Clastic rock lithology identification method and device based on satellite remote sensing image and medium

Abstract

The invention discloses a clastic rock lithology identification method and device based on a satellite remote sensing image and a medium. The clastic rock lithology identification method based on the satellite remote sensing image comprises the steps that a to-be-identified area is determined, the satellite remote sensing image of the to-be-identified area is acquired, and the satellite remote sensing image comprises at least one type of target rock area and at least one type of background ground feature area; performing image recognition on the input satellite remote sensing image through the trained lithology recognition model to obtain a lithology recognition result; and in response to manual lithology interpretation operation, performing lithology labeling on each type of target rock area to obtain a lithology type corresponding to each type of target rock area. Based on the technical scheme of the embodiment of the invention, the accuracy of rock lithology identification can be improved, and the cost of rock lithology identification can be reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent lithology identification, and in particular to a method, device and medium for identifying clastic rock lithology based on satellite remote sensing images. Background Art

[0002] With the development of oil and gas exploration technology, the direction of veins can be analyzed according to the lithology of rocks, and then the surface rock geological map of a certain area can be quickly produced to provide a scientific basis for determining rock porosity and oil saturation, studying geological reservoir characteristics, calculating reserves and geological modeling.

[0003] At present, the methods for identifying rock properties include: using the spectral characteristics of carbonate rocks to identify rock properties, and for the Jimusar area of ​​Xinjiang Uygur Autonomous Region, using remote sensing to interpret sedimentary rock properties based on the theory of rock spectra. However, the above methods are often costly and cannot achieve a certain degree of accuracy. Summary of the invention

[0004] The present invention provides a method, device and medium for identifying clastic rock lithology based on satellite remote sensing images, so as to solve the technical problems of high cost and poor accuracy in rock lithology identification.

[0005] According to one aspect of the present invention, a method for identifying lithology of clastic rocks based on satellite remote sensing images is provided, wherein the method comprises:

[0006] Determine an area to be identified, and obtain a satellite remote sensing image of the area to be identified, wherein the satellite remote sensing image includes at least one type of target rock area and at least one type of background object area;

[0007] The input satellite remote sensing image is subjected to image recognition through the trained lithology recognition model to obtain a lithology recognition result; wherein the lithology recognition result includes at least one type of the target rock area highlighted and a regional category label corresponding to each type of the target rock area, and the lithology recognition model is obtained by training a fully convolutional network based on sample remote sensing images;

[0008] In response to the manual lithology interpretation operation, lithology annotation is performed on each type of the target rock area to obtain the lithology category corresponding to each type of the target rock area.

[0009] According to another aspect of the present invention, a device for identifying lithology of clastic rocks based on satellite remote sensing images is provided, wherein the device comprises:

[0010] An image acquisition module is used to determine the area to be identified and acquire a satellite remote sensing image of the area to be identified, wherein the satellite remote sensing image includes at least one type of target rock area and at least one type of background object area;

[0011] A region recognition module, used to perform image recognition on the input satellite remote sensing image through the trained lithology recognition model to obtain a lithology recognition result; wherein the lithology recognition result includes at least one type of the target rock area highlighted and a region category label corresponding to each type of the target rock area, and the lithology recognition model is obtained by training a fully convolutional network based on sample remote sensing images;

[0012] The lithology identification module is used to label the lithology of each type of target rock area in response to the manual lithology interpretation operation, and obtain the lithology category corresponding to each type of target rock area.

[0013] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0014] at least one processor; and

[0015] a memory communicatively connected to the at least one processor; wherein,

[0016] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the method for identifying lithology of clastic rocks based on satellite remote sensing images as described in any embodiment of the present invention.

[0017] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the method for identifying lithology of clastic rocks based on satellite remote sensing images as described in any embodiment of the present invention when executed.

[0018] The technical solution of the embodiment of the present invention determines the area to be identified, obtains the satellite remote sensing image of the area to be identified, wherein the satellite remote sensing image includes at least one type of target rock area and at least one type of background feature area; performs image recognition on the input satellite remote sensing image through the trained lithology recognition model to obtain a lithology recognition result; wherein the lithology recognition result includes at least one type of the target rock area highlighted and the regional category label corresponding to each type of the target rock area, and the lithology recognition model is obtained by training the full convolutional network based on the sample remote sensing image; in response to the manual interpretation of lithology operation, lithology is annotated for each type of the target rock area to obtain the lithology category corresponding to each type of the target rock area. The accuracy of rock lithology recognition is improved and the cost of rock lithology recognition is reduced.

