Rock burst migration identification method based on Weibull hybrid model

Abstract

The invention provides a rock burst migration identification method based on a Weibull hybrid model, and relates to the technical field of coal mine safety, and the method comprises the steps: obtaining historical micro-seismic monitoring data of any reference coal mine, and constructing a corresponding reference anomaly identification model which is used for determining the anomaly probability of any reference coal mine; target micro-seismic data of the target coal mine are obtained, the target micro-seismic data are mapped to any reference anomaly identification model, a target data set corresponding to the target micro-seismic data is obtained, and the target data set comprises multiple target samples corresponding to the target micro-seismic data; and obtaining an abnormal probability of the target sample in any reference abnormal identification model, and determining whether the target micro-seismic data is abnormal based on the abnormal probability. The problem of constructing a coal mine micro-seismic anomaly identification model under the conditions of insufficient micro-seismic data and unmarked data is solved, and the micro-seismic anomaly state identification capability is improved.

Description

Technical Field

[0001] The present application relates to the field of coal mine safety technology, and in particular to a method for identifying rock burst migration based on a Weibull mixed model. Background Art

[0002] Abnormal microseismic events may be precursor information for the occurrence of geological disasters such as rock burst. At present, the research on the identification of abnormal microseismic events mainly focuses on two aspects: (1) Analyzing the changes and trends of microseismic energy based on the mechanical properties of coal and rock masses to identify large-energy or abnormal microseismic events. (2) Using machine learning methods, such as linear regression and discriminant models, to establish a microseismic anomaly identification model.

[0003] Machine learning-based methods often require a certain amount of microseismic historical data, especially abnormal data, to establish a supervised microseismic safety status identification model. However, in some scenarios, such as the initial mining face, microseismic historical data is scarce, and there is no abnormal data, making it difficult to build a microseismic anomaly identification model in the early stages of mining, which affects the accuracy of microseismic anomaly identification. Summary of the invention

[0004] The present application aims to solve one of the technical problems in the related art at least to some extent.

[0005] To this end, the purpose of this application is to propose a rock burst migration identification method based on the Weibull mixed model to achieve the construction of a microseismic anomaly identification model under unlabeled small samples.

[0006] To achieve the above objectives, the present application proposes a method for identifying rock burst migration based on a Weibull mixed model, comprising:

[0007] Acquire historical microseismic monitoring data of any reference coal mine, and construct a corresponding reference anomaly identification model, wherein the reference anomaly identification model is used to determine the anomaly probability of any reference coal mine;

[0008] Acquire target microseismic data of a target coal mine, and map the target microseismic data to any reference anomaly identification model to obtain a target data set corresponding to the target microseismic data, wherein the target data set includes a plurality of target samples corresponding to the target microseismic data;

[0009] The abnormal probability of the target sample in any of the reference abnormality identification models is obtained, and whether the target microseismic data is abnormal is determined based on the abnormal probability.

[0010] The rock burst migration identification method based on Weibull mixed model provided in the present application aims to solve the current problem of abnormal identification under the lack of microseismic data in coal mines and unlabeled data. By acquiring the historical microseismic monitoring data of any reference coal mine, a corresponding reference abnormal identification model is constructed, wherein the reference abnormal identification model is used to determine the abnormal probability of any reference coal mine; the target microseismic data of the target coal mine is acquired, and the target microseismic data is mapped to any reference abnormal identification model to obtain a target data set corresponding to the target microseismic data, wherein the target data set includes target samples corresponding to multiple target microseismic data; the abnormal probability of the target sample in any reference abnormal identification model is acquired, and whether the target microseismic data is abnormal is determined based on the abnormal probability. By mapping the target microseismic data to the reference abnormal identification model of the reference coal mine, the difference between the scene where the target microseismic data is located and the scene where the reference coal mine is located is reduced, so that the reference abnormal identification model is migrated to the new microseismic identification scene, and the abnormal situation of the target coal mine is determined, which solves the problem of constructing a microseismic abnormal identification model for coal mines under the lack of microseismic data and unlabeled data, and improves the recognition ability of abnormal microseismic states.

