Emulsion explosive detonation velocity prediction method and system based on digital image

Abstract

The invention relates to the technical field of detonation velocity prediction, in particular to an emulsion explosive detonation velocity prediction method and system based on a digital image. The prediction method comprises the following steps: S1, acquiring a cross section image of an emulsion explosive based on image acquisition equipment, establishing a mapping relation between the cross section image of the emulsion explosive and a detonation velocity measured value, and dividing the mapping relation into a training set, a test set and a verification set; s2, establishing a detonation velocity prediction model, and training, testing and verifying the detonation velocity prediction model by respectively adopting the training set, the test set and the verification set to obtain an emulsion explosive detonation velocity prediction model; and S3, obtaining a cross section image of the emulsion explosive during production, and outputting a predicted detonation velocity value in real time based on the cross section image of the emulsion explosive during production and the emulsion explosive detonation velocity prediction model. The prediction system is used for realizing the prediction method and comprises image acquisition equipment and processing equipment, the image acquisition equipment is in communication connection with the processing equipment, the real-time prediction of the detonation velocity of the emulsion explosive can be realized, and the detonation velocity detection efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of detonation velocity prediction, and particularly to a method and system for predicting the detonation velocity of emulsion explosive based on digital images. Background Art

[0002] The detonation velocity refers to the propagation speed of the explosion flame or its chemical reaction in the cartridge, which is a key parameter for measuring the performance and quality of industrial explosives.

[0003] In the manufacturing process of emulsion explosives, due to the lack of instruments for real-time monitoring of detonation velocity, the current production process cannot achieve the immediate detection of the detonation velocity of emulsion explosives. Enterprises in the civilian explosive industry usually adopt the method of conducting off-line tests on the detonation velocity of the finished emulsion explosives after production. However, the above method has hysteresis, cannot achieve real-time prediction of the detonation velocity of emulsion explosives, and has low detonation velocity detection efficiency. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for predicting the detonation velocity of emulsion explosive based on digital images, which can achieve real-time prediction of the detonation velocity of emulsion explosives and improve the detonation velocity detection efficiency.

[0005] To solve the above technical problem, the present invention provides the following technical solutions:

[0006] In the first aspect, the present invention provides a method for predicting the detonation velocity of emulsion explosive based on digital images, including the following steps:

[0007] S1: Obtain the cross-sectional image of the emulsion explosive based on the image acquisition device, establish the mapping relationship between the cross-sectional image of the emulsion explosive and the measured detonation velocity value, and divide it into a training set, a test set, and a validation set;

[0008] S2: Establish a detonation velocity prediction model, and use the training set, the test set, and the validation set to train, test, and validate the detonation velocity prediction model respectively to obtain the detonation velocity prediction model of the emulsion explosive;

[0009] S3: Obtain the cross-sectional image of the emulsion explosive during production, and based on the cross-sectional image of the emulsion explosive during production and the detonation velocity prediction model of the emulsion explosive, output the predicted detonation velocity value in real time;

[0010] S2 includes the following steps:

[0011] S21: Preprocess the cross-sectional image of the emulsion explosive to obtain a preprocessed image;

[0012] S22: Perform binary segmentation on the preprocessed image to obtain a binary image;

[0013] S23: Use the Canny edge extraction model to obtain the bubbles in the binary image and generate a bubble image;

[0014] S24: Input the bubble image into the detonation velocity prediction model to obtain high-dimensional features and low-dimensional features and output the detonation velocity prediction value.

[0015] By adopting the above technical solution, by collecting the cross-sectional image of the emulsion explosive and based on the cross-sectional image and the emulsion explosive detonation velocity prediction model, the detonation velocity of the cross-sectional image collected in real time is predicted, thereby realizing the real-time automatic prediction of the detonation velocity of the emulsion explosive based on digital images. The whole process saves time and effort, can realize the real-time prediction of the detonation velocity of the emulsion explosive, and improves the detonation velocity detection efficiency. Since the present invention predicts the detonation velocity by collecting the image of the emulsion explosive in real time, the detonation velocity predicted by the present invention is the real-time detonation velocity, which can meet the needs of civil explosive enterprises to predict the detonation velocity of emulsion explosives online in real time, so that they can adjust the production situation according to the detection results.

[0016] Optionally, the detonation velocity prediction model uses the Huber loss function as the model objective function for backpropagation, and the expression of the Huber loss function is:

[0017]

[0018] where y represents the measured detonation velocity value, f(x) represents the predicted detonation velocity value, and δ represents the hyperparameter of the Huber loss function.