[0019] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 is a flow chart of a clastic rock lithology identification method based on satellite remote sensing images provided according to the first embodiment of the present invention;

[0022] Figure 2 is a scene diagram of a rock lithology identification process provided according to an embodiment of the present invention;

[0023] Figure 3 is a flow chart of a clastic rock lithology identification method based on satellite remote sensing images provided according to the second embodiment of the present invention;

[0024] Figure 4 is a schematic diagram of adjusting network parameters of a fully convolutional neural network provided according to an embodiment of the present invention;

[0025] Figure 5 It is a structural diagram of a FCN fully convolutional neural network provided according to an embodiment of the present invention;

[0026] Figure 6 is a schematic structural diagram of a clastic rock lithology identification device based on satellite remote sensing images provided according to Embodiment 3 of the present invention;

[0027] Figure 7 It is a schematic diagram of the structure of an electronic device for implementing the clastic rock lithology identification method based on satellite remote sensing images according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0029] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0030] Embodiment 1

[0031] Figure 1 A flowchart of a clastic rock lithology identification method based on satellite remote sensing images is provided for the first embodiment of the present invention. This embodiment is applicable to the case of intelligent lithology identification. The method can be executed by a clastic rock lithology identification device based on satellite remote sensing images. The clastic rock lithology identification device based on satellite remote sensing images can be implemented in the form of hardware and / or software. The clastic rock lithology identification device based on satellite remote sensing images can be configured in computer software. Figure 1 As shown, the method includes:

[0032] S110, determining an area to be identified, and acquiring a satellite remote sensing image of the area to be identified, wherein the satellite remote sensing image includes at least one type of target rock area and at least one type of background object area.

[0033] Among them, the area to be identified can be understood as an area to be identified for rock lithology. In the embodiment of the present invention, the area to be identified can be preset according to the scene requirements and is not specifically limited here. Optionally, in the scenario where the rock lithology to be identified is the lithology of clastic rock, the area to be identified can be an area containing clastic rock. Or in the scenario where the rock lithology to be identified is the lithology of carbonate rock, the area to be identified can be an area containing carbonate rock.

[0034] The satellite remote sensing image can be understood as a remote sensing image corresponding to the area to be identified obtained based on satellite remote sensing technology. It should be understood that the satellite remote sensing image can provide a large amount of multi-spectral and multi-resolution data. Among them, the satellite remote sensing image includes rich information such as the color and texture of the rock. Based on the satellite remote sensing image, the overall appearance of the rock can be intuitively displayed.

[0035] The target rock region can be understood as the region where the target rock is located. The target rock can be understood as the rock of the lithology to be identified. In the embodiment of the present invention, the target rock can be preset according to the scene requirements and is not specifically limited here. Optionally, the target rock can be clastic rock or carbonate rock, etc.

[0036] The background object area can be understood as the object area in the satellite remote sensing image except the target rock area.

[0037] Optionally, the satellite remote sensing image includes at least one of a blue band, a green band and a red band. It should be understood that in the satellite remote sensing image, the blue band can clearly obtain the boundary information of the intersection of objects. The green band can well reflect many information characteristics of the water body. The red band can separate buildings and vegetation in areas where urban artificial objects and vegetation are mixed. In an embodiment of the present invention, acquiring multi-band satellite remote sensing images can ensure the richness and accuracy of the acquired data of the area to be identified, so as to ensure the accuracy of rock lithology identification based on the satellite remote sensing images.

[0038] S120. Perform image recognition on the input satellite remote sensing image through the trained lithology recognition model to obtain a lithology recognition result; wherein the lithology recognition result includes at least one type of the target rock area highlighted and the regional category label corresponding to each type of the target rock area, and the lithology recognition model is obtained by training a fully convolutional network based on sample remote sensing images.

[0039] The lithology identification model may be understood as a model having the function of classifying different lithology areas.

