Flotation froth image generation method and device based on de-noising diffusion probability model

Abstract

The invention discloses a flotation froth image generation method and device based on a denoising diffusion probability model, a storage medium and computer equipment. The method comprises the steps that multiple flotation froth original images are acquired; cutting each flotation froth original image to a preset size to obtain flotation froth cutting images so as to form a flotation froth cutting image set; training a de-noising diffusion probability model by using the flotation froth clipping image set, so that the de-noising diffusion probability model iteratively optimizes model weight parameters in the process of sequentially performing noise adding and de-noising on the flotation froth clipping image based on a loss function to obtain a first weight parameter file; and inputting a randomly generated Gaussian noise image into the trained denoising diffusion probability model, and carrying out step-by-step denoising on the input Gaussian noise image based on the first weight parameter file by using the trained denoising diffusion probability model to generate a plurality of first flotation froth expansion images. And high-quality flotation froth images can be obtained with low cost.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a flotation foam image generation method and device based on a denoising diffusion probability model, a storage medium, and a computer device. Background Art

[0002] Flotation is one of the most important unit operations in mineral processing. It is an industrial mineral processing method that uses the differences in physical and chemical properties of the mineral particle surface to separate different minerals. Among them, the most widely used is the froth flotation method. The main process is that bubbles carry selective mineral particles up in the slurry, and then scrape off the foam formed on the surface of the slurry to achieve the purpose of selective separation of minerals.

[0003] In recent years, with the continuous development of image processing technology and computer technology, machine vision technology has been widely used in mineral froth flotation process. At present, machine vision technology has been widely used in tasks such as flotation process grade prediction, abnormal condition identification, flotation process operation status evaluation and abnormal condition tracing, and has achieved very good results. However, these tasks all require a large amount of image data to establish and train models, including flotation froth images, operating variables, flotation process status information and other data. Due to the high pixel count of flotation froth images and limited storage space of industrial computers, it is relatively difficult to collect and annotate data. It is difficult to save a large amount of offline froth image data at the flotation process site, so the demand for data sets is difficult to meet. Summary of the invention

[0004] In view of this, the embodiment of the present application provides a flotation foam image generation method and device, storage medium, and computer equipment based on a denoising diffusion probability model, which not only improves the richness of the flotation foam image, but also improves the generalization ability of the model through model training, so that it can generate high-quality and extended images that are suitable for different flotation states. In addition, the solution also reduces the cost of acquiring image data and realizes the ability to quickly generate images, providing strong support for applications such as flotation equipment operation status monitoring and fault warning.

[0005] According to one aspect of the present application, a flotation froth image generation method based on a denoising diffusion probability model is provided, the method comprising:

[0006] Acquire multiple original images of flotation foam;

[0007] Cutting each of the flotation foam original images to a preset size to obtain a flotation foam cut image to form a flotation foam cut image set;

[0008] Using the flotation foam cropped image set to train a denoising diffusion probability model, so that the denoising diffusion probability model iteratively optimizes model weight parameters in the process of sequentially denoising and denoising the flotation foam cropped images based on a loss function, to obtain a first weight parameter file;

[0009] A randomly generated Gaussian noise image is input into the trained denoising diffusion probability model, and the input Gaussian noise image is gradually denoised using the trained denoising diffusion probability model based on the first weight parameter file to generate a plurality of first flotation foam expansion images.

[0010] Optionally, after cropping each of the flotation foam original images to a preset size to obtain a flotation foam cropped image to form a flotation foam cropped image set, the method further comprises:

[0011] According to the bubble distribution characteristics corresponding to multiple image types, each of the flotation foam cropped images is cropped twice to obtain a preset number of flotation foam classification images corresponding to each image type, so as to form a flotation foam classification image set corresponding to each image type.

[0012] Optionally, after forming a flotation foam classification image set corresponding to each image type, the method further comprises:

[0013] For the flotation foam classification image set corresponding to each image type, a denoising diffusion probability model is trained using the flotation foam classification image set, so that the denoising diffusion probability model iteratively optimizes model weight parameters in the process of sequentially denoising and denoising the flotation foam classification images based on a loss function to obtain a second weight parameter file, a randomly generated Gaussian noise image is input into the trained denoising diffusion probability model, and the trained denoising diffusion probability model is used to gradually denoise the input Gaussian noise image based on the second weight parameter file to generate multiple second flotation foam expansion images corresponding to the image type.

[0014] Optionally, the image types include a single large bubble image, a large number of small bubble images, a mixed image of large and small bubbles, and a medium-sized bubble image.