[0019] Optionally, the step S23 includes:

[0020] Input the binary image into the Canny edge extraction model to extract several closed regions in the binary image where n represents the number of closed regions in the binary image. Screen the closed regions according to the shape characteristics of the bubbles, and draw the screened closed regions on the blank image to generate the bubble image.

[0021] Optionally, the step S24 includes:

[0022] Input the bubble image into the detonation velocity prediction model. The detonation velocity prediction model is designed based on the Encoder-Decoder architecture. The high-dimensional features of the bubble image are obtained by the Encoder part, the high-dimensional features are reduced in dimension by the Decoder part to obtain low-dimensional features, and the detonation velocity prediction value is output by the fully connected layer.

[0023] Optionally, the Encoder part consists of three layers of downsampling convolution. Each layer of downsampling convolution increases the input data dimension and reduces the input data size, thereby obtaining the high-dimensional features of the input data. The formula for each layer of downsampling convolution is:

[0024] Down(x) = B(σ(Conv 3×3 (x)))

[0025] Among them, B represents the batch normalization operation, σ represents the ReLU activation function, and Conv 3×3 represents a 3×3 convolution.

[0026] Optionally, the Decoder part is composed of three layers of upsampling convolutions. Each layer of upsampling convolution reduces the input data dimension and increases the input data size, converting the high-dimensional features of the Encoder into low-dimensional features. The formula for each layer of upsampling convolution is:

[0027] Up(x) = B(σ(Conv 3×3 (x))).

[0028] Optionally, the output of the full connection layer for predicting the detonation velocity of emulsion explosive includes:

[0029] Flatten the output of the Decoder part, and then input it into the full connection layer to predict the detonation velocity of emulsion explosive. The formula for the structure of the full connection layer is:

[0030] L(x) = Linear(σ(Linear(B(σ(Linear(x))))))

[0031] Among them, Linear represents the full connection layer.

[0032] Optionally, the preprocessing includes clipping processing and masking processing.

[0033] Optionally, the step S22 includes:

[0034] Input the preprocessed image into the binary segmentation stage, and perform binary segmentation operations through gray processing and binary threshold to generate a binary image; the setting of the binary threshold is to convert the pixel values of the cross-sectional image of emulsion explosive into gray scale, record the gray scale values of all collected cross-sectional images of emulsion explosive to construct a gray scale distribution histogram, and select the maximum and minimum gray scale values to be retained as the binary threshold according to the gray scale distribution histogram.

[0035] In a second aspect, the present invention provides a system for predicting the detonation velocity of emulsion explosive based on digital images, which is used to implement the method for predicting the detonation velocity of emulsion explosive based on digital images as described in the first aspect, including an image acquisition device and a processing device, and the image acquisition device and the processing device are communicatively connected.

[0036] In summary, the present invention at least includes the following beneficial technical effects:

[0037] The method for predicting the detonation velocity of emulsion explosive based on digital images can collect the cross-sectional images of the emulsion explosive, extract the bubble map of the current emulsion explosive from the cross-sectional images, and input the obtained bubble map into a pre-trained emulsion explosive detonation velocity prediction model to predict the detonation velocity of the current emulsion explosive. By collecting the cross-sectional images of the emulsion explosive and based on the cross-sectional images and the emulsion explosive detonation velocity prediction model, the detonation velocity of the cross-sectional images collected in real time is predicted, thereby realizing the online real-time automatic detection of the detonation velocity of the emulsion explosive based on digital images. The whole process saves time and effort and can greatly improve the detection efficiency of the detonation velocity of the emulsion explosive. Since the present invention collects the images of the emulsion explosive in real time, the detonation velocity predicted by the present invention is the real-time detonation velocity, which can meet the needs of civil explosive enterprises to predict the detonation velocity of emulsion explosives in real time online, enabling them to adjust the production process parameters according to the detection results. Description of the Drawings

[0038] Figure 1 It is a flowchart of the method for predicting the detonation velocity of emulsion explosive based on digital images in an embodiment of the present invention.

[0039] Figure 2 It is a flowchart of step S2 in the method for predicting the detonation velocity of emulsion explosive based on digital images.

[0040] Figure 3 It is a structural diagram of the detonation velocity prediction network model based on the Encoder-Decoder architecture in an embodiment of the present invention.

[0041] Figure 4 It is a schematic diagram of the process of processing digital images of emulsion explosive in an embodiment of the present invention.

[0042] Figure 5 It is a schematic diagram of the structure of the detonation velocity prediction system of emulsion explosive based on digital images in an embodiment of the present invention.