[0040] The lithology recognition result can be understood as the region segmentation result corresponding to the satellite remote sensing image obtained by performing image recognition on the input satellite remote sensing image through the trained lithology recognition model. In an embodiment of the present invention, the lithology recognition result may include at least one type of the target rock area highlighted and the region category label corresponding to each type of the target rock area.

[0041] Optionally, the second segmentation result may also include a background object area that is not highlighted.

[0042] In the embodiment of the present invention, the target rock region can be understood as the region where the target rock is located. The lithology of the target rock can include multiple types. For example, when the target rock is a clastic rock, the lithology of the target rock can include conglomerate, sandstone, siltstone, and mudstone. Therefore, for example, the lithology identification result can include a highlighted conglomerate region, sandstone region, siltstone region, and mudstone region.

[0043] The region category labels can be understood as labels for distinguishing multiple types of target rock regions. Exemplarily, the region category labels can be labels of different colors.

[0044] The sample remote sensing image can be understood as sample data of the trained lithology recognition model. In an embodiment of the present invention, the sample remote sensing image can be a remote sensing image that does not contain a sample label. In other words, the present invention trains the fully convolutional network based on unsupervised learning to obtain the lithology recognition model.

[0045] The fully convolutional network can be understood as a model of the trained lithology recognition model. Optionally, the fully convolutional network can be a shallow fully convolutional network (FCN).

[0046] S130. In response to the manual lithology interpretation operation, lithology labeling is performed on each type of the target rock region to obtain a lithology category corresponding to each type of the target rock region.

[0047] The manual lithology interpretation operation may be understood as an operation of labeling the lithology of each type of target rock region.

[0048] The lithology category can be understood as the category of lithology corresponding to each type of the target rock area. It can be understood that each type of the target rock area can correspond to a type of lithology category of the target rock. Optionally, in the scenario where the target rock is clastic rock, the lithology category can include conglomerate, sandstone, siltstone, mudstone, etc., which are classified based on the particle size of the clastic rock.

[0049] Specifically, the lithology of each type of target rock area is manually labeled to obtain the lithology category corresponding to each type of target rock area, so as to obtain a rock lithology recognition image representing the lithology category corresponding to each type of target rock area (refer to Figure 2 ).

[0050] The technical solution of the embodiment of the present invention determines the area to be identified, obtains the satellite remote sensing image of the area to be identified, wherein the satellite remote sensing image includes at least one type of target rock area and at least one type of background feature area; performs image recognition on the input satellite remote sensing image through the trained lithology recognition model to obtain a lithology recognition result; wherein the lithology recognition result includes at least one type of the target rock area highlighted and the regional category label corresponding to each type of the target rock area, and the lithology recognition model is obtained by training the full convolutional network based on the sample remote sensing image; in response to the manual interpretation of lithology operation, lithology is annotated for each type of the target rock area to obtain the lithology category corresponding to each type of the target rock area. The accuracy of rock lithology recognition is improved and the cost of rock lithology recognition is reduced.

[0051] Embodiment 2

[0052] Figure 3 The flowchart of a clastic rock lithology identification method based on satellite remote sensing images provided in the second embodiment of the present invention is an addition to the determination of the area to be identified in the above embodiment. Figure 3 As shown, the method includes:

[0053] S210, obtaining a sample remote sensing image, and performing image segmentation on the sample remote sensing image using the Filsenzwab algorithm and the fully convolutional network to obtain a first segmentation result and a second segmentation result, respectively.

[0054] The Felzenszwalb algorithm can be understood as an image segmentation algorithm.

[0055] The first segmentation result can be understood as a segmentation result obtained by performing image segmentation on the current sample remote sensing image based on the Felzenszwalb algorithm. The first segmentation result may include multiple unlabeled pixel segmentation regions. Optionally, the pixel segmentation region is an irregular pixel segmentation region.

[0056] The second segmentation result can be understood as a segmentation result obtained by performing lithology recognition on the current sample remote sensing image based on the full convolutional network.

[0057] Optionally, the performing image segmentation on the sample remote sensing image by using the Filsenzwab algorithm and the fully convolutional network to obtain a first segmentation result and a second segmentation result respectively includes:

[0058] Performing image segmentation on the sample remote sensing image by using the Filsenzwab algorithm to obtain a first segmentation result, wherein the first segmentation result includes a plurality of unlabeled pixel segmentation regions;

[0059] The input sample remote sensing image is segmented by the fully convolutional network to obtain a second segmentation result, wherein the second segmentation result includes at least one type of rock segmentation area highlighted and a segmentation category label corresponding to each type of rock segmentation area. Optionally, the second segmentation result may also include a background feature area that is not highlighted. The rock segmentation area can be understood as the segmentation area where the target rock is located in the sample remote sensing image.

