Aggregate digital image generation method and device, electronic equipment and storage medium

Abstract

The invention relates to an aggregate digital image generation method and device, electronic equipment and a storage medium, and belongs to the technical field of image processing.The aggregate digital image generation method comprises the steps that an aggregate real image set corresponding to target aggregate is acquired, and the aggregate real image set serves as input of an image generation model; determining the output of the image generation model as an aggregate digital image set corresponding to the target aggregate; wherein the image generation model is obtained by training based on an infoGAN-L model, and the infoGAN-L model is obtained by replacing a generator activation function of the infoGAN model with a Leaky ReLU function. According to the method, the generation efficiency, quality and diversity of the aggregate digital image set corresponding to the target aggregate are improved, and the reliability of subsequent aggregate morphological characteristic research is improved.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a method, device, electronic equipment and storage medium for generating digital images of aggregates. Background Art

[0002] Aggregate is one of the main components of asphalt mixture. Its angularity, shape and other morphological characteristics are important factors affecting the mechanical properties of asphalt mixture. In recent years, with the development of computer technology and image processing technology, researchers have gradually established sophisticated digital aggregate models to explore the influence of aggregate morphology on the performance of asphalt mixture.

[0003] The current aggregate generation method mostly uses the random aggregate method, which mainly simplifies the complex particle morphology and constructs basic aggregate shapes such as circles, ellipses, and polygons. However, the aggregates generated by this method are quite different from the real aggregates in shape, and it is impossible to establish a high-precision virtual asphalt mixture. Therefore, some scholars began to use image processing methods to collect real digital images of aggregates and then randomly place them to establish asphalt mixture specimens. However, this method can only obtain a small number of real aggregate images, which can neither fully represent all possible aggregate characteristics nor cover all aggregate variability, which ultimately leads to the lack of universality and reliability of the research results.

[0004] Existing approaches to characterize and expand aggregate morphology using image processing techniques lack quality and diversity in the images they generate, resulting in low accuracy in subsequent studies of aggregate morphological characteristics. Summary of the invention

[0005] In view of this, it is necessary to provide a method, device, electronic device and storage medium for generating digital images of aggregates, so as to solve the problem that the images generated by the existing digital image generation schemes of aggregates are lacking in quality and diversity, thus resulting in low accuracy of subsequent research on aggregate morphological characteristics.

[0006] In order to solve the above problems, the present invention provides a method for generating a digital image of aggregate, comprising: Acquire a real aggregate image set corresponding to the target aggregate, and use the real aggregate image set as input of the image generation model; Determining the output of the image generation model as an aggregate digital image set corresponding to the target aggregate; The image generation model is trained based on the infoGAN-L model, and the infoGAN-L model is obtained by replacing the generator activation function of the infoGAN model with the LeakyReLU function.

[0007] In one possible implementation, the image generation model is trained based on a sample aggregate real image set, and the training goal of the image generation model is to ensure that the similarity between the aggregate digital image set corresponding to the sample aggregate real image set generated by the image generation model and the sample aggregate real image set is greater than a preset threshold.

[0008] In a possible implementation, the batch size of the image generation model during training is 16, the learning rate is 0.0001, the number of training times is 2000, and the convolution kernel size is 5*5.

[0009] In one possible implementation, the image generation effect of the image generation model is evaluated based on the similarity of the image distribution and two-dimensional morphological parameters of the first aggregate digital image set with the real aggregate image, as well as the diversity and edge detection results of the first aggregate digital image set. The first aggregate digital image set is the aggregate digital image set corresponding to the real image set of the sample aggregate generated by the image generation model.

[0010] In a possible implementation, the two-dimensional morphological parameters include roundness and convex hull algorithm angularity.

[0011] In a possible implementation, the similarity between the image distribution of the first aggregate digital image set and the real aggregate image is determined based on a t-SNE algorithm.

[0012] In a possible implementation, the diversity of the first aggregate digital image set is determined based on the Fréchet distance of the first aggregate digital image set.

