Construction progress identification method and device, computer equipment and storage medium

Abstract

The invention relates to a construction progress recognition method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring satellite remote sensing image data of an area containing a to-be-identified construction site, and determining coordinate information of the to-be-identified construction site based on the satellite remote sensing image data; based on the coordinate information and the satellite remote sensing image data of the construction site to be recognized, obtaining a first probability set, a second probability set and a third probability set containing multiple construction stage probabilities through a scene classification model obtained through pre-training, a convolutional neural network model obtained through pre-training and a residual network model obtained through pre-training; and determining the construction progress of the to-be-identified construction site according to the first probability set, the second probability set and the third probability set which all contain the plurality of construction stage probabilities. By adopting the method, the construction progress can be identified and monitored accurately.

Description

Technical Field

[0001] The present disclosure relates to the field of image processing technology, and in particular to a construction progress recognition method, device, computer equipment, and storage medium. Background Art

[0002] With the rapid development of various shooting equipment and computer vision technology, various infrastructures have emerged in various regions of the country. However, most of the more common construction progress automatic monitoring algorithms in the industry can only identify the three states of construction projects: not started, started, and completed, and cannot reflect the specific progress of the construction process in detail.

[0003] High-resolution remote sensing satellites have the advantages of wide coverage, short imaging cycle, and high resolution. They can remotely, dynamically, and accurately monitor the progress of projects and assist in the planning, implementation, and management of major projects. In recent years, with the accelerated advancement of my country's commercial aerospace process, the number of sub-meter commercial remote sensing satellites has continued to grow, spatial resolution has continued to increase, and spectral information has become richer, making it possible to use satellite remote sensing images to automatically monitor construction progress. Domestic policies also guide and call on major / key projects across the country to actively adopt advanced technical means to carry out progress verification. Therefore, the use of satellite remote sensing data for construction progress monitoring is urgently needed in practical applications.

[0004] How to use satellite remote sensing data to accurately identify and monitor construction progress is an urgent problem that needs to be solved. Summary of the invention

[0005] Based on this, it is necessary to provide a construction progress identification method, device, computer equipment, and storage medium to address the above technical problems.

[0006] In a first aspect, the present disclosure provides a construction progress identification method. The method comprises:

[0007] Acquire satellite remote sensing image data of an area containing a construction site to be identified, and determine coordinate information of the construction site to be identified based on the satellite remote sensing image data;

[0008] Based on the coordinate information of the construction site to be identified and the satellite remote sensing image data, a pre-trained scene classification model, a pre-trained convolutional neural network model, and a pre-trained residual network model are used to obtain a first probability set, a second probability set, and a third probability set, each of which contains multiple construction stage probabilities;

[0009] The scene classification model, the convolutional neural network model and the residual network model are respectively trained using sample training data; the sample training data is obtained based on training remote sensing image label data; the training remote sensing image label data includes labels for characterizing construction stages and training remote sensing images corresponding to the construction stages;

[0010] The construction progress of the to-be-identified construction site is determined according to the first probability set, the second probability set and the third probability set, each of which contains multiple construction stage probabilities.

[0011] In one embodiment, determining the construction progress of the to-be-identified construction site according to the first probability set, the second probability set, and the third probability set, each of which contains multiple construction stage probabilities, includes:

[0012] Adding the probability of each construction stage in the first probability set, the second probability set and the third probability set to obtain a total probability of each construction stage;

[0013] Determine the target construction stage according to the maximum probability of the total probability of each construction stage;

[0014] The construction progress of the to-be-identified construction site is determined according to the target construction stage.

[0015] In one embodiment, the method further comprises:

[0016] Acquiring training remote sensing image area data, wherein the training remote sensing image area data contains construction sites in multiple different construction stages;

[0017] Labeling construction sites in multiple different construction stages contained in the training remote sensing image area data;

[0018] The scene classification model, the convolutional neural network model and the residual network model are trained respectively using the labeled training remote sensing image area data.

[0019] In one embodiment, labeling the construction sites in different construction stages in the training remote sensing image area data includes:

[0020] Determining the construction stage of each construction site in the training remote sensing image based on predetermined landmark features of the construction stage;

[0021] Use different construction labels to label different construction stages;

[0022] According to the construction stage and the preset division ratio, the construction site in the training remote sensing image is divided into a training set and a data set.

[0023] In one embodiment, the method further comprises:

[0024] Acquiring training remote sensing image area data, wherein the training remote sensing image area data contains construction sites in multiple different construction stages;

[0025] The training remote sensing image area data contains construction sites in multiple different construction stages, and the construction start labels of the construction sites are determined according to the construction stages, wherein the construction start labels include construction start labels and non-construction start labels;

[0026] The target detection model is trained based on the construction site's construction start label and the training remote sensing image area data.