[0060] The segmentation category label can be understood as a label for distinguishing multiple types of rock segmentation regions. Exemplarily, the rock segmentation regions can be different color labels. In an embodiment of the present invention, the segmentation category label is annotated on each pixel in the current rock segmentation region.

[0061] S220. Determine a sample segmentation result corresponding to the sample remote sensing image based on the first segmentation result and the second segmentation result, wherein the sample segmentation result includes at least one type of highlighted sample segmentation area and a pseudo label corresponding to each type of the sample segmentation area.

[0062] The sample segmentation result can be understood as a segmentation result determined by combining the first segmentation result and the second segmentation result. The sample segmentation region in the sample segmentation result can be understood as a segmentation region determined based on the pixel segmentation region in the first segmentation result and the rock segmentation region in the second segmentation result. The pseudo label in the sample segmentation result is a label that characterizes the boundary information of each type of the sample segmentation region, determined based on the probability of the segmentation category label appearing in the pixel segmentation region.

[0063] Specifically, the specific process of performing image segmentation on the sample remote sensing image by using the Filsenzwab algorithm to obtain the first segmentation result may be:

[0064] The satellite remote sensing image is subjected to image segmentation processing, and the optimal Felzenszwalb algorithm is adopted. The algorithm is a graph-based segmentation method, which retains the details of low-variability image areas and ignores the details of high-variability image areas. It has a single scale parameter that affects the size of the segmented fragments, and the running time is approximately linearly related to the number of graph edges.

[0065] Detailed steps of Felzenszwalb segmentation algorithm:

[0066] Step 1: Calculate the dissimilarity between each pixel and its four-neighborhood or eight-neighborhood.

[0067] Step 2: Sort the edges according to the non-decreasing edge weights of dissimilarity to obtain e 1 ,e2 …e N .

[0068] Step 3: Select e 1 .

[0069] Step 4: For the currently selected edge e n Make a merge judgment. Let the connected vertices be (v i ,v j ), if the merging conditions are met: (1)v i ,v j Do not belong to the same area, that is, Id(v i )≠Id(v j ); (2) the dissimilarity is not greater than the internal dissimilarity of the two, that is, w i,j ≤Mint(C i ,C j ), execute Step 5; otherwise, execute Step 6.

[0070] Step 5: Update the threshold and class label, and change Id(v i ), Id(v j ) is uniformly labeled as Id(v i ) and set the dissimilarity threshold of this class to

[0071] Note: Since the edges with small dissimilarity are merged first, w i,j is the largest edge in the current merged region, that is, Int(C i ∪C j )=w i,j .

[0072] Step 6: If n≤N, select the next edge in sequence and go to Step 4, otherwise end.

[0073] Through testing, it can be seen that in order to make the program have a suitable running time and good stability, the segmentation scale is set to 30, the satellite remote sensing image segmentation processing is completed, and the first segmentation result is obtained.

[0074] Specifically, the specific step of determining the sample segmentation result corresponding to the sample remote sensing image based on the first segmentation result and the second segmentation result may be:

[0075] The first segmentation result is an irregular pixel area with different color features. Most of these areas retain the effective information of image segmentation, but do not contain semantic information. Therefore, they cannot correspond to the features extracted by the neural network, that is, they cannot be used as labels required for training, so the synthetic pseudo-label method is used.