[0013] The present invention also provides a device for generating a digital image of aggregates, comprising: An acquisition module, used for acquiring a real aggregate image set corresponding to a target aggregate, and using the real aggregate image set as an input of an image generation model; A determination module, used for determining the output of the image generation model as an aggregate digital image set corresponding to the target aggregate; The image generation model is trained based on the infoGAN-L model, and the infoGAN-L model is obtained by replacing the generator activation function of the infoGAN model with the LeakyReLU function.

[0014] The present invention also provides an electronic device, including a memory and a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the aggregate digital image generation method as described above is implemented.

[0015] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for generating a digital image of aggregate as described above is implemented.

[0016] The beneficial effects of the present invention are as follows: the aggregate digital image generation method, device, electronic device and storage medium provided by the present invention generate an aggregate digital image set corresponding to the target aggregate through the infoGAN-L model. While improving the quality and diversity of the generated images, since the generator activation function of the infoGAN model is modified from the ReLU function to the LeakyReLU function, the learning ability of the network is improved, the gradient disappearance is avoided, thereby increasing the convergence speed, and further improving the efficiency of image generation. The present invention improves the generation efficiency, quality and diversity of the aggregate digital image set corresponding to the target aggregate, thereby improving the reliability of subsequent research on aggregate morphological characteristics. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic diagram of a flow chart of an embodiment of a method for generating a digital image of aggregate provided by the present invention; Figure 2 A schematic diagram of an embodiment of the infoGAN model network structure provided by the present invention; Figure 3 A schematic diagram of an embodiment of the infoGAN model generator and discriminator network structure provided by the present invention; Figure 4 A schematic structural diagram of an embodiment of a device for generating digital images of aggregates provided by the present invention; Figure 5 A schematic structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0019] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0020] In the description of the present invention, reference to an "embodiment" means that a particular feature, structure, or characteristic described in conjunction with the embodiment may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive of other embodiments. It is explicitly and implicitly understood by those skilled in the art that the described embodiments may be combined with other embodiments.

[0021] Aggregate is one of the main components of asphalt mixture. Its angularity, shape and other morphological characteristics are important factors affecting the mechanical properties of asphalt mixture. In recent years, with the development of computer technology and image processing technology, researchers have gradually established sophisticated digital aggregate models to explore the influence of aggregate morphology on the performance of asphalt mixture.

[0022] The current aggregate generation method mostly uses the random aggregate method, which mainly simplifies the complex particle morphology and constructs basic aggregate shapes such as circles, ellipses, and polygons. However, the aggregates generated by this method are quite different from the real aggregates in shape, and it is impossible to establish a high-precision virtual asphalt mixture. Therefore, some scholars began to use image processing methods to collect real digital images of aggregates and then randomly place them to establish asphalt mixture specimens. However, this method can only obtain a small number of real aggregate images, which can neither fully represent all possible aggregate characteristics nor cover all aggregate variability, which ultimately leads to the lack of universality and reliability of the research results.

[0023] Existing approaches to characterize and expand aggregate morphology using image processing techniques lack quality and diversity in the images they generate, resulting in low accuracy in subsequent studies of aggregate morphological characteristics.

[0024] In order to solve the above problems, the present invention provides a method for generating a digital image of aggregate.

[0025] The specific embodiments are described in detail below: A specific embodiment of the present invention discloses a method for generating a digital image of aggregates, combining Figure 1 Come and see, Figure 1 The schematic flow chart of an embodiment of the method for generating a digital image of aggregate provided by the present invention includes step S101 and step S102, wherein: In step S101, a real aggregate image set corresponding to the target aggregate is obtained, and the real aggregate image set is used as an input of an image generation model; In step S102, the output of the image generation model is determined as an aggregate digital image set corresponding to the target aggregate; The image generation model is trained based on the infoGAN-L model, and the infoGAN-L model is obtained by replacing the generator activation function of the infoGAN model with the LeakyReLU function.

[0026] During implementation, first, a real aggregate image set corresponding to the target aggregate can be obtained. For example, the real aggregate image set corresponding to the target aggregate can be collected through an aggregate image testing system (AIMS). The target aggregate can be limestone, basalt, diabase, granite, gneiss, sandstone, etc.

[0027] After obtaining the real aggregate image set corresponding to the target aggregate, the real aggregate image set can be used as the input of the image generation model, and the output of the image generation model is determined as the aggregate digital image set corresponding to the target aggregate for subsequent aggregate morphological characteristics research.