[0027] In one embodiment, determining the coordinate information of the construction site to be identified based on the satellite remote sensing image data includes:

[0028] The trained target detection model is used to perform target detection on the satellite remote sensing image data to determine the coordinate information of the construction site to be identified.

[0029] In one embodiment, after determining the coordinate information of the construction site to be identified, the method further includes:

[0030] The coordinate information of the construction site to be identified is processed using a non-maximum suppression (NMS) post-processing algorithm.

[0031] In a second aspect, the present disclosure further provides a construction progress identification device. The device comprises:

[0032] A data acquisition module, used to acquire satellite remote sensing image data of an area containing a construction site to be identified, and determine coordinate information of the construction site to be identified based on the satellite remote sensing image data;

[0033] A model processing module, for obtaining a first probability set, a second probability set, and a third probability set, each containing a plurality of construction stage probabilities, based on the coordinate information of the construction site to be identified and the satellite remote sensing image data, using a pre-trained scene classification model, a pre-trained convolutional neural network model, and a pre-trained residual network model;

[0034] The scene classification model, the convolutional neural network model and the residual network model are respectively trained using sample training data; the sample training data is obtained based on training remote sensing image label data; the training remote sensing image label data includes labels for characterizing construction stages and training remote sensing images corresponding to the construction stages;

[0035] The construction progress determination module is used to determine the construction progress of the to-be-identified construction site based on the first probability set, the second probability set and the third probability set, each of which contains multiple construction stage probabilities.

[0036] In a third aspect, the present disclosure further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps in any of the above method embodiments are implemented.

[0037] In a fourth aspect, the present disclosure further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0038] In a fifth aspect, the present disclosure further provides a computer program product, wherein the computer program product comprises a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0039] In the above embodiments, by acquiring satellite remote sensing image data containing the construction site area to be identified, the overall situation of the area can be intuitively understood from a macro level, including the topography, traffic roads, other adjacent buildings, etc. around the construction site, providing rich background information and spatial references for the subsequent in-depth analysis of the construction site itself, avoiding focusing only on the local area and ignoring the possible impact of the surrounding environment on the construction. The coordinate information of the construction site is determined based on satellite remote sensing image data, so that the construction site has an accurate positioning in the geographic space. This not only facilitates the subsequent accurate marking of its location in geographic information systems such as maps, but also can be integrated and analyzed with other geographic data (such as urban planning data, infrastructure distribution data, etc.), which is conducive to the overall planning of regional construction resources and the evaluation of the impact of the construction site on the surrounding areas. The pre-trained scene classification model is usually trained based on a large amount of different scene image data. It can analyze the characteristics of the construction site to be identified from the perspective of the overall scene, and identify different scene modes presented in different construction stages, such as large-scale land leveling scenes in the infrastructure stage, towering scaffolding scenes in the main structure construction stage, and relatively regular facade scenes in the decoration stage, etc., to provide a basis for judging the construction progress based on the macroscopic scene level. CNN is good at automatically extracting deep features from image data, and can effectively capture the texture, shape, color and other detailed features of the construction site in satellite remote sensing images. For example, it can infer the construction stage based on the shape and distribution changes of the stacked building materials (which change as the construction stage progresses), the gradual clarity of the main building outline and other feature changes, and then output the probability of each construction stage, so that the judgment of the construction progress is more detailed and more accurate. ResNet can well solve the problem of gradient disappearance during deep network training, so that a deeper network structure can be constructed to deeply explore the potential features of different construction stages of construction sites in satellite remote sensing images and their inherent connections. For example, it can analyze the correlation between subtle changes in the internal structure of the building at different stages and changes in the appearance characteristics, and more comprehensively consider the various factors that affect the construction progress. The output construction stage probability is more valuable for reference. The three models can reduce the risk of overfitting of a single model. Moreover, the three models differ in structure, parameters and feature extraction methods, which can capture different information, improve generalization ability, and thus improve classification accuracy. Comprehensively considering the first probability set, the second probability set and the third probability set to determine the construction progress of the construction site to be identified is equivalent to integrating the analysis results of multiple models with different perspectives and different characteristics. Different models have their own advantages in analyzing image features and identifying construction stages. By integrating their probability outputs, we can learn from each other's strengths and weaknesses, so that the final construction progress is more in line with the actual situation, more scientific and objective. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the specific embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the specific embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0041] Figure 1 A schematic diagram of an application environment of a construction progress identification method in an embodiment;

[0042] Figure 2 A schematic diagram of a flow chart of a construction progress identification method in one embodiment;

[0043] Figure 3 This is a schematic diagram of the process of step S206 in one embodiment;

[0044] Figure 4 A flowchart of the steps of model training in one embodiment;

[0045] Figure 5 Schematic diagram of the process of step S404 in one embodiment;

[0046] Figure 6 A schematic diagram of a flow chart of steps for training a target detection model in one embodiment;

[0047] Figure 7 is a schematic block diagram of the structure of a construction progress identification device in one embodiment;

[0048] Figure 8 is a schematic diagram of the internal structure of a computer device in one embodiment;

[0049] Fig. 9 Schematic diagram of the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the present disclosure more clear, the present disclosure is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present disclosure and are not used to limit the present disclosure.