[0076] A remote sensing image is defined as and I∈R c×w×h , where {I i} represents the image pixel set. First, superpixel segmentation is performed on I to generate irregular pixel regions, i.e., the first segmentation result where {s r} is the set of regions after image segmentation, and R is the number of regions. Then the image I is passed to the shallow FCN network in a random order for feature extraction. In each iteration, the image I is convolved in turn and a feature map is generated. That is, the second segmentation result, so that each pixel i has a category label, and it is defined as m. According to the contour and feature map y of each region of the first segmentation result S, that is, the second segmentation result, each region {s r} corresponds to S in y, we can calculate m in this area {s r The value of m with the highest number of occurrences is recorded as M, that is, the region The pseudo label is M. As shown in formula L i (y,s)=argmax m ∥{y i =m|{i,j}∈{s r}}∥., where j represents the region {s r}, further, generate pseudo labels

[0077]

[0078] Optionally, determining a sample segmentation result corresponding to the sample remote sensing image based on the first segmentation result and the second segmentation result includes:

[0079] Determine a preliminary segmentation result based on the first segmentation result and the second segmentation result;

[0080] The preliminary segmentation result is post-processed by a majority filtering algorithm to obtain the sample segmentation result.

[0081] The preliminary segmentation result may be understood as a preliminary segmentation result determined based on the first segmentation result and the second segmentation result. In the embodiment of the present invention, the preliminary segmentation result needs to be subjected to classification post-processing to obtain the sample segmentation result.

[0082] Specifically, the majority filtering algorithm is used to replace the value of a pixel with the majority of the pixel values ​​in the neighborhood around the pixel, and small spots are removed, so that an image with smoother edges can be obtained, that is, the sample segmentation result, in preparation for vectorization.

[0083] S230, determining a true label of the sample segmentation result, and adjusting the network parameters of the full convolutional network by back-propagation based on the pseudo label and the true label to determine the lithology identification model.

[0084] The true label can be understood as a label that characterizes the boundary information of each type of sample segmentation region. Specifically, in the embodiment of the present invention, the pseudo label can be adjusted to obtain the true label.

[0085] The network parameters can be understood as model parameters of the fully convolutional network.

[0086] Optionally, determining a true label of the sample segmentation result includes:

[0087] In response to a manual intervention operation, the pseudo-label is adjusted to obtain the true label, wherein the true label is a label that characterizes boundary information of each type of sample segmentation region.

[0088] The manual intervention operation can be understood as adjusting the boundary information corresponding to the pseudo label to obtain the boundary information corresponding to the true label (refer to Figure 4 ).

[0089] Optionally, determining the lithology identification model includes:

[0090] When the preset conditions are met, the current fully convolutional network is determined as the lithology identification model, wherein the preset conditions are that the current adjustment parameter of the network parameter is less than the parameter adjustment threshold, and the current adjustment parameter is an adjustment parameter determined based on the current true label and the current pseudo label.

[0091] The current adjustment parameter may be understood as an adjustment parameter determined based on the difference between the boundary information corresponding to the current true label and the boundary information corresponding to the current pseudo label.

[0092] The parameter adjustment threshold can be understood as a threshold for ending the network parameter adjustment of the full convolutional network. In the embodiment of the present invention, the parameter adjustment threshold can be preset according to the scene requirements and is not specifically limited here.

[0093] In an embodiment of the present invention, the specific steps of training the fully convolutional network may be:

[0094] The shallow FCN neural network is an end-to-end, pixel-by-pixel fully convolutional network used for semantic segmentation of images. The overall network structure is divided into two parts: the full convolution part and the upsampling part.

[0095] Assume that the input size is 256×256×3 color image, after a convolution layer with 64 output channels and 3×3 convolution kernel size, a 256×256×64 feature map is obtained. Then the data is normalized, the feature number remains unchanged at 64, and the activation function is selected as ReLU function. After that, a 2×2 pooling layer is passed to obtain a 128×128×64 feature map. Repeat the above operation twice, but the result obtained in the last time does not enter the pooling layer. At this point, the downsampling is completed and a 64×64×64 feature map is obtained.

[0096] Then, the output size is set to be twice the input size, the data is upsampled, and then the data is processed in sequence through a convolution layer with an output channel number of 64 and a convolution kernel size of 3×3, normalized with a feature number of 64, and the activation function is ReLU function. After the above steps are repeated once, the data enters a convolution layer with an output channel number equal to the number of classification categories and a convolution kernel size of 1×1, and then normalized with a feature number equal to the number of classification categories. The final output size is the size of the original image, and the number of channels is the number of classification categories.