[0028] The image generation model used in the present invention is an improved Information Maximizing Generative Adversarial Nets (infoGAN), which can also be called infoGAN-L model. The improvement is that the generator activation function of the infoGAN model is changed from ReLU function to LeakyReLU function, thereby improving the learning ability of the network, avoiding gradient disappearance, and increasing the convergence speed. The infoGAN model itself can also improve the quality and diversity of generated images.

[0029] The aggregate digital image generation method provided by the present invention can be applied to the morphological characteristics research scenario of aggregates, and can also be applied to other image generation scenarios in small sample situations, and the present invention does not make specific limitations on this.

[0030] Compared with the prior art, the aggregate digital image generation method provided in this embodiment generates an aggregate digital image set corresponding to the target aggregate through the infoGAN-L model. While improving the quality and diversity of the generated images, since the generator activation function of the infoGAN model is modified from the ReLU function to the LeakyReLU function, the learning ability of the network is improved, the gradient disappearance is avoided, thereby increasing the convergence speed, and further improving the efficiency of image generation. The present invention improves the generation efficiency, quality and diversity of the aggregate digital image set corresponding to the target aggregate, thereby improving the reliability of subsequent research on aggregate morphological characteristics.

[0031] Exemplarily, the image generation model is trained based on a sample aggregate real image set, and the training goal of the image generation model is to ensure that the similarity between the aggregate digital image set corresponding to the sample aggregate real image set generated by the image generation model and the sample aggregate real image set is greater than a preset threshold.

[0032] Specifically, the image generation model can be trained based on the real image set of the sample collection. During training, the training goal is to generate a digital image set that is as similar as possible to the real image set of the sample collection.

[0033] Exemplarily, the batch size of the image generation model during training is 16, the learning rate is 0.0001, the number of training times is 2000, and the convolution kernel size is 5*5.

[0034] Specifically, when training the image generation model, the above training parameters may be referred to for model training to obtain the best training effect.

[0035] Exemplarily, the image generation effect of the image generation model is evaluated based on the similarity between the image distribution and two-dimensional morphological parameters of the first aggregate digital image set and the real aggregate image, as well as the diversity and edge detection results of the first aggregate digital image set. The first aggregate digital image set is the aggregate digital image set corresponding to the real image set of the sample aggregate generated by the image generation model.

[0036] Specifically, when evaluating the image generation effect of the image generation model, the image generation effect of the image generation model can be evaluated based on the image distribution and two-dimensional morphological parameters of the first aggregate digital image set (i.e., the aggregate digital image set corresponding to the real image set of the sample aggregate generated by the image generation model) and the similarity with the real aggregate image, as well as the diversity and edge detection results of the first aggregate digital image set.

[0037] Exemplarily, the two-dimensional morphological parameters include roundness and convex hull algorithm angularity.

[0038] Specifically, considering the evaluation efficiency and accuracy, the present invention uses the roundness and convex hull algorithm angularity in the two-dimensional morphological parameters to evaluate the shape of the aggregate. Roundness is an index used to evaluate the shape of the aggregate. The closer the roundness value is to 1, the closer the shape of the aggregate is to a circle. The more and more complex the edges and corners of the aggregate are, the greater the area difference before and after the convex hull is, and the greater the convex hull algorithm angularity value is.

[0039] Exemplarily, the similarity between the image distribution of the first aggregate digital image set and the real aggregate image is determined based on a t-SNE algorithm.

[0040] Specifically, when calculating the similarity between the image distribution of the first aggregate digital image set and the real aggregate image, the similarity between the two can be determined by the t-SNE algorithm.

[0041] Exemplarily, the diversity of the first aggregate digital image set is determined based on the Fréchet distance of the first aggregate digital image set.

[0042] Specifically, when calculating the diversity of the first aggregate digital image set, the diversity of the first aggregate digital image set may be determined by calculating the Fréchet distance between images in the first aggregate digital image set.