[0051] It should be noted that the terms "first", "second", etc. in the specification and claims of this article 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 this article described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, device, product or equipment that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0052] In this article, the term "and / or" is only a description of the association relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the related objects before and after are in an "or" relationship.

[0053] The present disclosure provides a construction progress identification method, which can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. The terminal 102 can obtain satellite remote sensing image data of the area containing the construction site to be identified in the server 104. The terminal 102 determines the coordinate information of the construction site to be identified based on the satellite remote sensing image data. Based on the coordinate information of the construction site to be identified and the satellite remote sensing image data, the terminal 102 obtains a first probability set, a second probability set, and a third probability set containing multiple construction stage probabilities through a pre-trained scene classification model, a pre-trained convolutional neural network model, and a pre-trained residual network model. Among them, the scene classification model, the convolutional neural network model, and the residual network model can be obtained by the terminal 102 or the server 104 respectively using sample training data; the sample training data is obtained based on the training remote sensing image label data; the training remote sensing image label data includes a label for characterizing the construction stage and a training remote sensing image corresponding to the construction stage. The terminal 102 determines the construction progress of the construction site to be identified based on the first probability set, the second probability set and the third probability set, each of which contains multiple construction stage probabilities. The terminal 102 may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices may be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers. It should be noted that this solution may also be implemented by the terminal 102 or the server 104 alone.

[0054] In one embodiment, Figure 2 As shown, a construction progress identification method is provided, which is applied to Figure 1 The terminal 102 in the example is used as an example to illustrate, and the following steps are included:

[0055] S202: Acquire satellite remote sensing image data of an area containing a construction site to be identified, and determine coordinate information of the construction site to be identified based on the satellite remote sensing image data.

[0056] Among them, satellite remote sensing image data is obtained by observing the earth's surface and recording the image information obtained through remote sensing sensors carried by artificial satellites.

[0057] Specifically, it is possible to connect to a remote sensing satellite to obtain satellite remote sensing image data taken by a remote sensing satellite in a certain area. The remote sensing image data can usually be high-resolution remote sensing image data. For high-resolution satellite remote sensing images, professional remote sensing image processing software, such as ENVI, ArcGIS, etc., can be used to directly visually interpret the image, identify the location of the construction site, and then use the coordinate measurement tool in the software to read the pixel coordinates of the construction site in the image. Then, according to the geographic coordinate information and projection method of the image, the pixel coordinates are converted into actual geographic coordinates. Image processing algorithms and machine learning models can also be used to automatically process and analyze satellite remote sensing images, extract characteristic information of the construction site, such as shape, texture, spectrum, etc., and then determine its location and range. For example, by constructing a convolutional neural network (CNN) model, a large number of labeled construction site images are trained so that it can automatically identify construction sites in new images and output their location coordinates.

[0058] In some exemplary embodiments, after acquiring satellite remote sensing image data, the image can be subjected to radiation correction, orthorectification, registration fusion, mosaic cropping and image enhancement preprocessing. Among them, the image enhancement uses a standard deviation stretching method to remove outliers in the image, enhance the contrast of ground objects, and reduce the image from 16 bits to 8 bits to obtain red, green and blue three-band orthophotos, thereby ensuring the accuracy of subsequent processing.

[0059] S204, based on the coordinate information of the construction site to be identified and the satellite remote sensing image data, a pre-trained scene classification model, a pre-trained convolutional neural network model and a pre-trained residual network model are used to obtain a first probability set, a second probability set and a third probability set, each of which contains multiple construction stage probabilities.