[0011] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0013] Figure 1 A schematic diagram of a flow chart of a rock burst migration identification method based on a Weibull mixed model provided in an embodiment of the present application;

[0014] Figure 2 A flow chart of another rock burst migration identification method based on the Weibull mixed model provided in an embodiment of the present application. DETAILED DESCRIPTION

[0015] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0016] The following describes a method for identifying rock burst migration based on a Weibull mixed model according to an embodiment of the present application with reference to the accompanying drawings.

[0017] Figure 1A flow chart of a method for identifying rock burst migration based on a Weibull mixed model provided in an embodiment of the present application.

[0018] Coal mine microseismic data usually refers to the signals generated by geological activities such as geological deformation, rock movement or geological stress release caused by mining activities such as excavation. Abnormal microseismic events may be precursor information for geological disasters such as rock burst. Analyzing microseismic data, especially detecting abnormal microseismic events, is of great significance for preventing coal mine accidents and ensuring the safety of miners.

[0019] The method of establishing a microseismic anomaly identification model based on machine learning is a commonly used microseismic anomaly identification method. It often requires a certain amount of microseismic historical data, especially abnormal data, to establish a supervised microseismic safety status identification model. However, in some scenarios, such as the initial mining face, microseismic historical data is scarce, and there is no abnormal data, making it difficult to build a microseismic anomaly identification model in the early stages of mining. The diversity and complexity of coal mine geological structure, coal seam occurrence, rock layer structure, etc. often cause differences in the distribution of microseismic data. If it is not processed, directly migrating the microseismic data or anomaly identification model of other coal mines or working faces to the current scenario will cause the degradation of model performance due to the difference in data distribution, affecting the accuracy of microseismic anomaly identification, and even causing certain safety hazards.

[0020] To address this problem, the present invention provides a method for identifying rock burst migration based on a Weibull mixed model to achieve the construction of a microseismic anomaly identification model under unlabeled small samples. Figure 1 As shown, the rock burst migration identification method based on the Weibull mixed model includes the following steps:

[0021] Step 101, obtaining historical microseismic monitoring data of any reference coal mine, and constructing a corresponding reference anomaly identification model, wherein the reference anomaly identification model is used to determine the anomaly probability of any reference coal mine.

[0022] Optionally, the reference coal mine corresponds to annotated historical microseismic monitoring data, and a reference identification model is constructed based on the historical microseismic monitoring data and the corresponding labels.

[0023] The reference identification model may be a decision tree model, and the labels of the historical microseismic monitoring data are used to mark whether the historical microseismic monitoring data is normal or abnormal.

[0024] As an example, by inputting historical microseismic monitoring data into a reference anomaly identification model, the corresponding anomaly probability and anomaly identification result can be output.

[0025] Step 102, obtaining target microseismic data of the target coal mine, and mapping the target microseismic data to any reference anomaly recognition model to obtain a target data set corresponding to the target microseismic data, wherein the target data set includes a plurality of target samples corresponding to the target microseismic data.

[0026] Optionally, the target coal mine is a coal mine that lacks historical data and needs to identify abnormal microseismic conditions, and the corresponding target microseismic data is collected in real time.

[0027] It is understandable that the target coal mine lacks historical data, and accordingly, there is no corresponding anomaly identification model to perform anomaly identification on the collected target data. It is necessary to use the existing anomaly identification models of multiple reference coal mines to perform corresponding anomaly identification.

[0028] Optionally, the reference identification model reflects the data distribution of historical microseismic monitoring data. By mapping the target microseismic data to any reference identification model, the data distribution difference between the target microseismic data and the historical microseismic monitoring data can be reduced, and the target sample of the target microseismic data under the reference anomaly identification model can be obtained. The target sample reflects the data distribution of the target microseismic data under the reference anomaly identification model.

[0029] In the embodiment of the present application, target samples corresponding to a plurality of target microseismic data constitute a target data set.

[0030] Furthermore, the target microseismic data are respectively input into a plurality of reference identification models to obtain a plurality of target data sets.

[0031] It is understandable that the diversity and complexity of the geological structure, coal seam occurrence, rock structure, etc. of different reference coal mines lead to differences in the distribution of microseismic data corresponding to different reference coal mines. The same target microseismic data is mapped to different reference identification models, and the target samples obtained may be different.

[0032] Step 103: Obtain the abnormal probability of the target sample in any reference abnormality identification model, and determine whether the target microseismic data is abnormal based on the abnormal probability.