[0097] The forward propagation of the neural network requires a large number of feature maps to be stored in memory. The standard approach to solve this problem is to use small blocks of images in the training set instead of the complete image as input into the neural network. In order to avoid boundary effects, a block-based training scheme is adopted with a block size of 512 pixels and a step size of 448 pixels. The majority voting method is performed in the overlapping area to output the final merged result. Through this mechanism, the memory consumption is relatively small, and it can be quickly applied to large-scale complex remote sensing images. The specific method is: define a remote sensing image as I, perform superpixel segmentation on I, and generate an irregular pixel area defined as S. Divide the original image (I) and the image segmentation result (S) into N blocks in the same way to obtain and in Represents each image block, express The corresponding image segmentation results, and all {I ′ n} and {S ′ n Generate pseudo labels.

[0098] Calculate the gradient of the SoftMax Loss function to each parameter in the network, update the parameters according to the optimization condition that the pixels in the same segmentation object should belong to the same category, reduce the loss function, and perform back propagation.

[0099] The calculation formula of SoftMax Loss loss function is as follows:

[0100]

[0101] Where M represents the number of categories, y tc represents the true category of sample i, p ic It represents the predicted probability that the observed sample i belongs to category c.

[0102] Among them, y tc It takes 1 if c is equal to c, otherwise it takes 0.

[0103] The specific approach is to calculate the classification cross entropy loss function of the pseudo label and feature map, and back-propagate to the neural network. In each iteration, the network training and classification prediction are performed sequentially. When N blocks of images are trained, the network parameters have been updated N times. Then the trained network is used to predict the classification results of the image, and the prediction results also perform the same block-based steps.

[0104] In summary, the present invention has a high degree of automation and good image processing effect, can save costs, and efficiently serve the basic work of geological survey and research and mineral resource exploration. The cost required for accurate detection and identification of rocks is far lower than the expensive cost required for one-by-one and regular field surveys, which provides great convenience for assisting geologists in judging and identifying rock properties.

[0105] Furthermore, the weakly supervised remote sensing image segmentation method based on the Felzenszwalb segmentation algorithm and the shallow FCN is used for model training, which has good performance, high recognition accuracy and low memory usage. The system takes a short time to run, is closer to the actual situation in terms of visualization results, and can encode richer classification information. At the same time, it is very helpful for the segmentation of large-scale remote sensing images containing complex targets.

[0106] Furthermore, by training the improved FCN (ref. Figure 5 ) can intelligently identify the lithology of outcrop clastic rocks in satellite remote sensing images. In areas with good outcrop of clastic rocks, the interpretation of dykes using satellite remote sensing images has a special interpretation effect. The period and intersection relationship of dykes, as well as the study of magmatic activity and tectonic change period using the distribution characteristics of dykes are more intuitive and realistic than conventional geological work methods.

[0107] S240: Determine an area to be identified, and obtain a satellite remote sensing image of the area to be identified, wherein the satellite remote sensing image includes at least one type of target rock area and at least one type of background object area.

[0108] S250. Performing image recognition on the input satellite remote sensing image through the trained lithology recognition model to obtain a lithology recognition result; wherein the lithology recognition result includes at least one type of the target rock area highlighted and a regional category label corresponding to each type of the target rock area, and the lithology recognition model is obtained by training a fully convolutional network based on sample remote sensing images.

[0109] S260: In response to the manual lithology interpretation operation, lithology labeling is performed on each type of the target rock region to obtain the lithology category corresponding to each type of the target rock region.

[0110] The technical solution of the embodiment of the present invention is to obtain a sample remote sensing image, perform image segmentation on the sample remote sensing image through the Filsenzwab algorithm and the fully convolutional network, and obtain a first segmentation result and a second segmentation result respectively; determine the sample segmentation result corresponding to the sample remote sensing image based on the first segmentation result and the second segmentation result, wherein the sample segmentation result includes at least one type of sample segmentation area highlighted and a pseudo label corresponding to each type of the sample segmentation area; determine the true label of the sample segmentation result, and adjust the network parameters of the fully convolutional network based on the pseudo label and the true label by back propagation to determine the lithology identification model. The model accuracy of the determined lithology identification model is improved.

