lighting interference is reduced using Gaussian filtering, median filtering, or bilateral filtering algorithms of the noise reduction unit. Finally, threshold segmentation, edge detection, and region growing methods of the image segmentation unit are used to achieve precise separation of the rock region from the background, outputting an optimized image.
[0025] The image preprocessing module includes an image enhancement unit, a noise reduction unit, and an image segmentation unit. These units optimize the original image, improving image quality and laying the foundation for subsequent feature extraction and block size recognition. The image enhancement unit employs histogram equalization, sharpening, and contrast adjustment techniques to enhance image clarity and contrast between the rock and background. The noise reduction unit uses Gaussian filtering, median filtering, or bilateral filtering algorithms to effectively reduce noise caused by dust, lighting, and other interference. The image segmentation unit utilizes threshold segmentation, edge detection, and region growing methods to accurately separate the rock region from the background, laying the foundation for subsequent processing.
[0026] As an example, a high-resolution camera mounted on a drone can be used to take aerial photos of the blasting area, obtain images of the rocks after the blasting, and then perform histogram equalization enhancement, Gaussian filtering noise reduction, and threshold segmentation on the original images to improve image quality.
[0027] Step 2: Based on the preprocessed image, automatically extract image features using the convolutional neural network model in the feature extraction module.
[0028] In some embodiments, automatic feature extraction is performed: the optimized image output from step 1 is input into the feature extraction module, and convolutional neural network models such as ResNet and EfficientNet, which have been pre-trained by transfer learning, are used to automatically mine deep features such as texture and contour of the rock image and generate feature vectors with uniform dimensions. No manual design of feature parameters is required, which ensures the efficiency and accuracy of feature extraction.
[0029] The feature extraction module employs convolutional neural network models such as ResNet and EfficientNet, pre-trained through transfer learning, to automatically extract deep features such as texture and contours from rock images without requiring manual design of feature parameters. In other words, the feature extraction module uses convolutional neural network models, including ResNet and EfficientNet, for automatic image feature extraction.
[0030] As an example, a pre-trained ResNet50 model can be used to extract image features and form a feature vector.
[0031] Step 3: Based on the extracted image features, the rock block size is segmented, classified, and its size is measured using the block size recognition module.
[0032] In some embodiments, block segmentation, classification, and size measurement: Based on the feature vector generated in step 2, the block recognition module uses the YOLOv8 Seg model to complete the accurate segmentation and category labeling of a single rock block. Then, combined with image processing algorithms, key size parameters such as the equivalent diameter of each rock block are calculated, and a list of recognition results containing block category, size data, confidence level, and coordinate position is output.
[0033] The block size recognition module uses the YOLOv8 Seg model for block size segmentation and classification, and combines image processing algorithms to complete the measurement of rock block size. The size measurement covers key parameters such as equivalent diameter, and the classification includes categories such as small blocks, medium blocks, large blocks and extra-large blocks.
[0034] It should be noted that the block size recognition module processing flow can be as follows: Figure 3 As shown.
[0035] As an example, the YOLOv8 Seg model can be used to segment and classify rock blocks, and the OpenCV library can be used to calculate the size of each block (such as equivalent diameter and area).
[0036] Step 4: Based on the segmentation, classification, and size measurement results, use the block size distribution analysis module to perform statistical analysis on the identification results, generate a block size distribution curve, and obtain the analysis results.
[0037] In some embodiments, the block size analysis module includes a block size distribution line chart and a block size distribution bar chart for statistical analysis of the identification results. Block size distribution analysis: The identification result list from step 3 is imported into the block size distribution analysis module. The block size statistics unit calculates the proportion of rocks in each of the 13 particle size ranges: 0-10cm, 10-20cm, 20-30cm, 30-40cm, 40-50cm, 50-60cm, 60-70cm, 70-80cm, 80-90cm, 90-100cm, 100-110cm, 110-120cm, and above 120cm. The block size distribution curve plotting unit uses Python's Matplotlib library to generate a cumulative distribution curve, visually presenting the block size distribution pattern.
[0038] The block size distribution analysis module includes: a block size statistics unit, a block size distribution curve plotting unit, and a block size visualization unit.
[0039] The block size statistical unit divides the block size into 13 intervals: 0-10cm, 10-20cm, 20-30cm, 30-40cm, 40-50cm, 50-60cm, 60-70cm, 70-80cm, 80-90cm, 90-100cm, 100-110cm, 110-120cm, and above 120cm. The percentage of rock in each interval is then counted.
[0040] The block size distribution curve plotting unit uses Python's Matplotlib library to generate a cumulative distribution curve. The horizontal axis of the cumulative distribution curve represents the block size particle size, and the vertical axis represents the cumulative quantity percentage. The curve is smoothed and fitted to intuitively present the block size distribution pattern.
[0041] The block size visualization unit supports overlaying core identification information such as the category, equivalent diameter, and confidence level of a single rock block onto the original image in the form of a labeled box, and simultaneously displays the block size distribution curve and interval statistics. It also provides a data interaction query function, allowing users to quickly retrieve detailed size data and statistical proportions of the corresponding rock block by clicking on the labeled box or chart interval, intuitively presenting the overall situation and local details of the blasting block size detection.
[0042] As an example, statistical analysis can be performed on the identified block sizes to generate a block size distribution line chart, such as... Figure 4 As shown in the block size distribution histogram, Figure 5 As shown.
[0043] Step 5: Use the results visualization module to display the identification and analysis results in the form of images and charts.