[0043] Description of the reference numerals: 1. Processing device; 2. Image acquisition device. Detailed Embodiments

[0044] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0045] The terms used in the following embodiments of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification and appended claims of the present invention, the singular forms "a", "an", "the", "above-mentioned", "said", "this" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used in the present invention refers to and includes any or all possible combinations of one or more of the listed items. The term "exemplary" means "serving as an example, embodiment or illustration", and any embodiment described as "exemplary" here does not have to be construed as superior or better than other embodiments. The terms "first" and "second" are only used for descriptive purposes and cannot be construed as implying or suggesting 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 such features. In the description of the embodiments of the present invention, unless otherwise specified, the meaning of "a plurality" is two or more.

[0046] An embodiment of the present invention provides a method for predicting the detonation velocity of emulsion explosive based on digital images.

[0047] Reference Figure 1 , a method for predicting the detonation velocity of emulsion explosive based on digital images, comprising the following steps:

[0048] S1: Based on the image acquisition device 2, obtain the cross-sectional image of the emulsion explosive, establish the mapping relationship between the cross-sectional image of the emulsion explosive and the measured detonation velocity value, and divide it into a training set, a test set, and a validation set.

[0049] When producing the emulsion explosive, use the image acquisition device 2 to photograph the cross-section of the emulsion explosive to obtain the cross-sectional image of the emulsion explosive. The processing device 1 can obtain the cross-sectional image of the emulsion explosive through the image acquisition device 2.

[0050] After obtaining the cross-sectional image of the emulsion explosive, obtain the measured detonation velocity value of the emulsion explosive through manual measurement, establish the mapping relationship between the cross-sectional image of the emulsion explosive and the measured detonation velocity value, that is, make the cross-sectional image of the emulsion explosive and the measured detonation velocity value correspond one by one, and then divide it into a training set, a test set, and a validation set.

[0051] S2: Establish a detonation velocity prediction model, and use the training set, the test set, and the validation set to train, test, and validate the detonation velocity prediction model respectively to obtain the detonation velocity prediction model of the emulsion explosive.

[0052] The processing device 1 establishes a detonation velocity prediction model, and then trains the detonation velocity prediction model with a training set. After the training is completed, the detonation velocity prediction model is tested with a test set. After the test is completed, the detonation velocity prediction model is verified with a validation set, so as to complete the iterative training of the detonation velocity prediction model and obtain an emulsion explosive detonation velocity prediction model.

[0053] It should be understood that whether using a training set, a test set or a validation set, essentially, the emulsion explosive cross-sectional image is used to iteratively train the detonation velocity prediction model.

[0054] The detonation velocity prediction model uses the Huber loss function as the model objective function for backpropagation. The expression of the Huber loss function is:

[0055]

[0056] where y represents the measured detonation velocity value, f(x) represents the predicted detonation velocity value, and δ represents the hyperparameter of the Huber loss function.

[0057] Specifically, referring to Figure 2 , the step S2 includes the iterative training of the following steps S21 - S24.

[0058] S21: Preprocess the emulsion explosive cross-sectional image to obtain a preprocessed image.

[0059] After the processing device 1 obtains the emulsion explosive cross-sectional image, it preprocesses the emulsion explosive cross-sectional image. The preprocessing includes cropping processing and masking processing. After the preprocessing is completed, the preprocessed image I p .

[0060] Specifically, the masking processing fixes the imaging position of the emulsion explosive in the image acquisition device 2, that is, the emulsion explosive will be in the same position in each collected emulsion explosive cross-sectional image. By setting a blank image with the same resolution as the emulsion explosive cross-sectional image, setting the pixel values of the positions where the emulsion explosive appears to 1 and the pixel values of the remaining positions to 0, multiplying the blank image by the collected emulsion explosive cross-sectional image can extract only the emulsion explosive image in the emulsion explosive cross-sectional image.

[0061] S22: Perform binary segmentation on the preprocessed image to obtain a binary image.

[0062] The processing device 1 inputs the preprocessed image I p into the binary segmentation stage, and performs binary segmentation operations through gray-scale processing and binary segmentation thresholds to generate a binary image I b .

[0063] Among them, the binarization threshold is set according to the statistical results after the pixel value distribution of the cross-sectional image of the emulsion explosive is statistically analyzed. Specifically, when setting the binarization threshold, the pixel values of the cross-sectional image of the emulsion explosive are converted to grayscale, and the grayscale values of all collected cross-sectional images of the emulsion explosive are recorded to construct a grayscale distribution histogram, which will show a similar normal distribution. The maximum and minimum grayscale values selected from the grayscale distribution histogram are used as the binarization threshold.

[0064] S23: Use the Canny edge extraction model to obtain the bubbles in the binarized image and generate a bubble image.