[0015] Optionally, multiple flotation foam original images are obtained, including:

[0016] A camera is set above the flotation tank of the gold flotation equipment;

[0017] Under various flotation states of the gold ore flotation equipment, the camera collects images at a preset frequency as the flotation foam original images, wherein the flotation state includes a normal production state and an abnormal operating state.

[0018] Optionally, after generating the first flotation froth expansion image, the method further comprises:

[0019] constructing a first sample set based on the flotation foam cropped image and the first flotation foam expanded image, and marking the flotation status of each image in the first sample set;

[0020] training a flotation state recognition model based on the first sample set;

[0021] The flotation state recognition model is used to perform flotation state recognition on the flotation foam image collected on-site by the gold mine flotation equipment.

[0022] Optionally, after generating the second flotation froth expansion image, the method further comprises:

[0023] constructing a second sample set based on the flotation foam classification image and the second flotation foam expansion image, and marking the image type of each image in the second sample set;

[0024] Training an image category recognition model based on the second sample set;

[0025] Acquire continuous flotation foam images collected on-site by the gold mine flotation equipment, and divide the flotation foam images into multiple image regions, perform image type recognition on each image region respectively through the image type recognition model, and determine the regional image type corresponding to each image region;

[0026] According to the regional image types of each image area corresponding to a plurality of consecutive flotation foam images, the bubble movement characteristics corresponding to the gold mine flotation equipment are determined, and the flotation state is identified according to the bubble movement characteristics, wherein the bubble movement characteristics include bubble movement speed and / or bubble rupture speed.

[0027] According to another aspect of the present application, a flotation froth image generating device based on a denoising diffusion probability model is provided, the device comprising:

[0028] An image acquisition module, used for acquiring a plurality of original images of flotation foam;

[0029] An image processing module, used for cropping each of the flotation foam original images to a preset size to obtain a flotation foam cropped image to form a flotation foam cropped image set;

[0030] A model training module, for training a denoising diffusion probability model using the flotation foam cropped image set, so that the denoising diffusion probability model iteratively optimizes model weight parameters in the process of sequentially denoising and denoising the flotation foam cropped images based on a loss function, and obtains a first weight parameter file;

[0031] The image generation module is used to input a randomly generated Gaussian noise image into the trained denoising diffusion probability model, and use the trained denoising diffusion probability model to gradually denoise the input Gaussian noise image based on the first weight parameter file to generate multiple first flotation foam expansion images.

[0032] Optionally, the image processing module is further used to:

[0033] According to the bubble distribution characteristics corresponding to multiple image types, each of the flotation foam cropped images is cropped twice to obtain a preset number of flotation foam classification images corresponding to each image type, so as to form a flotation foam classification image set corresponding to each image type.

[0034] Optionally, the model training module is used to: train a denoising diffusion probability model using the flotation foam classification image set corresponding to each image type, so that the denoising diffusion probability model iteratively optimizes model weight parameters in the process of sequentially denoising and denoising the flotation foam classification images based on a loss function, thereby obtaining a second weight parameter file;

[0035] The image generation module is further used to: input a randomly generated Gaussian noise image into the trained denoising diffusion probability model, and use the trained denoising diffusion probability model to gradually denoise the input Gaussian noise image based on the second weight parameter file to generate a plurality of second flotation foam expansion images corresponding to the image type.

[0036] Optionally, the image types include a single large bubble image, a large number of small bubble images, a mixed image of large and small bubbles, and a medium-sized bubble image.

[0037] Optionally, the image acquisition module is further used to:

[0038] A camera is set above the flotation tank of the gold flotation equipment;

[0039] Under various flotation states of the gold ore flotation equipment, the camera collects images at a preset frequency as the flotation foam original images, wherein the flotation state includes a normal production state and an abnormal operating state.

[0040] Optionally, the device further includes: an identification module, configured to:

[0041] constructing a first sample set based on the flotation foam cropped image and the first flotation foam expanded image, and marking the flotation status of each image in the first sample set;

[0042] training a flotation state recognition model based on the first sample set;

[0043] The flotation state recognition model is used to perform flotation state recognition on the flotation foam image collected on-site by the gold mine flotation equipment.

[0044] Optionally, the device further includes: an identification module, further configured to:

[0045] constructing a second sample set based on the flotation foam classification image and the second flotation foam expansion image, and marking the image type of each image in the second sample set;

[0046] Training an image category recognition model based on the second sample set;

[0047] Acquire continuous flotation foam images collected on-site by the gold mine flotation equipment, and divide the flotation foam images into multiple image regions, perform image type recognition on each image region respectively through the image type recognition model, and determine the regional image type corresponding to each image region;

[0048] According to the regional image types of each image area corresponding to a plurality of consecutive flotation foam images, the bubble movement characteristics corresponding to the gold mine flotation equipment are determined, and the flotation state is identified according to the bubble movement characteristics, wherein the bubble movement characteristics include bubble movement speed and / or bubble rupture speed.