[0044] In some embodiments, the results are visualized: the results visualization module simultaneously displays the original image, segmented and labeled image (recognition results are overlaid in the form of labeled boxes), and block size distribution chart. Users can quickly retrieve detailed data and statistical proportions of the corresponding rock blocks by clicking on the labeled boxes or chart intervals. Report generation and blasting parameter optimization: the report generation module automatically generates reports based on standardized templates, including project information, image processing procedures, block size recognition results, and distribution analysis data.
[0045] The results visualization module supports the display of the original image, segmented labeled image, and analysis charts. It overlays the recognition results onto the original image in the form of labeled boxes and supports user interaction.
[0046] As an example, the recognition results can be displayed on the main interface, such as... Figure 6 The above is provided for reference by engineering and technical personnel.
[0047] In summary, utilizing deep learning technology to automatically extract image features eliminates the need for manual feature design, improving both efficiency and accuracy. Advanced image segmentation and classification algorithms enable precise identification of rock block sizes in complex scenes, enhancing recognition accuracy and robustness. The block size analysis module rapidly generates block size distribution curves, providing a scientific basis for blasting effect evaluation and blasting parameter optimization. The results visualization module facilitates intuitive understanding of the blasting effects for users.
[0048] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.
Claims
1. A method for intelligent identification of burst block density based on deep learning, characterized in that, Includes the following steps: Step 1: Use an image acquisition device to acquire images of the rock after blasting, and use an image preprocessing module to enhance, reduce noise, and segment the images of the rock after blasting to obtain a preprocessed image; Step 2: Based on the preprocessed image, automatically extract image features using the convolutional neural network model in the feature extraction module; Step 3: Based on the extracted image features, the rock block size is segmented, classified, and its size is measured using the block size recognition module; Step 4: Based on the segmentation, classification, and size measurement results, use the block size distribution analysis module to perform statistical analysis on the identification results, generate a block size distribution curve, and obtain the analysis results; Step 5: Use the results visualization module to display the identification and analysis results in the form of images and charts.
2. The method for intelligent identification of blasting block size based on deep learning according to claim 1, characterized in that, The image preprocessing module includes: an image enhancement unit, a noise reduction unit, and an image segmentation unit; the image enhancement unit uses histogram equalization, sharpening, and contrast adjustment techniques to enhance the image; the noise reduction unit uses Gaussian filtering, median filtering, or bilateral filtering algorithms to reduce noise; and the image segmentation unit uses threshold segmentation, edge detection, and region growing methods to separate the rock region from the background.
3. The method for intelligent identification of blasting block size based on deep learning according to claim 1, characterized in that, The feature extraction module uses ResNet and EfficientNet convolutional neural network models and is pre-trained through transfer learning to automatically extract deep features such as texture and contour of rock images.
4. The method for intelligent identification of blasting block size based on deep learning according to claim 1, characterized in that, The block size recognition module uses the YOLOv8 Seg model for block size segmentation and classification, and combines image processing algorithms to complete the measurement of rock block size. The size measurement covers key parameters such as equivalent diameter, and the classification includes categories such as small blocks, medium blocks, large blocks and extra-large blocks.
5. The method for intelligent identification of blasting block size based on deep learning according to claim 1, characterized in that, The block size distribution analysis module includes: a block size statistics unit, a block size distribution curve plotting unit, and a block size visualization unit.
6. The method for intelligent identification of blasting block size based on deep learning according to claim 5, characterized in that, The steps implemented by the block size statistics unit include: The particle size was divided into 13 intervals: 0-10cm, 10-20cm, 20-30cm, 30-40cm, 40-50cm, 50-60cm, 60-70cm, 70-80cm, 80-90cm, 90-100cm, 100-110cm, 110-120cm, and above 120cm. The proportion of rocks in each interval was counted.
7. The method for intelligent identification of blasting block size based on deep learning according to claim 5, characterized in that, The steps implemented by the block size distribution curve drawing unit include: The cumulative distribution curve was generated using Python's Matplotlib library. The x-axis of the cumulative distribution curve represents the particle size, and the y-axis represents the cumulative quantity percentage. The curve was then smoothed and fitted.
8. The method for intelligent identification of blasting block size based on deep learning according to claim 5, characterized in that, The steps implemented by the block-level visualization unit include: The core identification information of a single rock block, such as its category, equivalent diameter, and confidence level, is overlaid on the original image as a bounding box, and the block size distribution curve and interval statistics are displayed simultaneously. The system also provides a data interaction query function, allowing users to retrieve the size data and statistical proportion of the corresponding rock block by clicking on the bounding box or the chart interval.
9. The method for intelligent identification of blasting block size based on deep learning according to claim 1, characterized in that, The results visualization module supports the display of original images, segmented labeled images, and analysis charts. It overlays the recognition results onto the original image in the form of labeled boxes and supports user interaction.
10. A deep learning-based intelligent identification system for blasting block density, characterized in that, include: Image acquisition equipment for collecting images of rocks after blasting, a data processing and computing server for running deep learning models and processing data, a data storage device for storing various types of data, and a software system running on the data processing and computing server; the software system includes an image preprocessing module for optimizing the original images, a feature extraction module for automatically extracting rock features from the images, a block size recognition module for realizing rock block size segmentation, classification and size measurement, a block size distribution analysis module for statistical analysis of the recognition results, and a result visualization module for presenting the recognition and analysis results; each hardware device and software module realizes bidirectional data interaction through a data transmission interface, forming a closed-loop processing system.