[0065] The processing device 1 inputs the binarized image I b into the Canny edge extraction model, and can extract several closed regions in the binarized image I b where n represents the number of closed regions in the binarized image I According to the shape characteristics of the bubbles, the closed regions are screened, and the screened closed regions are drawn on a blank image, that is, the RGB pixel values in the closed regions are set to 1, and the remaining pixel values are set to 0 to generate a bubble image I b g .

[0066] Among them, the screening conditions are mainly obtained by manually detecting and statistically analyzing the appearance characteristics of the bubbles in the emulsion explosive, that is, the size of the bubbles that appear, the contour shape needs to be approximately circular, and there are no overlapping bubbles. The pixel area range of the bubbles should be within 0 - 50 pixel (pixels), and the value of this pixel area range can be adaptively modified according to the actual collected images.

[0067] The schematic diagrams of the images generated in the process of steps S21 - S23 are as Figure 3 shown.

[0068] S24: Input the bubble image into the detonation velocity prediction model to obtain high-dimensional features and low-dimensional features and output the detonation velocity prediction value.

[0069] As Figure 4 shown, the processing device 1 inputs the bubble image I g into the detonation velocity prediction model, which is designed based on the Encoder-Decoder architecture. Then, the Encoder part obtains the high-dimensional features of the bubble image I g , and then the Decoder part reduces the dimension of the high-dimensional features to obtain low-dimensional features, and finally the fully connected layer outputs the detonation velocity prediction value.

[0070] ​When the predicted detonation velocity value and the measured detonation velocity value are within the preset error range, it means that the iterative training using the training set is completed, and an emulsion explosive detonation velocity prediction model is obtained. Then, the emulsion explosive detonation velocity prediction model can be tested and verified using the test set and the validation set in turn.

[0071] Specifically, the Encoder part consists of three layers of downsampling convolution. Each layer of downsampling convolution increases the dimension of the input data and reduces the size of the input data, thereby obtaining the high-dimensional features of the input data. The formula for each layer of downsampling convolution is:

[0072] Down(x) = B(σ(Conv 3×3 (x)))

[0073] where B represents the batch normalization operation, σ represents the ReLU activation function, and Conv 3×3 represents a 3×3 convolution.

[0074] The Decoder part consists of three layers of upsampling convolution. Each layer of upsampling convolution reduces the dimension of the input data and increases the size of the input data, converting the high-dimensional features of the Encoder into low-dimensional features. The formula for each layer of upsampling convolution is:

[0075] Up(x) = B(σ(Conv 3×3 (x)))

[0076] Flatten the output of the Decoder part, that is, arrange the two-dimensional data in a one-dimensional data manner in sequence, and then input it into the fully connected layer to predict the detonation velocity of the emulsion explosive. The formula for the fully connected layer structure is:

[0077] L(x) = Linear(σ(Linear(B(σ(Linear(x))))))

[0078] where Linear represents the fully connected layer.

[0079] S3: Obtain the cross-sectional image of the emulsion explosive during production, and based on the cross-sectional image of the emulsion explosive during production and the emulsion explosive detonation velocity prediction model, output the predicted detonation velocity value in real time.

[0080] Deploy the emulsion explosive detonation velocity prediction model to the actual detection environment for real-time emulsion explosive detonation velocity prediction experiments. After the processing device 1 obtains the cross-sectional image of the emulsion explosive during production through the image acquisition device 2, it can output the predicted detonation velocity value in real time through the trained emulsion explosive detonation velocity prediction model.

[0081] The embodiment of the present invention also provides an emulsion explosive detonation velocity prediction system based on digital images.

[0082] Reference Figure 5, An emulsion explosive detonation velocity prediction system based on digital images, comprising: an image acquisition device 2 and a processing device 1, and the image acquisition device 2 and the processing device 1 are communicatively connected. The various change methods and specific examples in the method provided in the above embodiments are applicable to the emulsion explosive detonation velocity prediction system based on digital images in this embodiment. Through the foregoing detailed description of the emulsion explosive detonation velocity prediction method based on digital images, those skilled in the art can clearly know the implementation method of the emulsion explosive detonation velocity prediction system based on digital images in this embodiment. For the sake of simplicity of the specification, it will not be elaborated here.

[0083] Among them, the processing device 1 can be implemented in various forms, including devices such as tablet computers, palmtop computers, laptop computers, and desktop computers.