[0049] According to another aspect of the present application, a storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned flotation foam image generation method based on the denoising diffusion probability model is implemented.

[0050] According to another aspect of the present application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein when the processor executes the program, the above-mentioned flotation foam image generation method based on the denoising diffusion probability model is implemented.

[0051] By means of the above technical scheme, the embodiment of the present application provides a flotation foam image generation method and device based on a denoising diffusion probability model, a storage medium, and a computer device. First, a camera is set above the flotation tank of the gold mine flotation equipment to collect the original flotation foam images under various flotation states at a preset frequency. Subsequently, the collected original images are cropped to a uniform size to form a flotation foam cropped image set. Next, the denoising diffusion probability model is trained using this image set so that the model can learn the ability to recover the flotation foam image from the noise image. Finally, a randomly generated Gaussian noise image is input to the trained model, and the model generates multiple flotation foam expanded images through a step-by-step denoising process. The embodiment of the present application not only improves the richness of the flotation foam image, but also improves the generalization ability of the model through model training, so that it can generate high-quality expanded images that are adapted to different flotation states. In addition, the scheme also reduces the acquisition cost of image data and realizes the ability to quickly generate images, providing strong support for applications such as flotation equipment operation status monitoring and fault warning.

[0052] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0054] Figure 1 A schematic flow chart of a flotation foam image generation method based on a denoising diffusion probability model provided in an embodiment of the present application is shown;

[0055] Figure 2 A schematic diagram of an original image of flotation foam provided in an embodiment of the present application is shown;

[0056] Figure 3 A schematic diagram of a flotation foam cropping image provided by an embodiment of the present application is shown;

[0057] Figure 4 A schematic diagram of a flotation foam classification image provided by an embodiment of the present application is shown;

[0058] Figure 5 A structural schematic diagram of a flotation foam image generating device based on a denoising diffusion probability model provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0059] The present application will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0060] In this embodiment, a flotation foam image generation method based on a denoising diffusion probability model is provided. Figure 1 As shown, the method includes:

[0061] Step 101, obtaining a plurality of original images of flotation foam.

[0062] Optionally, step 101 includes: setting a camera above the flotation cell of the gold mine flotation equipment; in various flotation states of the gold mine flotation equipment, collecting images as the flotation foam original images by the camera at a preset frequency, wherein the flotation state includes a normal production state and an abnormal operating state.

[0063] In an embodiment of the present application, a camera can be installed above the flotation cell of the gold flotation equipment to collect images under different flotation conditions (including normal production and abnormal working conditions). Doing so can ensure that the collected data fully reflects various situations in actual production, and provide rich and real basic data for subsequent image processing and model training. This helps to cover images of various working conditions so that the model can learn and understand the changes in the flotation process more comprehensively, among which images of abnormal working conditions help the model identify and predict potential problems, and improve the system's response speed and accuracy to abnormal situations.

[0064] Step 102 , cropping each of the flotation foam original images to a preset size to obtain a flotation foam cropped image to form a flotation foam cropped image set.

[0065] In the embodiment of the present application, the original images are uniformly cropped to a predetermined preset size to form a standard flotation foam cropped image set. This ensures that all images input to the model have the same size and key area focus, reduces noise caused by differences in image size and composition, facilitates model learning, and standardization simplifies model processing complexity and speeds up training. The preset size of the flotation foam cropped image can be 512×512. Figure 2 The schematic diagram of the original image of flotation foam is shown in Figure 3 Shown is a schematic representation of a cropped image of a flotation foam.

[0066] Step 103, using the flotation foam cropped image set to train a denoising diffusion probability model, so that the denoising diffusion probability model iteratively optimizes model weight parameters in the process of sequentially denoising and denoising the flotation foam cropped images based on a loss function, and obtains a first weight parameter file.

[0067] In the embodiment of the present application, a denoising diffusion probability model is trained using a flotation foam cropped image set, and the model weight parameters are iteratively optimized during the denoising and denoising process to form a first weight parameter file containing the optimized model weight parameters. The model learns how to extract useful features from complex images by combating noise, and these features are essential for distinguishing different flotation states. The denoising process enhances the model's ability to ignore insignificant details in the image, thereby improving recognition accuracy in complex environments.