[0060] Among them, the scene classification model can be a YOLOv8 model. The YOLOv8 scene classification model is based on the Darknet framework, uses CSPDarknet53 as the backbone network, combines PANet for feature fusion, and sets detection heads at different scales for prediction. The convolutional neural network model can be VGG16. The algorithm structure of VGG16 consists of 13 convolutional layers and 3 fully connected layers, a total of 16 layers. Its convolutional layers are organized into 5 blocks, each of which uses the same 3x3 convolution kernel size, followed by a ReLU activation function and a maximum pooling layer. The residual network model can be a ResNet50 model. The algorithm structure of ResNet50 is based on a deep residual network and contains 50 layers of neural networks. Its core consists of 4 groups of residual blocks, each of which contains multiple residual units, and each residual unit has three convolutional layers (1x1, 3x3, 1x1) for feature extraction and dimensionality change. The input image first passes through a 7x7 convolutional layer and a maximum pooling layer, then passes through four groups of residual blocks in sequence, and finally outputs the classification result through a global average pooling layer and a fully connected layer. The construction stage can be the stage that a construction site is in during the construction process, for example, it can include seven stages: unstarted, earthwork excavation, pile foundation engineering, main construction, ground building roofing, outdoor construction, and project completion. In the unstarted stage, the image shows a natural surface or dust-proof net coverage and there are no trucks, excavators, or any other construction vehicles and equipment. In the earthwork excavation stage, there are excavators or trucks entering the site, there may be traces of excavation, and there are no buildings and foundations. In the pile foundation engineering stage, there are white squares or pile foundation holes on the surface, but there is no building height. In the main construction stage, there is building height, the roof color is inconsistent, and there must be a tower crane. In the ground building roofing stage, the roof color is consistent, and there is no greening and road. In the outdoor construction stage, there are traces of greening or road construction. In the project completion stage, the greening and roads have been completed, and there are no traces of construction. It should be noted that the above-mentioned construction stages are only used for illustration. In actual processes, those skilled in the art can flexibly adjust the construction stages. The scene classification model, the convolutional neural network model, and the residual network model are respectively trained using sample training data; the sample training data are obtained based on training remote sensing image label data; the training remote sensing image label data includes labels for characterizing construction stages and training remote sensing images corresponding to the construction stages. Taking the above-mentioned 7 construction stages as an example, the label for unstarted construction can be 1, the label for earth excavation can be 2, the label for pile foundation engineering can be 3, the label for main construction can be 4, the label for ground building completion can be 5, the label for outdoor construction can be 6, and the label for project completion can be 7.

[0061] Specifically, the coordinate information of the construction site to be identified and the satellite remote sensing image data can be respectively input into the pre-trained scene classification model to obtain a first probability set containing multiple construction stage probabilities. The coordinate information of the construction site to be identified and the satellite remote sensing image data can be input into the pre-trained convolutional neural network model to obtain a second probability set. The coordinate information of the construction site to be identified and the satellite remote sensing image data can be input into the pre-trained residual network model to obtain a third probability set.

[0062] S206: Determine the construction progress of the construction site to be identified based on the first probability set, the second probability set and the third probability set, each of which contains multiple construction stage probabilities.

[0063] Specifically, the current construction progress is determined according to the probability of each construction stage in the first probability set, the second probability set and the third probability set.

[0064] In the above construction progress identification method, by obtaining satellite remote sensing image data containing the construction site area to be identified, the overall situation of the area can be intuitively understood from a macro level, including the topography, traffic roads, other adjacent buildings, etc. around the construction site, providing rich background information and spatial references for subsequent in-depth analysis of the construction site itself, avoiding focusing only on the local area and ignoring the possible impact of the surrounding environment on the construction. The coordinate information of the construction site is determined based on satellite remote sensing image data, so that the construction site has a precise positioning in the geographic space. This not only facilitates the subsequent accurate marking of its location in geographic information systems such as maps, but also can be integrated and analyzed with other geographic data (such as urban planning data, infrastructure distribution data, etc.), which is conducive to the overall planning of regional construction resources and the evaluation of the impact of the construction site on the surrounding areas. The pre-trained scene classification model is usually trained based on a large amount of different scene image data. It can analyze the characteristics of the construction site to be identified from the perspective of the overall scene, and identify different scene modes presented in different construction stages, such as large-scale land leveling scenes in the infrastructure stage, towering scaffolding scenes in the main structure construction stage, and relatively regular facade scenes in the decoration stage, etc., to provide a basis for judging the construction progress based on the macroscopic scene level. CNN is good at automatically extracting deep features from image data, and can effectively capture the texture, shape, color and other detailed features of the construction site in satellite remote sensing images. For example, it can infer the construction stage based on the shape and distribution changes of the stacked building materials (which change as the construction stage progresses), the gradual clarity of the main building outline and other feature changes, and then output the probability of each construction stage, so that the judgment of the construction progress is more detailed and more accurate. ResNet can well solve the problem of gradient disappearance during deep network training, so that a deeper network structure can be constructed to deeply explore the potential features of different construction stages of construction sites in satellite remote sensing images and their inherent connections. For example, it can analyze the correlation between subtle changes in the internal structure of the building at different stages and changes in the appearance characteristics, and more comprehensively consider the various factors that affect the construction progress. The output construction stage probability is more valuable for reference. The three models can reduce the risk of overfitting of a single model. Moreover, the three models differ in structure, parameters and feature extraction methods, which can capture different information, improve generalization ability, and thus improve classification accuracy. Comprehensively considering the first probability set, the second probability set and the third probability set to determine the construction progress of the construction site to be identified is equivalent to integrating the analysis results of multiple models with different perspectives and different characteristics. Different models have their own advantages in analyzing image features and identifying construction stages. By integrating their probability outputs, we can learn from each other's strengths and weaknesses, so that the final construction progress is more in line with the actual situation, more scientific and objective.