[0043] The technical solution of the present invention is better described below with reference to a specific embodiment: The present invention adopts infoGAN as the basic algorithm framework, and obtains the derivative model infoGAN-L on this basis to realize the refined generation of aggregate digital images, solve the problem that the generative adversarial network algorithm generates aggregate images poorly, provide guidance for the development of high-quality aggregate digital image generation methods, and provide support for the establishment of a high-precision asphalt mixture mesoscopic model.

[0044] The present invention uses the infoGAN-L model to generate refined aggregate images of two particle sizes, 4.75-9.5 mm and 9.5-13.2 mm. The results show that the infoGAN-L model can achieve efficient generation of digital images of aggregates, and the morphological features of the generated digital images of aggregates are very similar to those of real aggregate images, and the generated digital images of aggregates are of high quality and diversity.

[0045] The specific steps of the aggregate digital image generation method provided by the present invention are as follows: 1. Collection of digital image training sets.

[0046] The Aggregate Image Test System (AIMS) is a test system that integrates an aggregate image acquisition module and an analysis program. The aggregate image acquisition module is mainly composed of a high-magnification camera, an aggregate tray, a lighting system, and a microscope. This module can be used to obtain a real two-dimensional digital image of the aggregate; the analysis program is a data analysis system that performs statistical analysis of the morphological characteristics of the acquired images.

[0047] The present invention selects diabase as raw material. First, the diabase aggregate is subjected to a series of pretreatments such as screening, washing, and drying. Secondly, AIMS is used to obtain a two-dimensional image of the processed aggregate. Finally, the collected aggregate images are optimized to form a training set for the generation model. The number of aggregate images collected at each particle size is 500, totaling 1000.

[0048] 2. Selection of evaluation parameters for aggregate morphological characteristics.

[0049] With the development of computer technology, the evaluation index of aggregate morphological characteristics has gradually developed from one-dimensional index to two-dimensional index to three-dimensional evaluation index. One-dimensional evaluation index uses aggregate diameter, flatness and other characteristics to characterize morphological characteristics; two-dimensional evaluation index uses shape, area, roundness, convex hull algorithm angularity and other characteristics to characterize morphological characteristics; three-dimensional evaluation index uses aggregate three-dimensional sphericity value and other characteristics to characterize morphological characteristics. The present invention selects two-dimensional evaluation indexes with high efficiency and good accuracy to study aggregate morphological characteristics according to actual conditions.

[0050] After comparative analysis, the present invention selects roundness Round and convex hull algorithm angularity CHA as evaluation indicators of aggregate morphological characteristics.

[0051] 3. Build the infoGAN-L model.

[0052] (1) Generator and discriminator.

[0053] infoGAN is a variant of generative adversarial networks that achieves control and interpretability of generated images by combining noise vectors in the latent space with specific category information.

[0054] Combination Figure 2 Come and see, Figure 2 This is a schematic diagram of an embodiment of the network structure of the infoGAN model provided by the present invention. The infoGAN model consists of a generator G and a discriminator D, and an additional continuous variable c is added to the noise vector input to the generator G to control specific attributes of the image. The discriminator D is connected to the tail of the auxiliary network Q, which is essentially a classifier for implicit coding of samples. Q outputs the code c'. If the implicit code c input by the generator can have a clear impact on the generated sample, that is, the two are highly correlated. In order to improve the correlation between the implicit code c and the generated sample, so that a certain dimension of the implicit code can correspond to certain semantic information of the generated sample, infoGAN needs to maximize the mutual information between the two.

[0055] infoGAN introduces mutual information into the loss function and adds a regularization term to the loss function. , and its training objective function is:

[0056] in, Generate data for the continuous variables c and The mutual information term between is a hyperparameter.

[0057] In the infoGAN model, the goal of the generator G is to generate fake samples that can deceive the discriminator D, and its loss function is:

[0058] The goal of the discriminator D is to judge true and false samples as accurately as possible, and its loss function is:

[0059] For the joint network of the generator G and the auxiliary network Q, in order to make the implicit code c and the output There is a strong correlation, and the loss function is:

[0060] Combination Figure 3 Come and see, Figure 3 This is a schematic diagram of an embodiment of the infoGAN model generator and discriminator network structure provided by the present invention. The generator and discriminator in the present invention both perform 4 deconvolution operations, wherein each deconvolution layer of the generator is followed by a batch normalization (BN) layer and a ReLU layer. The discriminator is followed by a BN layer, a Dropout layer, and a LeakyReLU layer after each deconvolution layer. The BN layer is used to normalize samples between layers during training to avoid the gradient vanishing problem; the Dropout layer is a way to regularize the objective function, which can reduce the degree of overfitting of the model, and the Dropout layer takes a value of 0.5; ReLU and LeakyReLU are the activation functions of the generator and the discriminator, respectively.