[0068] Specifically, the training of the denoising diffusion probability model can include the following processes: Forward diffusion process: At the beginning of training, the model simulates a forward diffusion process, that is, gradually adding noise to the flotation foam cropped image set until the image is completely converted into Gaussian noise. Learning denoising process: After the forward diffusion process, the model enters the learning stage. In this stage, the model needs to learn how to restore the original flotation foam image from the noisy image. For example, it is achieved through supervised learning, that is, the model will be provided with the original image and the corresponding noise image pair, and then the model parameters are adjusted through an optimization algorithm (such as gradient descent) to minimize the difference between the denoised output and the original image. Iterative optimization: By continuously iterating the above-mentioned forward diffusion and learning denoising processes, the model gradually optimizes its weight parameters until it can accurately restore the original image of the flotation foam from the noise image.

[0069] Step 104 , inputting a randomly generated Gaussian noise image into the trained denoising diffusion probability model, and using the trained denoising diffusion probability model to gradually denoise the input Gaussian noise image based on the first weight parameter file to generate a plurality of first flotation foam expansion images.

[0070] In the embodiment of the present application, further, after the model training is completed, a randomly generated Gaussian noise image can be input to it, so as to perform a reverse diffusion process: after the noise image is input, the model starts to perform a reverse diffusion process, that is, gradually remove the noise in the image until the image related to the flotation foam is restored. This process is the inverse process of the forward diffusion process. By using the weight parameters learned in the training stage, the model can gradually transform the noise image into an image with flotation foam characteristics. After the reverse diffusion process, the model finally generates a plurality of expanded images related to the flotation foam. These images can be used for further analysis and research, or as reference data for monitoring the operating status of the flotation equipment. Through the above process, the denoising diffusion probability model plays a key role in the flotation foam image generation scheme. It can not only recover high-quality flotation foam images from noise images, but also generate a large number of expanded images, providing rich data support for subsequent image analysis and research. By generating additional image samples, the diversity of the data set is increased, which helps to avoid overfitting and improve the generalization ability of the model. The generated images can be used to simulate different flotation conditions, provide a visual reference for process adjustment and optimization, and help improve the gold mine recovery rate and overall production efficiency.

[0071] In the embodiment of the present application, the DDPM model is trained using a flotation foam cropped image set, so that the model can learn the intrinsic characteristics and laws of the flotation foam image. This training method can improve the generalization ability of the model, so that the model can also generate high-quality extended images when processing unseen flotation foam images. By generating flotation foam extended images through the DDPM model, the size of the flotation foam image data set can be significantly increased without increasing the actual acquisition cost and complexity. These extended images can make up for the problem of insufficient data in the actual acquisition process and provide richer data support for subsequent image analysis and research. In addition, since the acquisition of the original image includes normal production status and abnormal operating conditions, the trained DDPM model can generate flotation foam images reflecting different flotation states. This ability helps to monitor the operating status of the flotation equipment, detect abnormalities in time and take corresponding treatment measures, so as to ensure the stability and safety of production. Compared with traditional image acquisition and processing methods, the flotation foam image generation scheme based on DDPM does not require a large amount of manpower and material resources, and can significantly reduce the cost of acquiring image data. At the same time, since the generated expanded images are of high quality and large quantity, they can replace the actually collected images to a certain extent, further reducing the cost of data collection and processing.

[0072] By applying the technical solution of this embodiment, a flotation foam image generation method based on a denoising diffusion probability model (DDPM) is provided. First, a camera is set above the flotation tank of the gold mine flotation equipment to collect the original flotation foam images under various flotation states at a preset frequency. Subsequently, the collected original images are cropped to a uniform size to form a flotation foam cropped image set. Next, the denoising diffusion probability model is trained using this image set so that the model can learn the ability to recover the flotation foam image from the noise image. Finally, a randomly generated Gaussian noise image is input to the trained model, and the model generates multiple flotation foam expanded images through a step-by-step denoising process. The embodiment of the present application not only improves the richness of the flotation foam image, but also improves the generalization ability of the model through model training, so that it can generate high-quality expanded images that are adapted to different flotation states. In addition, the scheme also reduces the acquisition cost of image data and realizes the ability to quickly generate images, providing strong support for applications such as flotation equipment operation status monitoring and fault warning.

[0073] In the embodiment of the present application, optionally, after step 104, the method further includes:

[0074] Step 105: construct a first sample set based on the flotation foam cropped image and the first flotation foam expanded image, and mark the flotation state of each image in the first sample set; train a flotation state recognition model based on the first sample set; and perform flotation state recognition on the flotation foam images collected on-site by the gold mine flotation equipment through the flotation state recognition model.