[0065] In one embodiment, Figure 3As shown, determining the construction progress of the to-be-identified construction site according to the first probability set, the second probability set and the third probability set, each of which contains multiple construction stage probabilities, includes:

[0066] S302: Add the probabilities of each construction stage in the first probability set, the second probability set, and the third probability set to obtain a total probability of each construction stage.

[0067] S304, determining the target construction stage according to the maximum probability in the total probability of each construction stage.

[0068] S306: Determine the construction progress of the construction site to be identified according to the target construction stage.

[0069] In some exemplary embodiments, the above 7 construction stages are taken as an example for explanation.

[0070] Assume that the probability values ​​of the i-th construction site in the image based on the YOLOv8, ResNet50 and VGG16 models are , and , the voting weight is

[0071] (1)

[0072] (2)

[0073] (3)

[0074] (4)

[0075] x=0 indicates that the construction site is predicted to be in the unstarted stage, x=1 indicates that the construction site is predicted to be in the earthwork excavation stage, x=2 indicates that the construction site is in the pile foundation stage, x=3 indicates that the construction site is in the main construction stage, x=4 indicates that the construction site is in the ground building roof stage, x=5 indicates that the construction site is in the outdoor construction stage, and x=6 indicates that the construction site is in the project completion stage. Using the voting mechanism, the probability values ​​predicted by the three models are added together. It is the sum of the probability values ​​predicted by the three models. The final prediction result is determined based on the value. The final prediction result can be the construction stage. The current construction progress percentage is determined based on the construction stage.

[0076] In this embodiment, the probability of each construction stage in the first probability set, the second probability set and the third probability set output from the scene classification model, the convolutional neural network model and the residual network model is added together, so as to realize the effective integration of the information provided by multiple different models. Each model analyzes the satellite remote sensing image data from its own unique perspective and method, extracts the characteristics of the construction stage of the construction site and outputs the probability. Through this addition and aggregation method, the one-sidedness and limitations that may be generated by relying on a single model to judge are avoided, and the advantages of each model are fully gathered, so that the information used to judge the construction progress is more comprehensive and rich. The total probability is obtained by adding the multiple probabilities corresponding to each construction stage, which is a scientific processing method based on the principle of probability statistics. In probability theory, the comprehensive consideration of the probabilities of multiple independent events for the same result can more accurately reflect the possibility of the occurrence of the result. Here, the probability of the construction stage output by each model is equivalent to an independent event. Combining them to calculate the total probability is more in line with the logic of quantitative analysis of complex real situations, so that the total probability of each construction stage finally obtained can more reasonably reflect the possibility of the construction stage in actual situations.

[0077] In one embodiment of the device, Figure 4 As shown, the method also includes:

[0078] S402, obtaining training remote sensing image area data, wherein the training remote sensing image area data contains construction sites in multiple different construction stages.

[0079] Specifically, the training remote sensing image area data used for model training may contain construction sites in multiple different construction stages.

[0080] S404: Label the construction sites in different construction stages contained in the training remote sensing image area data.

[0081] Specifically, based on the actual construction situation of the construction site, the construction stage is accurately marked. The labeling personnel need to have certain knowledge of construction engineering and the ability to interpret remote sensing images. They can judge the construction stage by observing the appearance of the building in the image, the layout of the facilities on the site, the stacking of construction materials, etc., and assign the correct label.

[0082] S406, using the labeled training remote sensing image area data to respectively train the scene classification model, the convolutional neural network model and the residual network model.

[0083] Specifically, the YOLOv8 model is trained using the labeled training remote sensing image area data, Adam is selected as the optimizer, the cross entropy loss function is used as the loss function, the initial learning rate is 0.01, the number of training iterations is 100, and the batchsize is set to 16. The labeled training remote sensing image area data is divided into a sample set and a validation set and then input into YOLOv8 for training. The ResNet50 network model is trained using the labeled training remote sensing image area data, Adam is selected as the optimizer, the cross entropy loss function is used as the loss function, the initial learning rate is 0.001, the number of training iterations is 100, and the batchsize is set to 16. The labeled training remote sensing image area data is divided into a sample set and a validation set and then input into ResNet50 for training. The VGG16 network model is trained using the labeled training remote sensing image area data. The Adam optimizer is selected, the cross entropy loss function is used as the loss function, the initial learning rate is 0.01, the number of training iterations is 100, and the batch size is set to 16. The labeled training remote sensing image area data is divided into a sample set and a validation set and input into VGG16 for training.

[0084] In this embodiment, three different models are trained, and these three models are complementary to each other. VGG16 has stronger generalization ability than YOLOv8, but its classification effect on small target images is inferior to YOLOv8. ResNet50 has a different structure from VGG16, and the combination of the two can enhance the generalization ability and ensure the accuracy of the final recognition.