[0061] (2) Activation function

[0062] Activation functions are the mathematical equations that determine the output of a neural network, and they must be computationally efficient because they are calculated across thousands or even millions of neurons in a deep neural network for each data sample.

[0063] In the original infoGAN model, the activation function of the generator uses the ReLU function. The gradient of the ReLU function can only take two values: 0 and 1. Its advantage is that it has high computational efficiency. However, it may also hinder the training process. Usually, the input value of the activation function has a bias term. Suppose the bias term becomes too small, so that the value of the input activation function is always negative. At this time, the neuron can no longer learn, which is called the neuron "death" problem. Moreover, the negative gradient is 0, which can easily cause information loss. The LeakyReLU function can solve this problem.

[0064] LeakyReLU is very similar to ReLU, with the difference being that the input is less than 0. For the part where the LeakyReLU input is less than 0, the value is negative and has a small gradient.

[0065] The benefit of using the LeakyReLU function is that during the back-propagation process, the gradient can also be calculated for the part of the LeakyReLU activation function input that is less than zero. This helps improve the learning ability of the network, avoid gradient vanishing, and increase the convergence speed.

[0066] Taking into account the advantages of the LeakyReLU function, the present invention changes the generator activation function of the infoGAN model from the ReLU function to the LeakyReLU function. The network structure is consistent with the original infoGAN model. The adjusted model is called the infoGAN-L model.

[0067] 4. Generate digital image of aggregate.

[0068] The present invention adopts the infoGAN-L model to take 500 processed aggregate images with a particle size of 4.75-9.5 mm and 500 processed aggregate images with a particle size of 9.5-13.2 mm as the input of the model, and trains the aggregate image generation model according to a batch size of 16, a learning rate of 0.0001, a training number of 2000 times, and a convolution kernel size of 5*5.

[0069] 5. Evaluation of aggregate digital image generation effect.

[0070] (1) Similarity between the distribution of aggregate digital images and the distribution of aggregate original images.

[0071] t-SNE is a popular learning method for nonlinear dimensionality reduction. It embeds high-dimensional data into low-dimensional (2D or 3D) space, displays the representation of generated images in low-dimensional space, and retains the overall structure of the data, making it easier to observe the distribution between images and thus evaluate image quality.

[0072] The present invention uses t-SNE to evaluate the quality of generated images and concludes that the distribution of aggregate images generated by the infoGAN-L model is similar to that of the original images.

[0073] (2) Diversity and quality of aggregate digital images.

[0074] The present invention introduces the FID index to evaluate the diversity and quality of aggregate digital images generated by the infoGAN-L model. When calculating the FID value, a set of samples are first extracted from the real data distribution and the generated image respectively, and then the feature vectors are extracted from these samples using the pre-trained Inception network. The FID value is obtained by calculating the mean and covariance matrix of the two distributions and calculating the Fréchet distance between them. The smaller the FID value, the closer the image generated by the model is to the real image.

[0075] For digital images of aggregates with a particle size of 4.75-9.5 mm, the FID value of the aggregate image generated by the infoGAN-L model is 150.167. For digital images of aggregates with a particle size of 9.5-13.2 mm, the FID value of the aggregate image generated by the infoGAN-L model is 63.308. In summary, the infoGAN-L model performs well in all particle size ranges.

[0076] (3) Edge contour detection of aggregate digital images.