[0075] In this embodiment, after the first flotation foam expanded image is generated, the generated flotation foam expanded image can also be used to construct a sample set, and the flotation state recognition model can be trained based on the sample set, and then the flotation state recognition of the flotation foam image collected on-site at the gold mine flotation equipment can be performed using the model. Specifically, in step 104, the first flotation foam expanded image has been generated by a denoising diffusion probability model (DDPM). These expanded images, together with the original flotation foam cropped image, constitute the first sample set. Each image in the first sample set is marked with a flotation state, that is, the flotation state (such as normal, abnormal, etc.) corresponding to each image is clarified. The first sample set and its corresponding flotation state mark are further used to train the flotation state recognition model, for example, the model can be a classifier, such as a deep neural network (DNN), a convolutional neural network (CNN), etc., for identifying the flotation state of the flotation foam image. After the training is completed, the flotation state recognition model is deployed to the site of the gold mine flotation equipment. When a new flotation foam image is collected on site, the flotation state is identified through the model to determine whether the current flotation equipment is in a normal or abnormal state. The extended images generated by the denoising diffusion probability model in the embodiment of the present application greatly enrich the data set of flotation foam images, so that the trained flotation state recognition model can cover more flotation states and image changes. Since the sample set contains extended images generated by the model, these images may be slightly different from the actual collected images visually, which helps to enhance the generalization ability of the model and enable it to better adapt to the complex environment in the actual site. By real-time recognition of the flotation foam images collected on site by the flotation state recognition model, the abnormal state of the flotation equipment can be discovered in time, and timely warning information can be provided to the operator, thereby avoiding the occurrence of production accidents. Traditional manual monitoring of the flotation state requires a lot of manpower and time, and is prone to errors. Automatic monitoring through the flotation state recognition model can greatly reduce labor costs, improve the accuracy and efficiency of monitoring, thereby helping to maintain the continuity and stability of production and improve production efficiency.

[0076] In the embodiment of the present application, optionally, after step 102, the following may also be included:

[0077] Step 201, according to the bubble distribution characteristics corresponding to the multiple image types, each of the flotation foam cropped images is cropped twice to obtain a preset number of flotation foam classification images corresponding to each image type, so as to form a flotation foam classification image set corresponding to each image type. The image types include a single large bubble image, a large number of small bubble images, a large and small bubble mixed image, and a medium-sized bubble image.

[0078] In this embodiment, after obtaining the flotation foam cropped image, the flotation foam cropped image can be further cropped according to different image types (such as a single large bubble image, a large number of small bubble images, a mixed image of large and small bubbles, and an image of medium-sized bubbles). The purpose of the secondary cropping is to obtain a preset number of flotation foam classification images corresponding to each image type, thereby forming a flotation foam classification image set corresponding to each image type. Such a classification image set will be used for subsequent processing and training. Figure 4 As shown in the figure, there are four types of image schematic diagrams, among which Little_big represents a single large bubble image, which refers to an image mainly occupied by large bubbles, Many_small represents a large number of small bubble images, which refers to an image mainly occupied by small bubbles, Mixed represents a mixed image of large and small bubbles, which refers to an image where large and small bubbles occupy a similar surface area, and Some_normal represents a medium-sized bubble image, which refers to an image mainly occupied by medium-sized bubbles. In addition, the specific classification criteria for large bubbles, small bubbles, and medium-sized bubbles here can be defined manually, and secondary cropping can be completed by manual cropping. By performing secondary cropping and classification on the flotation foam images, it can be ensured that the images in each image set have similar features and quality, which helps to improve the accuracy and efficiency of subsequent image processing.

[0079] Further, optionally, step 201 may further include:

[0080] Step 202, for the flotation foam classification image set corresponding to each image type, use the flotation foam classification image set to train a denoising diffusion probability model, so that the denoising diffusion probability model iteratively optimizes the model weight parameters based on the loss function in the process of sequentially denoising and denoising the flotation foam classification images to obtain a second weight parameter file, input a randomly generated Gaussian noise image into the trained denoising diffusion probability model, and use the trained denoising diffusion probability model to gradually denoise the input Gaussian noise image based on the second weight parameter file to generate a plurality of second flotation foam expansion images corresponding to the image type.

[0081] Furthermore, for the flotation foam classification image set corresponding to each image type, these image sets can also be used to train a denoising diffusion probability model. During the training process, the model will perform iterative operations of adding noise and denoising on the flotation foam classification image based on the loss function to optimize the weight parameters of the model. Once the model training is completed, a randomly generated Gaussian noise image can be input into the model, and the input Gaussian noise image can be gradually denoised based on the weight parameters obtained by training (the second weight parameter file), thereby generating a plurality of second flotation foam expansion images corresponding to the original image type. The embodiment of the present application uses different types of flotation foam image sets to train the denoising diffusion probability model, which can enhance the generalization ability of the model so that it can handle more diverse input images, not just a specific type of image. By inputting a randomly generated Gaussian noise image into the trained denoising diffusion probability model and generating a corresponding flotation foam expansion image, the required image data is generated in a targeted manner, which helps to solve the problem of uneven distribution of data types in the flotation image data set, thereby effectively expanding the original data set.