[0085] In one embodiment, Figure 5 As shown, the labeling of the construction sites in different construction stages in the training remote sensing image area data includes:

[0086] S502: Determine the construction stage of each construction site in the training remote sensing image according to predetermined landmark features of the construction stage.

[0087] S504, using different construction labels to label different construction stages.

[0088] S506: Divide the construction site in the training remote sensing image into a training set and a data set according to the construction stage and a preset division ratio.

[0089] Specifically, different construction stages have different corresponding features. Therefore, the construction stage of each construction site in the training remote sensing image can be determined based on the sign features that appear in each construction stage. Due to the different construction stages, the corresponding construction labels for each construction stage are also different. Different perspective labels are required to label different construction stages. The division ratio is preset according to actual needs and experience. The common practice is to divide it according to the ratio of 80% as the training set and 20% as the validation set. Of course, this ratio can also be adjusted according to the specific situation, such as 70% training set and 30% validation set.

[0090] In some exemplary embodiments, the training remote sensing images are cropped into slices of 512×512 pixels according to the center point of each construction site vector. The sample library for distinguishing the construction progress stages is obtained by dividing the sample library into 7 categories according to the 7 stage labels, and is divided into a training set and a validation set according to a ratio of 8:2.

[0091] In one embodiment, Figure 6 As shown, the method also includes:

[0092] S602, obtaining training remote sensing image area data, wherein the training remote sensing image area data contains construction sites in multiple different construction stages.

[0093] S604, the training remote sensing image area data contains construction sites in multiple different construction stages, and the construction start labels of the construction sites are determined according to the construction stages, wherein the construction start labels include construction-in-progress labels and not-construction-started labels.

[0094] S606: Train a target detection model based on the construction site's construction start label and the training remote sensing image area data.

[0095] The target detection model may be YOLOv8. The "under construction" label indicates that the project is under construction, and the "not started" label indicates that the project has not been started or has been completed.

[0096] Specifically, training remote sensing image area data can also be obtained. Since some construction stages have been completed or finished, the construction site's start label can be determined according to the construction stage. The construction site's start label and the training remote sensing image area data are used to train the target detection model.

[0097] In some exemplary embodiments, the training remote sensing image area data is set to 0 (unstarted label) according to the 7-stage labels of each construction site vector, and the stages marked as 1, 2, and 7 are not used as detection and positioning targets. All stages marked as 2, 3, 4, and 5 are changed to 1, indicating that they are under construction. The image and vector patch data are sliced ​​with reference to the format of the DOTA data set, and the image slices and label data are saved in tif and txt formats, respectively. Among them, the tif slice size is set to 1024×1024, and the overlapping pixels are 256. When the truncated target area is less than 50% of the original area, the label information corresponding to the target is discarded (to remove too many targets to ensure the accuracy of the final training model); the location information of the target in the txt file is represented by the coordinates of the four corner points, and a sample data set is obtained, which is divided into a training set and a validation set in an 8:2 ratio.

[0098] In one embodiment, determining the coordinate information of the construction site to be identified based on the satellite remote sensing image data includes:

[0099] The trained target detection model is used to perform target detection on the satellite remote sensing image data to determine the coordinate information of the construction site to be identified.

[0100] Specifically, first, the acquired satellite remote sensing image data is organized into a format acceptable to the target detection model. Common deep learning frameworks (such as TensorFlow, PyTorch, etc.) generally require the input data to be in a specific tensor format. For example, the original format of the image data (which may be common image formats such as GeoTIFF, JPEG, etc.) is read through the corresponding image processing library (such as the PIL library in Python, OpenCV library, etc.) and converted into a tensor with appropriate dimensions (such as [batch size, number of channels, height, width], for a single image, the batch size is usually 1, and the number of channels may be 3 for color images) to ensure that the data format matches the input requirements of the target detection model. Depending on the input size requirements of the target detection model and the computing resources, the original satellite remote sensing image may need to be cropped or scaled. If the image size is too large and exceeds the range that the model can handle, it can be cropped according to certain rules and divided into multiple sub-images of appropriate sizes (it should be noted that the integrity of the relevant information of the construction site should be guaranteed as much as possible during cropping); or it can be scaled proportionally to adjust the image size to meet the settings of the model input layer for the image height and width, such as scaling the original larger image to the fixed input size required by the model (such as 600×600 pixels, etc.), so that the subsequent model can perform efficient target detection operations. Input the preprocessed satellite remote sensing image data into the loaded target detection model and start the forward propagation process of the model. In this process, the model will perform a series of convolution, pooling, full connection and other operations on the input image in turn according to its internal network structure and parameter configuration, automatically extract feature information in the image, and identify and locate the target (construction site) based on these features. Different target detection models have different detection mechanisms. The YOLO series model divides the image into multiple grid cells, directly predicts the category and location information of the target in each cell, and outputs the detection results at one time. If the ultimate goal is to obtain the coordinate information of the construction site in the actual geographic space (rather than just the pixel coordinates in the image), a coordinate conversion operation is required. This requires prior knowledge of the geographic reference information of the satellite remote sensing image, such as the projection method of the image (such as the commonly used Gauss-Kruger projection, Mercator projection, etc.), the geographic coordinate system (such as the WGS84 coordinate system, etc.) and the resolution of the image (the actual ground distance represented by each pixel) and other parameters. By using professional geographic information system (GIS) software (such as ArcGIS) or related coordinate conversion algorithm libraries, the pixel coordinates in the image are converted into actual geographic coordinates (in the form of longitude, latitude, etc.) according to the above parameters, so as to accurately determine the actual geographic location coordinate information of the construction site on the earth.For example, given the geographic coordinate conversion relationship corresponding to a certain pixel coordinate of a known image and information such as the image resolution, the previously detected pixel coordinates of the construction site [100, 200, 300, 400] can be converted into actual geographic coordinates (such as longitude 116.3°, latitude 39.9° and other specific coordinate values) through corresponding calculations, thereby achieving accurate determination of the coordinate information of the construction site to be identified.