[0077] One of the characteristics of a real aggregate image is that it has clear boundary contours, and these contours are closed. To this end, the present invention uses an edge contour detection method to calculate the percentage of the number of closed contours in the total number of generated aggregate images, so as to quantitatively evaluate the quality of digital aggregate images generated by different models. The Canny edge detection algorithm is used to extract the edge of the image, and then the contour of the generated image is detected based on the findContours function in OpenCV, and its perimeter and area are calculated to determine whether the contour is closed. If it is closed, the specific perimeter and area values ​​are returned, and if it is not closed, 0 is returned. The digital aggregate image generated by the infoGAN-L model has a contour closure rate of 99% for aggregates with a particle size of 4.75-9.5mm, and a contour closure rate of 100% for aggregates with a particle size of 9.5-13.2mm.

[0078] (4) Distribution probability of morphological parameters.

[0079] The contour features of the digital image of aggregate generated by the infoGAN-L model were extracted, and the roundness and convex hull algorithm angularity of the digital image generated by the infoGAN-L model were calculated. The average roundness of the digital image of aggregate with a particle size of 4.75-9.5 mm generated by the infoGAN-L model is 0.697, and the average roundness of the original image is 0.741. By comparison, the digital image of aggregate generated by the infoGAN-L model is close to the roundness of the original aggregate image; the convex hull algorithm angularity is 10.564, and the average convex hull algorithm angularity of the original image is 6.241. The digital image of aggregate generated by the infoGAN-L model is close to the convex hull algorithm angularity of the original image. Based on the results of morphological parameter distribution analysis, it is shown that the morphological characteristics of the digital image of aggregate generated by the infoGAN-L model are very similar to those of the original aggregate image.

[0080] The infoGAN-L model can realize the efficient generation of digital images of aggregates. The morphological characteristics of the digital images of aggregates generated by it are very similar to those of real aggregate images, and can be used to establish high-precision virtual asphalt mixtures. The generated digital images of aggregates are of high quality and diversity, which can lay the foundation for the establishment of the mesostructure of asphalt mixtures and the study of the morphological characteristics of asphalt mixtures.

[0081] The embodiment of the present invention also provides a device for generating a digital image of aggregates, Figure 4 Come and see, Figure 4 This is a schematic structural diagram of an embodiment of a device for generating a digital image of aggregate provided by the present invention. The device for generating a digital image of aggregate 400 comprises: An acquisition module 401 is used to acquire a real aggregate image set corresponding to a target aggregate, and use the real aggregate image set as an input of an image generation model; A determination module 402 is used to determine the output of the image generation model as an aggregate digital image set corresponding to the target aggregate; The image generation model is trained based on the infoGAN-L model, and the infoGAN-L model is obtained by replacing the generator activation function of the infoGAN model with the LeakyReLU function.

[0082] The specific implementation of each module of the aggregate digital image generation device can refer to the description of the above-mentioned aggregate digital image generation method, and has similar beneficial effects, which will not be repeated here.

[0083] The embodiment of the present invention further provides an electronic device, Figure 5 Come and see, Figure 5 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. The electronic device 500 includes a processor 501, a memory 502, and a computer program stored in the memory 502 and executable on the processor 501. When the processor 501 executes the program, the aggregate digital image generation method as described above is implemented.

[0084] As a preferred embodiment, the electronic device 500 further includes a display 503 for displaying the aggregate digital image generation method executed by the processor 501 as described above.

[0085] Exemplarily, the computer program may be divided into one or more modules / units, one or more modules / units are stored in the memory 502, and are executed by the processor 501 to complete the present invention. One or more modules / units may be a series of computer program instruction segments that can complete specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device 500. For example, the computer program may be divided into the acquisition module 401 and the determination module 402 in the above embodiment, and the specific functions of each module are as described above, and are not repeated here one by one.

[0086] The electronic device 500 may be a desktop computer, a notebook, a PDA or a smart phone with an adjustable camera module.

[0087] Among them, the processor 501 may be an integrated circuit chip with signal processing capabilities. The above-mentioned processor 501 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0088] The memory 502 may be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), etc. The memory 502 is used to store programs, and the processor 501 executes the program after receiving an execution instruction. The process definition method disclosed in any of the above embodiments of the present invention may be applied to the processor 501, or implemented by the processor 501.

[0089] The display 503 may be an LCD display screen or an LED display screen, for example, a display screen on a mobile phone.