[0082] Further, optionally, step 202 may further include:

[0083] Step 203, based on the flotation foam classification image and the second flotation foam expansion image, construct a second sample set, and mark the image type of each image in the second sample set; train an image type recognition model based on the second sample set; obtain continuous flotation foam images collected on-site by the gold mine flotation equipment, and divide the flotation foam images into multiple image areas, perform image type recognition on each image area respectively through the image type recognition model, and determine the regional image type corresponding to each image area; determine the bubble movement characteristics corresponding to the gold mine flotation equipment according to the regional image types of each image area corresponding to the multiple continuous flotation foam images, and perform flotation state recognition according to the bubble movement characteristics, wherein the bubble movement characteristics include bubble movement speed and / or bubble rupture speed.

[0084] In this embodiment, a new data set, namely, a second sample set, is constructed based on the second flotation foam expanded image (i.e., the artificially generated image) previously generated by the denoising diffusion probability model and the flotation foam classification image obtained by secondary cropping. This sample set contains not only the original classification image, but also the expanded image, which increases the diversity and scale of the data set. Each image in the second sample set is labeled with an image type, and the labeled type is consistent with the type defined in step 201, such as a single large bubble image, a large number of small bubble images, etc. An image type recognition model is trained using the labeled second sample set. This model will be able to identify which image type the input flotation foam image belongs to. Further, flotation foam images are continuously collected from the site of the gold mine flotation equipment, and these images are segmented into multiple image regions. This step is to analyze the state of the flotation foam in more detail. The trained image type recognition model is used to perform image type recognition on each segmented image region to determine the regional image type corresponding to each region. Thus, based on the regional image type of each image region in a plurality of consecutive flotation foam images, the motion characteristics of the bubble, such as the moving speed and / or bursting speed of the bubble, are analyzed. Then, the flotation state is identified according to these bubble movement characteristics to determine the flotation state of the gold flotation equipment. The embodiment of the present application generates a second flotation foam expansion image and constructs a second sample set, which not only increases the size of the data set, but also improves the diversity of the data set, which is crucial for training a more robust and accurate image type recognition model. By dividing the flotation foam image into multiple image regions and performing image type recognition on each region, fine-grained analysis of the flotation foam image can be achieved, and the local characteristics of the flotation foam can be understood more accurately. Based on the movement characteristics of bubbles in multiple consecutive flotation foam images, the flotation state is identified, and the operating status and flotation effect of the gold flotation equipment can be monitored in real time, potential problems can be discovered and solved in time, and flotation efficiency can be improved. Further, by automatically analyzing and identifying the flotation foam image, a large amount of flotation process data can be collected. These data can be used to optimize flotation process parameters, improve the recovery rate and product quality of gold flotation, and compared with the traditional method of manual observation and analysis of flotation foam, the use of image processing and machine learning technology can greatly reduce manual intervention and improve the accuracy and efficiency of analysis.

[0085] It should be noted that, in the embodiment of the present application, the flotation state identification results of step 105 and step 203 can also be combined to perform a comprehensive determination of the flotation state of the gold ore flotation equipment.

[0086] By applying the technical solution of the embodiment of the present application, the flotation foam expanded image is generated by the denoising diffusion probability model (DDPM), which greatly enriches the data set of flotation foam images and improves the diversity and scale of the data set. The model is trained by using a variety of flotation foam image sets, which enhances the generalization ability of the model, enables it to process more diverse flotation foam images, and improves the recognition accuracy of the model in complex scenes. By performing secondary cropping and classification on the flotation foam images, a more refined flotation foam classification image set is constructed, which helps to more accurately identify the image types of flotation foam. The sample set is constructed by using the flotation foam classification image and the expanded image to train the flotation state recognition model and the image type recognition model, which improves the accuracy of flotation state recognition and can more accurately judge the operating state and flotation effect of the gold mine flotation equipment. In addition, the real-time monitoring of the flotation foam images collected on-site of the gold mine flotation equipment is realized, and the abnormal state of the flotation equipment can be discovered in time, and timely warning information can be provided. By analyzing the motion characteristics of bubbles in multiple consecutive flotation foam images, the flotation state can be more accurately identified, further enhancing the accuracy and reliability of the warning. Furthermore, through the automated analysis and recognition of flotation foam images, a large amount of flotation process data can be collected, which can be used to optimize flotation process parameters, improve the recovery rate and product quality of gold flotation, reduce dependence on manual observation and analysis, improve the automation level of the flotation process, reduce labor costs, and improve production efficiency. In addition, this solution is not only applicable to gold flotation equipment, but can also be extended to other similar flotation processes. It has high adaptability and flexibility. By training different image type recognition models and flotation state recognition models, it can adapt to the flotation foam image recognition needs in different scenarios.