[0101] In one embodiment, after determining the coordinate information of the construction site to be identified, the method further includes:

[0102] The coordinate information of the construction site to be identified is processed using a non-maximum suppression (NMS) post-processing algorithm.

[0103] Among them, non-maximum suppression is a post-processing algorithm widely used in fields such as target detection. Its core purpose is to eliminate redundant detection frames with high overlap and relatively low confidence, and retain the most representative detection frames that are most likely to be the target, so as to obtain more accurate and concise target detection results.

[0104] Specifically, the coordinate information of the construction site to be identified may be processed using a non-maximum suppression (NMS) post-processing algorithm, thereby eliminating redundant monitoring frames and obtaining an optimized construction site positioning result.

[0105] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0106] Based on the same inventive concept, the disclosed embodiment also provides a construction progress identification device for implementing the construction progress identification method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more construction progress identification device embodiments provided below can refer to the limitations of the construction progress identification method above, and will not be repeated here.

[0107] In one embodiment, Figure 7As shown, a construction progress identification device 700 is provided, comprising: a data acquisition module 702, a model processing module 704 and a construction progress determination module 706, wherein:

[0108] The data acquisition module 702 is used to acquire satellite remote sensing image data of an area containing a construction site to be identified, and determine the coordinate information of the construction site to be identified based on the satellite remote sensing image data;

[0109] The model processing module 704 is used to obtain a first probability set, a second probability set and a third probability set each containing a plurality of construction stage probabilities based on the coordinate information of the construction site to be identified and the satellite remote sensing image data, respectively, using a pre-trained scene classification model, a pre-trained convolutional neural network model and a pre-trained residual network model;

[0110] The scene classification model, the convolutional neural network model and the residual network model are respectively trained using sample training data; the sample training data is obtained based on training remote sensing image label data; the training remote sensing image label data includes labels for characterizing construction stages and training remote sensing images corresponding to the construction stages;

[0111] The construction progress determination module 706 is used to determine the construction progress of the to-be-identified construction site according to the first probability set, the second probability set and the third probability set, each of which contains multiple construction stage probabilities.

[0112] In one embodiment of the device, the construction progress determination module 706 is also used to add the probabilities of each construction stage in the first probability set, the second probability set and the third probability set to obtain the total probability of each construction stage; determine the target construction stage according to the maximum probability in the total probability of each construction stage; and determine the construction progress of the construction site to be identified according to the target construction stage.

[0113] In one embodiment of the device, the device also includes: a model training module, used to obtain training remote sensing image area data, the training remote sensing image area data contains construction sites in multiple different construction stages; label the construction sites in multiple different construction stages contained in the training remote sensing image area data; and use the labeled training remote sensing image area data to respectively train the scene classification model, the convolutional neural network model and the residual network model.

[0114] In one embodiment of the device, the model training module includes: a data processing module, used to determine the construction stage of each construction site in the training remote sensing image according to the predetermined landmark features of the construction stage; label different construction stages with different construction labels; and divide the construction sites in the training remote sensing image into training sets and data sets according to the construction stages and a preset division ratio.

[0115] In one embodiment of the device, the model training module is also used to obtain training remote sensing image area data, wherein the training remote sensing image area data contains construction sites in multiple different construction stages; the construction site in the training remote sensing image area data contains construction sites in multiple different construction stages, and the construction start label of the construction site is determined according to the construction stage, and the construction start label includes an under-construction label and a not-started label; and the target detection model is trained based on the construction start label of the construction site and the training remote sensing image area data.

[0116] In one embodiment of the device, the data acquisition module 702 is further used to perform target detection on the satellite remote sensing image data using the trained target detection model to determine the coordinate information of the construction site to be identified.