[0090] Understandably, Figure 5 The structure shown is only a schematic diagram of the structure of the electronic device 500. The electronic device 500 may also include Figure 5 More or fewer components as shown. Figure 5 Each component shown in the figure can be implemented by hardware, software or a combination thereof.

[0091] The electronic device provided according to the above-mentioned embodiment of the present invention can be implemented with reference to the specific description of the aggregate digital image generation method as described above according to the present invention, and has similar beneficial effects as the aggregate digital image generation method as described above, which will not be repeated here.

[0092] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the method for generating a digital image of aggregate as described above is implemented.

[0093] Generally speaking, the computer instructions for implementing the method of the present invention can be carried by any combination of one or more computer-readable storage media. Non-transitory computer-readable storage media can include any computer-readable media except for the signal itself that is temporarily propagating.

[0094] Computer-readable storage media may be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more conductors, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium may be any tangible medium that contains or stores a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0095] Computer program codes for performing the operations of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages, in particular, Python suitable for neural network computing and platform frameworks based on TensorFlow, PyTorch, etc. may be used. The program code may be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0096] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0097] The present invention discloses a method, device, electronic device and storage medium for generating digital images of aggregates. An infoGAN-L model is used to generate a digital image set of aggregates corresponding to a target aggregate. While improving the quality and diversity of the generated images, the generator activation function of the infoGAN model is modified from a ReLU function to a LeakyReLU function, thereby improving the learning ability of the network, avoiding gradient vanishing, thereby increasing the convergence speed, and further improving the efficiency of image generation. The present invention improves the generation efficiency, quality and diversity of the digital image set of aggregates corresponding to the target aggregate, thereby improving the reliability of subsequent research on aggregate morphological characteristics.

[0098] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field 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 generating a digital image of aggregate, characterized in that: include: Acquire a real aggregate image set corresponding to the target aggregate, and use the real aggregate image set as input of the image generation model; Determining the output of the image generation model as an aggregate digital image set corresponding to the target aggregate; The image generation model is trained based on the infoGAN-L model, and the infoGAN-L model is obtained by replacing the generator activation function of the infoGAN model with the LeakyReLU function.

2. The method for generating a digital image of aggregate according to claim 1, characterized in that: The image generation model is trained based on a sample aggregate real image set, and the training goal of the image generation model is to ensure that the similarity between the aggregate digital image set corresponding to the sample aggregate real image set generated by the image generation model and the sample aggregate real image set is greater than a preset threshold.

3. The method for generating a digital image of aggregate according to claim 2, characterized in that: The batch size of the image generation model during training is 16, the learning rate is 0.0001, the number of training times is 2000, and the convolution kernel size is 5*5.

4. The method for generating a digital image of aggregate according to claim 3, characterized in that: The image generation effect of the image generation model is evaluated based on the similarity between the image distribution and two-dimensional morphological parameters of the first aggregate digital image set and the real aggregate image, as well as the diversity and edge detection results of the first aggregate digital image set. The first aggregate digital image set is the aggregate digital image set corresponding to the real image set of the sample aggregate generated by the image generation model.

5. The method for generating a digital image of aggregate according to claim 4, characterized in that: The two-dimensional morphological parameters include roundness and convex hull algorithm angularity.

6. The method for generating a digital image of aggregate according to claim 4, characterized in that: The similarity between the image distribution of the first aggregate digital image set and the real aggregate image is determined based on a t-SNE algorithm.

7. The method for generating a digital image of aggregate according to claim 4, characterized in that: The diversity of the first aggregate digital image set is determined based on a Fréchet distance of the first aggregate digital image set.

8. A device for generating digital images of aggregates, characterized in that: include: An acquisition module, used to acquire a set of real aggregate images corresponding to the target aggregate, and use the set of real aggregate images as input to the image generation model; A determination module, used for determining the output of the image generation model as an aggregate digital image set corresponding to the target aggregate; The image generation model is trained based on the infoGAN-L model, and the infoGAN-L model is obtained by replacing the generator activation function of the infoGAN model with the LeakyReLU function.

9. An electronic device, characterized in that: It comprises a memory and a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the method for generating a digital image of aggregate according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the method for generating a digital image of aggregate as described in any one of claims 1 to 7 is implemented.

Images