[0087] In summary, the entire solution improves the accuracy and reliability of flotation foam image recognition, optimizes the flotation process, improves production efficiency, and reduces labor costs through steps such as data expansion, model training, real-time monitoring and early warning. At the same time, the solution also has high adaptability and flexibility and can be applied to similar flotation processes.

[0088] Further, as Figure 1 The specific implementation of the method, the embodiment of the present application provides a flotation foam image generation device based on a denoising diffusion probability model, such as Figure 5 As shown, the device comprises:

[0089] An image acquisition module, used for acquiring a plurality of original images of flotation foam;

[0090] An image processing module, used for cropping each of the flotation foam original images to a preset size to obtain a flotation foam cropped image to form a flotation foam cropped image set;

[0091] A model training module, for training a denoising diffusion probability model using the flotation foam cropped image set, so that the denoising diffusion probability model iteratively optimizes model weight parameters in the process of sequentially denoising and denoising the flotation foam cropped images based on a loss function, and obtains a first weight parameter file;

[0092] The image generation module is used to input a randomly generated Gaussian noise image into the trained denoising diffusion probability model, and use the trained denoising diffusion probability model to gradually denoise the input Gaussian noise image based on the first weight parameter file to generate multiple first flotation foam expansion images.

[0093] Optionally, the image processing module is further used to:

[0094] According to the bubble distribution characteristics corresponding to multiple image types, each of the flotation foam cropped images is cropped twice to obtain a preset number of flotation foam classification images corresponding to each image type, so as to form a flotation foam classification image set corresponding to each image type.

[0095] Optionally, the model training module is used to: train a denoising diffusion probability model using the flotation foam classification image set corresponding to each image type, so that the denoising diffusion probability model iteratively optimizes model weight parameters in the process of sequentially denoising and denoising the flotation foam classification images based on a loss function, thereby obtaining a second weight parameter file;

[0096] The image generation module is further used to: input a randomly generated Gaussian noise image into the trained denoising diffusion probability model, and use the trained denoising diffusion probability model to gradually denoise the input Gaussian noise image based on the second weight parameter file to generate a plurality of second flotation foam expansion images corresponding to the image type.

[0097] Optionally, the image types include a single large bubble image, a large number of small bubble images, a mixed image of large and small bubbles, and a medium-sized bubble image.

[0098] Optionally, the image acquisition module is further used to:

[0099] A camera is set above the flotation tank of the gold flotation equipment;

[0100] Under various flotation states of the gold ore flotation equipment, the camera collects images at a preset frequency as the flotation foam original images, wherein the flotation state includes a normal production state and an abnormal operating state.

[0101] Optionally, the device further includes: an identification module, configured to:

[0102] constructing a first sample set based on the flotation foam cropped image and the first flotation foam expanded image, and marking the flotation status of each image in the first sample set;

[0103] training a flotation state recognition model based on the first sample set;

[0104] The flotation state recognition model is used to perform flotation state recognition on the flotation foam image collected on-site by the gold mine flotation equipment.

[0105] Optionally, the device further includes: an identification module, further configured to:

[0106] constructing a second sample set based on the flotation foam classification image and the second flotation foam expansion image, and marking the image type of each image in the second sample set;

[0107] Training an image category recognition model based on the second sample set;

[0108] Acquire continuous flotation foam images collected on-site by the gold mine flotation equipment, and divide the flotation foam images into multiple image regions, perform image type recognition on each image region respectively through the image type recognition model, and determine the regional image type corresponding to each image region;

[0109] According to the regional image types of each image area corresponding to a plurality of consecutive flotation foam images, the bubble movement characteristics corresponding to the gold mine flotation equipment are determined, and the flotation state is identified according to the bubble movement characteristics, wherein the bubble movement characteristics include bubble movement speed and / or bubble rupture speed.

[0110] It should be noted that for other corresponding descriptions of the functional units involved in the flotation foam image generation device based on the denoising diffusion probability model provided in the embodiment of the present application, reference can be made to Figures 1 to 4 The corresponding description in the method will not be repeated here.

[0111] The embodiment of the present application also provides a computer device, which can be a personal computer, a server, a network device, etc. The computer device includes a bus, a processor, a memory and a communication interface, and can also include an input and output interface and a display device. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store location information. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the steps in each method embodiment are implemented.