[0111] Embodiment 3

[0112] Figure 6 The schematic diagram of the structure of a clastic rock lithology identification device based on satellite remote sensing images provided in the third embodiment of the present invention. Figure 6 As shown, the device includes: an image acquisition module 310, a region identification module 320 and a lithology identification module 330; wherein,

[0113] The image acquisition module 310 is used to determine the area to be identified and obtain the satellite remote sensing image of the area to be identified, wherein the satellite remote sensing image includes at least one type of target rock area and at least one type of background feature area; the area identification module 320 is used to perform image recognition on the input satellite remote sensing image through the trained lithology identification model to obtain a lithology identification result; wherein the lithology identification result includes at least one type of the target rock area highlighted and the regional category label corresponding to each type of the target rock area, and the lithology identification model is obtained by training the full convolutional network based on the sample remote sensing image; the lithology identification module 330 is used to respond to the manual interpretation of lithology operation, perform lithology labeling on each type of the target rock area, and obtain the lithology category corresponding to each type of the target rock area.

[0114] The technical solution of the embodiment of the present invention determines the area to be identified, obtains the satellite remote sensing image of the area to be identified, wherein the satellite remote sensing image includes at least one type of target rock area and at least one type of background feature area; performs image recognition on the input satellite remote sensing image through the trained lithology recognition model to obtain a lithology recognition result; wherein the lithology recognition result includes at least one type of the target rock area highlighted and the regional category label corresponding to each type of the target rock area, and the lithology recognition model is obtained by training the full convolutional network based on the sample remote sensing image; in response to the manual interpretation of lithology operation, lithology is annotated for each type of the target rock area to obtain the lithology category corresponding to each type of the target rock area. The accuracy of rock lithology recognition is improved and the cost of rock lithology recognition is reduced.

[0115] Optionally, the clastic rock lithology identification device based on satellite remote sensing images further includes: an image segmentation module, a pseudo-label confirmation module and a back propagation module; wherein,

[0116] The image segmentation module is used to obtain a sample remote sensing image before determining the area to be identified, and perform image segmentation on the sample remote sensing image by using the Filsenzwab algorithm and the fully convolutional network to obtain a first segmentation result and a second segmentation result respectively;

[0117] The pseudo label confirmation module is used to determine a sample segmentation result corresponding to the sample remote sensing image based on the first segmentation result and the second segmentation result, wherein the sample segmentation result includes at least one type of highlighted sample segmentation area and a pseudo label corresponding to each type of the sample segmentation area;

[0118] The back propagation module is used to determine the true label of the sample segmentation result, and adjust the network parameters of the full convolution network based on the pseudo label and the true label through back propagation to determine the lithology identification model.

[0119] Optionally, the image segmentation module is used to:

[0120] Performing image segmentation on the sample remote sensing image by using the Filsenzwab algorithm to obtain a first segmentation result, wherein the first segmentation result includes a plurality of unlabeled pixel segmentation regions;

[0121] The input sample remote sensing image is segmented by the fully convolutional network to obtain a second segmentation result, wherein the second segmentation result includes at least one type of rock segmentation area highlighted and a segmentation category label corresponding to each type of the rock segmentation area.

[0122] Optionally, the back propagation module includes: a true label determination unit, used to adjust the pseudo label in response to a manual intervention operation to obtain the true label, wherein the true label is a label characterizing boundary information of each type of sample segmentation area.

[0123] Optionally, the back propagation module includes: a model determination unit, used to determine the current full convolutional network as the lithology identification model when a preset condition is met, wherein the preset condition is that the current adjustment parameter of the network parameter is less than a parameter adjustment threshold, and the current adjustment parameter is an adjustment parameter determined based on the current true label and the current pseudo-label.

[0124] Optionally, the pseudo label confirmation module is used to:

[0125] Determine a preliminary segmentation result based on the first segmentation result and the second segmentation result;

[0126] The preliminary segmentation result is post-processed by a majority filtering algorithm to obtain the sample segmentation result.

[0127] Optionally, the satellite remote sensing image includes at least one of a blue band, a green band and a red band.

[0128] The device for identifying lithology of clastic rocks based on satellite remote sensing images provided in an embodiment of the present invention can execute the method for identifying lithology of clastic rocks based on satellite remote sensing images provided in any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

[0129] Embodiment 4

[0130] Figure 7 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0131] like Figure 7As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0132] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0133] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a clastic rock lithology identification method based on satellite remote sensing images.