[0033] Optionally, by inputting the target sample into any reference anomaly detection model and outputting the anomaly probability corresponding to the target sample, the anomaly probability corresponding to the target sample when the data distribution conforms to the reference anomaly detection model is obtained.

[0034] By inputting the target samples into different reference anomaly identification models respectively, the anomaly probabilities under multiple different microseismic data distribution conditions are obtained.

[0035] Furthermore, an integrated decision is made based on the various abnormal probabilities to obtain the final abnormal probability corresponding to the target sample. When the final abnormal probability is large, it means that the target microseismic data is abnormal. When the final abnormal probability is small, it means that the target microseismic data is normal.

[0036] As an example, an abnormality threshold is set, and when the final abnormality probability is greater than the abnormality threshold, the target microseismic data is determined to be abnormal; otherwise, the target microseismic data is determined to be normal.

[0037] In this embodiment, by acquiring the historical microseismic monitoring data of any reference coal mine, a corresponding reference anomaly identification model is constructed, wherein the reference anomaly identification model is used to determine the anomaly probability of any reference coal mine; the target microseismic data of the target coal mine is acquired, and the target microseismic data is mapped to any reference anomaly identification model to obtain a target data set corresponding to the target microseismic data, wherein the target data set includes a plurality of target samples corresponding to the target microseismic data; the anomaly probability of the target sample in any reference anomaly identification model is acquired, and whether the target microseismic data is abnormal is determined based on the anomaly probability. By mapping the target microseismic data to the reference anomaly identification model of the reference coal mine, the difference between the scene where the target microseismic data is located and the scene where the reference coal mine is located is reduced, so that the reference anomaly identification model is migrated to the new microseismic identification scene, and the abnormal situation of the target coal mine is determined, which solves the problem of constructing a microseismic anomaly identification model for coal mines in the absence of microseismic data and unlabeled data, and improves the ability to identify abnormal microseismic states.

[0038] This embodiment provides another rock burst migration identification method based on the Weibull mixed model. Figure 2 A flow chart of another rock burst migration identification method based on the Weibull mixed model provided in an embodiment of the present application.

[0039] like Figure 2 As shown, the rock burst migration identification method based on the Weibull mixed model may include the following steps:

[0040] Step 201, obtaining historical microseismic monitoring data of any reference coal mine, and constructing a corresponding reference anomaly identification model, wherein the reference anomaly identification model is used to determine the anomaly probability of any reference coal mine.

[0041] Optionally, for any reference coal mine, the corresponding reference anomaly identification model can be obtained through the following steps:

[0042] Obtain historical microseismic data and label data of any reference coal mine, wherein the label data is used to indicate whether the historical microseismic data is abnormal; input the historical microseismic data into the decision tree model to be trained, and train the decision tree model based on the difference between the output of the decision tree model and the label data; and use the trained decision tree model as a reference anomaly identification model.

[0043] Optionally, the label data may be represented by numbers, for example, normal data in historical microseismic monitoring data is recorded as -1, and abnormal data is recorded as 1.

[0044] In the embodiment of the present application, the decision tree model uses the XGBoost (eXtreme Gradient Boosting) algorithm, which is an integrated machine learning algorithm that combines multiple weak learners (usually decision trees) to form a strong learner. Xgboost uses a gradient optimization algorithm to find the best split point in the tree construction process, thereby achieving fast and accurate convergence.

[0045] As an example, we will refer to the historical microseismic monitoring data of coal mine 1 (D S1 ) and label data XGBoost is introduced to learn the mapping relationship between microseismic features and microseismic states. The custom loss function is iterated through the difference between the label data and the XGBoost output until the loss function converges. The trained decision tree model is obtained as a reference anomaly identification model. The function express.

[0046] Similarly, for the historical microseismic monitoring data of other reference coal mines 2 (D S2 ), …, refer to the historical microseismic monitoring data of coal mine N (D SN ), the obtained reference anomaly recognition model uses the function express.

[0047] XGBoost uses a gradient boosting algorithm, where each new decision tree attempts to correct the errors of the previous model, thereby gradually improving the ability to identify abnormal microseismic states.