[0084] As mentioned above, the above embodiments are only used to introduce the technical solutions of the present invention in detail, but the descriptions of the above embodiments are only used to help understand the method and its core idea of the present invention, and should not be construed as a limitation of the present invention. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for predicting detonation velocity of emulsion explosives based on digital images, characterized in that: The following steps are involved: S1: acquiring a cross-sectional image of the emulsion explosive based on the image acquisition device (2), establishing a mapping relationship between the cross-sectional image of the emulsion explosive and the measured value of the detonation velocity, and dividing the image into a training set, a test set, and a validation set; S2: Establishing a detonation velocity prediction model, respectively using a training set, a test set, and a validation set to train, test, and validate the detonation velocity prediction model, and obtaining an emulsion explosive detonation velocity prediction model; S3: obtaining a cross-sectional image of the emulsion explosive during production, and outputting a predicted detonation velocity value in real time based on the cross-sectional image of the emulsion explosive during production and an emulsion explosive detonation velocity prediction model; S2 includes the following steps: S21: preprocessing the emulsion explosive cross-section image to obtain a preprocessed image; S22: performing binary segmentation on the preprocessed image to obtain a binary image; S23: using the Canny edge extraction model to obtain bubbles in the binary image and generate a bubble image; S24: Input the bubble image into the detonation velocity prediction model, obtain high-dimensional features and low-dimensional features, and output the detonation velocity prediction value.

2. The method for predicting detonation velocity of emulsion explosives based on digital images according to claim 1, characterized in that: The detonation velocity prediction model uses the Huber loss function as the model objective function for back propagation. The expression of the Huber loss function is: Among them, y represents the measured value of detonation velocity, f(x) represents the predicted value of detonation velocity, and δ represents the hyperparameter of Huber loss function.

3. The method for predicting detonation velocity of emulsion explosives based on digital images according to claim 1, characterized in that: The step S23 comprises: Input the binary image into the Canny edge extraction model to extract several closed areas in the binary image Where n represents the number of closed areas in the binary image. The closed areas are screened according to the shape characteristics of the bubbles, and the screened closed areas are drawn on the blank image to generate a bubble image.

4. The method for predicting detonation velocity of emulsion explosives based on digital images according to claim 1, characterized in that: The step S24 comprises: The bubble image is input into the detonation velocity prediction model, which is designed based on the Encoder-Decoder architecture. The Encoder part obtains the high-dimensional features of the bubble image, and the Decoder part reduces the high-dimensional features to obtain low-dimensional features, and the fully connected layer outputs the detonation velocity prediction value.

5. The method for predicting detonation velocity of emulsion explosives based on digital images according to claim 4, characterized in that: The encoder part consists of three layers of downsampling convolution. Each layer of downsampling convolution increases the dimension of the input data and reduces the size of the input data, thereby obtaining the high-dimensional features of the input data. The formula for each layer of downsampling convolution is: Down(x)=B(σ(Conv 3×3 (x))) Among them, B represents the batch normalization operation, σ represents the ReLU activation function, Conv 3×3 Represents a 3×3 convolution.

6. The method for predicting detonation velocity of emulsion explosives based on digital images according to claim 5, characterized in that: The Decoder part consists of three layers of upsampling convolution. Each layer of upsampling convolution reduces the dimension of input data and increases the size of input data by converting the high-dimensional features of the Encoder into low-dimensional features. The formula for each layer of upsampling convolution is: Up(x)=B(σ(Conv 3×3 (x)))。 7. The method for predicting detonation velocity of emulsion explosives based on digital images according to claim 6, characterized in that: The detonation velocity prediction value output by the fully connected layer includes: The output of the decoder is flattened and then input into the fully connected layer to predict the detonation velocity of the emulsion explosive. The structural formula of the fully connected layer is: L(x)=Linear(σ(Linear(B(σ(Linear(x)))))) Among them, Linear represents the fully connected layer.

8. The method for predicting detonation velocity of emulsion explosives based on digital images according to claim 1, characterized in that: The preprocessing includes a clipping process and a masking process.

9. The method for predicting detonation velocity of emulsion explosives based on digital images according to claim 1, characterized in that: The step S22 comprises: The preprocessed image is input into the binarization segmentation stage, and the binarization segmentation operation is performed through grayscale processing and binarization threshold to generate a binarized image; the setting of the binarization threshold is to convert the pixel value of the emulsion explosive cross-section image into grayscale, record the grayscale values ​​of all collected emulsion explosive cross-section images to construct a grayscale distribution histogram, and select the maximum grayscale value and the minimum grayscale value retained as the binarization threshold according to the grayscale distribution histogram.

10. A digital image-based emulsion explosive detonation velocity prediction system, used to implement the digital image-based emulsion explosive detonation velocity prediction method according to any one of claims 1 to 9, characterized in that: It comprises an image acquisition device (2) and a processing device (1), wherein the image acquisition device (2) and the processing device (1) are communicatively connected.

Images