[0134] In some embodiments, the clastic rock lithology identification method based on satellite remote sensing images can be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the clastic rock lithology identification method based on satellite remote sensing images described above can be executed. Alternatively, in other embodiments, the processor 11 can be configured to execute the clastic rock lithology identification method based on satellite remote sensing images in any other appropriate manner (for example, by means of firmware).

[0135] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0136] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0137] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0138] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0139] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0140] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0141] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0142] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for identifying lithology of clastic rocks based on satellite remote sensing images, characterized in that: include: Determine an area to be identified, and obtain a satellite remote sensing image of the area to be identified, wherein the satellite remote sensing image includes at least one type of target rock area and at least one type of background object area; The input satellite remote sensing image is subjected to image recognition through the trained lithology recognition model to obtain a lithology recognition result; wherein the lithology recognition result includes at least one type of the target rock area highlighted and a regional category label corresponding to each type of the target rock area, and the lithology recognition model is obtained by training a fully convolutional network based on sample remote sensing images; In response to the manual lithology interpretation operation, lithology annotation is performed on each type of the target rock area to obtain the lithology category corresponding to each type of the target rock area.

2. The method according to claim 1, characterized in that Before determining the area to be identified, the method further includes: Acquire a sample remote sensing image, and perform image segmentation on the sample remote sensing image using the Filsenzwab algorithm and the fully convolutional network to obtain a first segmentation result and a second segmentation result respectively; Determining a sample segmentation result corresponding to the sample remote sensing image based on the first segmentation result and the second segmentation result, wherein the sample segmentation result includes at least one type of highlighted sample segmentation area and a pseudo label corresponding to each type of the sample segmentation area; The true label of the sample segmentation result is determined, and the network parameters of the full convolutional network are adjusted by back-propagation based on the pseudo label and the true label to determine the lithology identification model.

3. The method according to claim 2, characterized in that The sample remote sensing image is segmented by the Filsenzwab algorithm and the fully convolutional network to obtain a first segmentation result and a second segmentation result, respectively, including: Performing image segmentation on the sample remote sensing image by using the Filsenzwab algorithm to obtain a first segmentation result, wherein the first segmentation result includes a plurality of unlabeled pixel segmentation regions; The input sample remote sensing image is segmented by the fully convolutional network to obtain a second segmentation result, wherein the second segmentation result includes at least one type of rock segmentation area highlighted and a segmentation category label corresponding to each type of the rock segmentation area.

4. The method according to claim 2, characterized in that: Determining the true label of the sample segmentation result includes: In response to a manual intervention operation, the pseudo-label is adjusted to obtain the true label, wherein the true label is a label that characterizes boundary information of each type of sample segmentation region.

5. The method according to claim 2, characterized in that: Determining the lithology identification model includes: When the preset conditions are met, the current fully convolutional network is determined as the lithology identification model, wherein the preset conditions are that the current adjustment parameter of the network parameter is less than the parameter adjustment threshold, and the current adjustment parameter is an adjustment parameter determined based on the current true label and the current pseudo label.

6. The method according to claim 2, characterized in that The determining the sample segmentation result corresponding to the sample remote sensing image based on the first segmentation result and the second segmentation result includes: Determine a preliminary segmentation result based on the first segmentation result and the second segmentation result; The preliminary segmentation result is post-processed by a majority filtering algorithm to obtain the sample segmentation result.

7. The method according to claim 1, characterized in that The satellite remote sensing image includes at least one of a blue band, a green band and a red band.

8. A clastic rock lithology identification device based on satellite remote sensing images, characterized in that: include: An image acquisition module is used to determine the area to be identified and acquire a satellite remote sensing image of the area to be identified, wherein the satellite remote sensing image includes at least one type of target rock area and at least one type of background object area; A region recognition module, used to perform image recognition on the input satellite remote sensing image through the trained lithology recognition model to obtain a lithology recognition result; wherein the lithology recognition result includes at least one type of the target rock area highlighted and a region category label corresponding to each type of the target rock area, and the lithology recognition model is obtained by training a fully convolutional network based on sample remote sensing images; The lithology identification module is used to respond to the manual lithology interpretation operation, mark the lithology of each type of target rock area, and obtain the lithology category corresponding to each type of target rock area.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the clastic rock lithology identification method based on satellite remote sensing images as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the clastic rock lithology identification method based on satellite remote sensing images according to any one of claims 1 to 7 when executed.

Images