[0048] Step 202, obtaining target microseismic data of the target coal mine and constructing a cumulative distribution function of the target microseismic data; based on the inverse function of the cumulative distribution function, mapping the target microseismic data to any reference anomaly recognition model to obtain a target data set corresponding to the target microseismic data.

[0049] Optionally, the acquired target microseismic data is recorded as The sample number of the target microseismic data is

[0050] The target data set includes target samples corresponding to multiple target microseismic data.

[0051] Optionally, obtain the probability density function of the target microseismic data.

[0052] In the embodiment of the present application, the Weilblull mixed model is used to obtain the probability density function of the target microseismic data:

[0053]

[0054] in, is the probability density function of the target microseismic data x, x>0. m is the number of Weibull distributions that make up the mixed model, α i is the scale parameter of the ith Weibull component, β i is the shape parameter of the ith Weibull component, π i is the weight vector of different components in the mixture model, and π i ∈[0,1],

[0055] Different modes of microseismic data are represented by different Weibull distributions of the mixed model, providing an unsupervised method for pattern recognition and clustering of microseismic data.

[0056] Furthermore, the parameters corresponding to the probability density function are optimized by maximizing the likelihood function.

[0057] In the embodiment of the present application, the likelihood function L is maximized by the EM (Expectation-Maximization) algorithm to optimize π i , α i , β i , i=1,…,m.

[0058] The likelihood function can be expressed by the following formula:

[0059]

[0060] in, for The jth sample of .

[0061] By maximizing the likelihood function for the parameter π corresponding to the probability density function i , α i , β i Optimize and obtain the optimized probability density function.

[0062] Furthermore, based on the optimized probability density function, a cumulative distribution function is obtained.

[0063] The Cumulative Distribution Function (CDF) is used to describe the probability that a random variable takes a value less than or equal to a given value, which can be expressed by the following formula:

[0064]

[0065] Optionally, based on the inverse function of the cumulative distribution function, the target microseismic data is mapped to any reference anomaly identification model.

[0066] The percent point function (PPF), also known as the inverse cumulative distribution function (ICDF), is the inverse function of the cumulative distribution function (CDF). The percent point function of historical microseismic monitoring data can be used to calculate the target microseismic data. Map it to the feature space where the historical microseismic monitoring data is located to reduce the difference in data distribution between it and the reference coal mine.

[0067] Using the density accumulation function of the target domain and the percentile function of the source domain Adjust the distribution of microseismic data in the target domain to reduce the distribution difference between it and the source domain. First, calculate the CDF value of the target microseismic data.

[0068]

[0069] Secondly, it is mapped to any reference anomaly identification model to make it conform to the distribution of historical microseismic monitoring data corresponding to any reference coal mine.

[0070]

[0071] …

[0072]

[0073] in, is the jth target microseismic data mapped to the historical microseismic monitoring data of reference coal mine 1 (D S1 ) target sample; is the jth target microseismic data mapped to the historical microseismic monitoring data of reference coal mine 2 (D S2 ) target sample; is the mapping of the jth target microseismic data to the historical microseismic monitoring data of the reference coal mine N (D SN ) target sample.

[0074] Finally, we obtain a target dataset consisting of multiple target samples, denoted as

[0075]

[0076] By reducing the distribution difference between the target microseismic data and the historical microseismic monitoring data, the supervised classification model pre-trained with the historical microseismic monitoring data can be used for anomaly detection of the target microseismic data.

[0077] Step 203: Obtain the abnormal probability of the target sample in any reference abnormality identification model, and determine whether the target microseismic data is abnormal based on the abnormal probability.

[0078] Optionally, the target sample is input into any reference anomaly identification model and the anomaly probability is output.

[0079] Input the target samples included in each target data set into any reference anomaly recognition model and output the corresponding anomaly probability:

[0080]

[0081] …

[0082]

[0083] in, represents the abnormal probability set of the target sample evaluated by the reference abnormality identification model corresponding to reference coal mine 1, Indicates the 1st, 2nd, ..., target samples (i.e. ) is abnormal. Similarly, Represents the abnormal probability set of the target samples evaluated by the reference anomaly identification model corresponding to reference coal mines 2,…,N.