[0112] Those skilled in the art will appreciate that the structure of the above-mentioned computer device is only a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components, or combine certain components, or have a different arrangement of components.

[0113] In one embodiment, a computer-readable storage medium is provided. The computer-readable storage medium may be non-volatile or volatile, and stores a computer program thereon. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0114] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0115] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0116] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0117] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0118] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A flotation froth image generation method based on a denoising diffusion probability model, characterized in that: The method comprises: Acquire multiple original images of flotation foam; Cutting each of the flotation foam original images to a preset size to obtain a flotation foam cut image to form a flotation foam cut image set; Using the flotation foam cropped image set to train a denoising diffusion probability model, so that the denoising diffusion probability model iteratively optimizes model weight parameters in the process of sequentially denoising and denoising the flotation foam cropped images based on a loss function, to obtain a first weight parameter file; A randomly generated Gaussian noise image is input into the trained denoising diffusion probability model, and the input Gaussian noise image is gradually denoised using the trained denoising diffusion probability model based on the first weight parameter file to generate a plurality of first flotation foam expansion images.

2. The method according to claim 1, characterized in that After cropping each of the flotation foam original images to a preset size to obtain a flotation foam cropped image to form a flotation foam cropped image set, the method further includes: According to the bubble distribution characteristics corresponding to multiple image types, each of the flotation foam cropped images is cropped twice to obtain a preset number of flotation foam classification images corresponding to each image type, so as to form a flotation foam classification image set corresponding to each image type.

3. The method according to claim 2, characterized in that After forming a flotation foam classification image set corresponding to each image type, the method further includes: For the flotation foam classification image set corresponding to each image type, a denoising diffusion probability model is trained using the flotation foam classification image set, so that the denoising diffusion probability model iteratively optimizes model weight parameters in the process of sequentially denoising and denoising the flotation foam classification images based on a loss function to obtain a second weight parameter file, a randomly generated Gaussian noise image is input into the trained denoising diffusion probability model, and the trained denoising diffusion probability model is used to gradually denoise the input Gaussian noise image based on the second weight parameter file to generate multiple second flotation foam expansion images corresponding to the image type.

4. The method according to claim 2, characterized in that: The image types include a single large bubble image, a large number of small bubble images, a mixed image of large and small bubbles, and a medium-sized bubble image.

5. The method according to claim 3 or 4, characterized in that: Get multiple flotation foam raw images, including: A camera is set above the flotation tank of the gold flotation equipment; Under various flotation states of the gold ore flotation equipment, the camera collects images at a preset frequency as the flotation foam original images, wherein the flotation state includes a normal production state and an abnormal operating state.

6. The method according to claim 5, characterized in that After generating the first flotation froth expansion image, the method further comprises: constructing a first sample set based on the flotation foam cropped image and the first flotation foam expanded image, and marking the flotation status of each image in the first sample set; training a flotation state recognition model based on the first sample set; The flotation state recognition model is used to perform flotation state recognition on the flotation foam image collected on-site by the gold mine flotation equipment.

7. The method according to claim 5, characterized in that After generating the second flotation froth expansion image, the method further comprises: constructing a second sample set based on the flotation foam classification image and the second flotation foam expansion image, and marking the image type of each image in the second sample set; Training an image category recognition model based on the second sample set; Acquire continuous flotation foam images collected on-site by the gold mine flotation equipment, and divide the flotation foam images into multiple image regions, perform image type recognition on each image region respectively through the image type recognition model, and determine the regional image type corresponding to each image region; According to the regional image types of each image area corresponding to a plurality of consecutive flotation foam images, the bubble movement characteristics corresponding to the gold mine flotation equipment are determined, and the flotation state is identified according to the bubble movement characteristics, wherein the bubble movement characteristics include bubble movement speed and / or bubble rupture speed.

8. A flotation froth image generation device based on a denoising diffusion probability model, characterized in that: The device comprises: An image acquisition module, used for acquiring a plurality of original images of flotation foam; An image processing module, used for cropping each of the flotation foam original images to a preset size to obtain a flotation foam cropped image to form a flotation foam cropped image set; A model training module, used for training a denoising diffusion probability model using the flotation foam cropped image set, so that the denoising diffusion probability model iteratively optimizes model weight parameters in the process of sequentially denoising and denoising the flotation foam cropped images based on a loss function, and obtains a first weight parameter file; The image generation module is used to input a randomly generated Gaussian noise image into the trained denoising diffusion probability model, and use the trained denoising diffusion probability model to gradually denoise the input Gaussian noise image based on the first weight parameter file to generate multiple first flotation foam expansion images.

9. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

10. A computer device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

Images