[0117] In one embodiment of the device, the device further comprises: an algorithm processing module for processing the coordinate information of the construction site to be identified using a non-maximum suppression (NMS) post-processing algorithm.

[0118] Each module in the above-mentioned construction progress identification device can be implemented in whole or in part by software, hardware or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0119] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Figure 8 As shown. The computer device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store satellite remote sensing image data. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a construction progress identification method is implemented.

[0120] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig. 9 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a construction progress identification method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0121] Those skilled in the art will understand that Figure 8 The structure shown in or 9 is merely a block diagram of a partial structure related to the scheme of the present disclosure, and does not constitute a limitation on the computer device to which the scheme of the present disclosure is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0122] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0123] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0124] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0125] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties.

[0126] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided by the present disclosure can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided by the present disclosure may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided by the present disclosure may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, etc., but are not limited to this.

[0127] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0128] The above-described embodiments only express several implementation methods of the present disclosure, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present disclosure, and these all belong to the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure shall be subject to the attached claims.

Claims

1. A construction progress identification method, characterized in that: The method comprises: Acquire satellite remote sensing image data of an area containing a construction site to be identified, and determine coordinate information of the construction site to be identified based on the satellite remote sensing image data; Based on the coordinate information of the construction site to be identified and the satellite remote sensing image data, a pre-trained scene classification model, a pre-trained convolutional neural network model, and a pre-trained residual network model are used to obtain a first probability set, a second probability set, and a third probability set, each of which contains multiple construction stage probabilities; The scene classification model, the convolutional neural network model and the residual network model are respectively trained using sample training data; the sample training data is obtained based on training remote sensing image label data; the training remote sensing image label data includes labels for characterizing construction stages and training remote sensing images corresponding to the construction stages; The construction progress of the to-be-identified construction site is determined according to the first probability set, the second probability set and the third probability set, each of which contains multiple construction stage probabilities.

2. The method according to claim 1, characterized in that: Determining the construction progress of the to-be-identified construction site according to the first probability set, the second probability set, and the third probability set, each of which contains multiple construction stage probabilities, includes: Adding the probability of each construction stage in the first probability set, the second probability set and the third probability set to obtain a total probability of each construction stage; Determine the target construction stage according to the maximum probability of the total probability of each construction stage; The construction progress of the to-be-identified construction site is determined according to the target construction stage.

3. The method according to claim 1, characterized in that: The method further comprises: Acquiring training remote sensing image area data, wherein the training remote sensing image area data contains construction sites in multiple different construction stages; Labeling construction sites in multiple different construction stages contained in the training remote sensing image area data; The scene classification model, the convolutional neural network model and the residual network model are trained respectively using the labeled training remote sensing image area data.

4. The method according to claim 3, characterized in that: The labeling of the construction sites in different construction stages contained in the training remote sensing image area data includes: Determining the construction stage of each construction site in the training remote sensing image based on predetermined landmark features of the construction stage; Use different construction labels to label different construction stages; According to the construction stage and the preset division ratio, the construction site in the training remote sensing image is divided into a training set and a data set.

5. The method according to claim 1, characterized in that The method further comprises: Acquiring training remote sensing image area data, wherein the training remote sensing image area data contains construction sites in multiple different construction stages; The training remote sensing image area data contains construction sites in multiple different construction stages, and the construction start labels of the construction sites are determined according to the construction stages, wherein the construction start labels include construction start labels and non-construction start labels; The target detection model is trained based on the construction site's construction start label and the training remote sensing image area data.

6. The method according to claim 5, characterized in that The step of determining the coordinate information of the construction site to be identified based on the satellite remote sensing image data includes: The trained target detection model is used to perform target detection on the satellite remote sensing image data to determine the coordinate information of the construction site to be identified.

7. The method according to claim 5, characterized in that After determining the coordinate information of the construction site to be identified, the method further includes: The coordinate information of the construction site to be identified is processed using a non-maximum suppression (NMS) post-processing algorithm.

8. A construction progress identification device, characterized in that: The device comprises: A data acquisition module, used to acquire satellite remote sensing image data of an area containing a construction site to be identified, and determine coordinate information of the construction site to be identified based on the satellite remote sensing image data; A model processing module, for obtaining a first probability set, a second probability set, and a third probability set, each containing a plurality of construction stage probabilities, based on the coordinate information of the construction site to be identified and the satellite remote sensing image data, respectively through a pre-trained scene classification model, a pre-trained convolutional neural network model, and a pre-trained residual network model; The scene classification model, the convolutional neural network model and the residual network model are respectively trained using sample training data; the sample training data is obtained based on training remote sensing image label data; the training remote sensing image label data includes labels for characterizing construction stages and training remote sensing images corresponding to the construction stages; The construction progress determination module is used to determine the construction progress of the to-be-identified construction site based on the first probability set, the second probability set and the third probability set, each of which contains multiple construction stage probabilities.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

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

Images