[0084] Furthermore, the abnormality degree corresponding to the target microseismic data is determined based on each abnormality probability; when the abnormality degree is greater than a preset abnormality threshold, the target microseismic data is determined to be abnormal.

[0085] As an example, the average value of each abnormal probability is calculated as the abnormality degree, which can be expressed by the following formula:

[0086]

[0087] in, Indicates the abnormality of each target microseismic data in the target domain.

[0088] Furthermore, the abnormal threshold is preset to 0. For the jth target microseismic data, if Confirm that it is abnormal, if Confirm that it is normal.

[0089] In this embodiment, by acquiring the historical microseismic monitoring data of any reference coal mine, a corresponding reference anomaly identification model is constructed, wherein the reference anomaly identification model is used to determine the anomaly probability of any reference coal mine; the target microseismic data of the target coal mine is acquired, and the cumulative distribution function of the target microseismic data is constructed; based on the inverse function of the cumulative distribution function, the target microseismic data is mapped to any reference anomaly identification model to obtain the target data set corresponding to the target microseismic data; the anomaly probability of the target sample in any reference anomaly identification model is acquired, and whether the target microseismic data is abnormal is determined based on the anomaly probability. By mapping the target microseismic data to the reference anomaly identification model of the reference coal mine, the difference between the scene where the target microseismic data is located and the scene where the reference coal mine is located is reduced, so that the reference anomaly identification model is migrated to the new microseismic identification scene, and the abnormal situation of the target coal mine is determined, which solves the problem of constructing a microseismic anomaly identification model for coal mines in the absence of microseismic data and unlabeled data, and improves the ability to identify abnormal microseismic states.

[0090] In the description of the aforementioned embodiments, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0091] In addition, the terms "first" and "second" are used for descriptive purposes only and should not 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 the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0092] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0093] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0094] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0095] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0096] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0097] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for identifying rock burst migration based on Weibull mixed model, characterized in that: The following steps are involved: Acquire historical microseismic monitoring data of any reference coal mine, and construct a corresponding reference anomaly identification model, wherein the reference anomaly identification model is used to determine the anomaly probability of any reference coal mine; Acquire target microseismic data of a target coal mine, and map the target microseismic data to any reference anomaly identification model to obtain a target data set corresponding to the target microseismic data, wherein the target data set includes a plurality of target samples corresponding to the target microseismic data; The abnormal probability of the target sample in any of the reference abnormality identification models is obtained, and whether the target microseismic data is abnormal is determined based on the abnormal probability.

2. The method according to claim 1, characterized in that Mapping the target microseismic data to any one of the reference anomaly identification models includes: Obtaining a cumulative distribution function of the target microseismic data; Based on the inverse function of the cumulative distribution function, the target microseismic data is mapped to any one of the reference anomaly identification models.

3. The method according to claim 2, characterized in that The step of obtaining the cumulative distribution function of the target microseismic data comprises: Acquire a probability density function of the target microseismic data; Optimizing the parameters corresponding to the probability density function by maximizing the likelihood function; Based on the optimized probability density function, the cumulative distribution function is obtained.

4. The method according to claim 1, characterized in that: The step of obtaining historical microseismic monitoring data of any reference coal mine and constructing a corresponding reference anomaly identification model includes: Acquire historical microseismic monitoring data and label data of any reference coal mine, wherein the label data is used to indicate whether the historical microseismic monitoring data is abnormal; Inputting the historical microseismic monitoring data into a decision tree model to be trained, and training the decision tree model based on the difference between the output of the decision tree model and the label data; The trained decision tree model is used as the reference abnormality identification model.

5. The rock burst migration identification method based on Weibull mixed model according to claim 1, characterized in that: The obtaining the abnormal probability of the target sample in any of the reference abnormality identification models includes: The target sample is input into any of the reference anomaly identification models, and the anomaly probability is output.

6. The method according to claim 1, characterized in that Determining whether the target microseismic data is abnormal based on the abnormal probability includes: Determining the abnormality degree corresponding to the target microseismic data based on each abnormality probability; When the abnormality is greater than a preset abnormality threshold, it is determined that the target microseismic data is abnormal.

7. The method according to claim 6, characterized in that The determining the abnormality degree corresponding to the target microseismic data based on each abnormality probability includes: The average value of each abnormality probability is calculated as the abnormality degree.

Images