CN114067205A - Light-weight arbitrary-scale double-time-phase image change detection method - Google Patents
Light-weight arbitrary-scale double-time-phase image change detection method Download PDFInfo
- Publication number
- CN114067205A CN114067205A CN202111335452.1A CN202111335452A CN114067205A CN 114067205 A CN114067205 A CN 114067205A CN 202111335452 A CN202111335452 A CN 202111335452A CN 114067205 A CN114067205 A CN 114067205A
- Authority
- CN
- China
- Prior art keywords
- change detection
- resolution
- image
- scale
- arbitrary
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000008859 change Effects 0.000 title claims abstract description 211
- 238000001514 detection method Methods 0.000 title claims abstract description 186
- 238000012549 training Methods 0.000 claims description 52
- 238000000034 method Methods 0.000 claims description 26
- 230000008569 process Effects 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 8
- 238000012952 Resampling Methods 0.000 claims description 5
- 238000007781 pre-processing Methods 0.000 claims description 5
- 238000003860 storage Methods 0.000 claims description 4
- 230000000007 visual effect Effects 0.000 claims description 4
- 238000010586 diagram Methods 0.000 description 6
- 238000005070 sampling Methods 0.000 description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 235000002566 Capsicum Nutrition 0.000 description 1
- 239000006002 Pepper Substances 0.000 description 1
- 235000016761 Piper aduncum Nutrition 0.000 description 1
- 235000017804 Piper guineense Nutrition 0.000 description 1
- 244000203593 Piper nigrum Species 0.000 description 1
- 235000008184 Piper nigrum Nutrition 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 150000003839 salts Chemical class 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4053—Scaling of whole images or parts thereof, e.g. expanding or contracting based on super-resolution, i.e. the output image resolution being higher than the sensor resolution
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
本发明公开了一种轻量型的任意尺度双时相影像变化检测方法，包括以下步骤：S1：获取现有的基于高分辨率遥感影像的变化检测数据集；同时获取待检测区域的双时相影像并构建多分辨率变化检测数据集；S2：构建轻量型的任意尺度变化检测模型；S3：利用所述的基于高分辨率遥感影像的变化检测数据集对轻量型的任意尺度变化检测模型的变化检测模块进行预训练；S4：利用多分辨率变化检测数据集对步骤S2构建的整个模型进行训练，得到训练好的轻量型的任意尺度变化检测模型；S5：将待检测区域的具有不同分辨率的双时相影像，输入在训练好的模型中，得到待检测区域的变化结果图。本发明实现了对任意尺度双时相影像的变化检测。
The invention discloses a lightweight arbitrary-scale dual-temporal image change detection method, comprising the following steps: S1: obtaining an existing high-resolution remote sensing image-based change detection data set; simultaneously obtaining a dual-temporal image of an area to be detected Image and build a multi-resolution change detection dataset; S2: Build a lightweight arbitrary-scale change detection model; S3: Use the high-resolution remote sensing image-based change detection dataset to detect lightweight arbitrary-scale changes The change detection module of the detection model is pre-trained; S4: Use the multi-resolution change detection data set to train the entire model constructed in step S2 to obtain a trained lightweight arbitrary-scale change detection model; S5: The area to be detected is The bi-temporal images with different resolutions are input into the trained model, and the result map of the change of the area to be detected is obtained. The invention realizes the change detection of bitemporal images of any scale.
Description
Technical Field
The invention relates to the field of remote sensing geographic information systems, in particular to a light-weight arbitrary-scale double-time-phase image change detection method.
Background
The change detection aims at identifying the surface change between the two time-phase images in the same region, and provides quantitative data for many important applications such as national resource investigation, ecological monitoring and protection, urban planning and the like. The remote sensing image has the characteristics of wide coverage range, rich information, high timeliness and the like, and is a main data source for extracting the ground feature information all the time. The high-resolution remote sensing image has more and more important functions in various cartography and application because of containing abundant space detail information.
In recent years, a method based on deep learning has become the most dominant solution for high-resolution image change detection. Given a double-time phase image with the same resolution, the methods effectively extract multi-level information in the image through convolution operation, and further mine change information. However, due to the limitation of remote sensing data acquisition, the change analysis is often required to be carried out by using double-phase images with different resolutions in practical production. To address the resolution difference of the change detection data, scholars have introduced sub-pixel localization (SPM) into the change detection. The SPM obtains a high-resolution change map by establishing a mapping relation between a previous time-phase high-resolution land cover map and a next coarse-resolution image. The method is widely used for solving the problem of resolution difference of Landsat-MODIS image change detection in the past, and achieves effective results. However, due to the intra-class heterogeneity and inter-class similarity, SPM-based methods are difficult to effectively apply to high-resolution remote sensing images. Therefore, research has recently been proposed to solve the problem of multi-scale change detection for high-resolution remote sensing images by using super-resolution. However, the traditional super-resolution model has large memory occupation, can only realize the super-resolution reconstruction with a specific scale, and is difficult to meet the diversified actual production needs.
In the prior art, the publication numbers are: the CN110969088A chinese invention patent, No. 4/7 in 2020, discloses a method for detecting changes in remote sensing images based on significance detection and a deep twin neural network, which comprises: preprocessing the two-time phase remote sensing image; carrying out normalization processing on the difference image; multi-scale segmentation and merging optimization; obtaining a significance detection graph; establishing a double-window depth twin convolution network model and training the model; and fusing the segmentation object and the pixel level change detection result through judgment to finally obtain a change detection result graph. The scheme can overcome the influence of salt and pepper noise to a certain extent, improves the detection precision, but cannot adapt to super-resolution of any scale, and the model is not light enough.
Disclosure of Invention
The invention provides a light-weight arbitrary-scale double-time-phase image change detection method for overcoming the defect that the conventional detection method cannot realize change detection of multi-scale remote sensing images.
The primary objective of the present invention is to solve the above technical problems, and the technical solution of the present invention is as follows:
a light-weight arbitrary-scale double-time-phase image change detection method comprises the following steps:
s1: acquiring a change detection data set based on the existing high-resolution remote sensing image; simultaneously acquiring double-temporal images of a region to be detected and constructing a multi-resolution change detection data set;
s2: constructing a light-weight arbitrary scale change detection model;
s3: pre-training a light-weight random scale change detection model by using the change detection data set based on the high-resolution remote sensing image;
s4: training the whole model constructed in the step S2 by using a multi-resolution change detection data set to obtain a trained light-weight arbitrary scale change detection model;
s5: and inputting the double time phase images with different resolutions of the region to be detected into the trained model to obtain a change result graph of the region to be detected.
Further, the specific process of acquiring the existing change detection dataset based on the high-resolution remote sensing image in step S1 is as follows:
s101: determining a data source of a change detection data set of the high-resolution remote sensing image;
s102: and selecting a data set meeting preset conditions according to the determined data source for downloading.
Further, the specific steps of obtaining the image of the area to be detected and constructing the multi-resolution change detection data set are as follows:
selecting a sample region in a region to be detected, and acquiring a clear and cloudless double-time-phase image with the same resolution in the sample region;
preprocessing and geographically registering the double-time images according to the acquired data description document of the double-time images;
marking the area with changed land coverage on the image of the sample area after geographic registration by visual interpretation, converting label vector data into raster data after marking is finished, collecting samples from the image of the sample area after geographic registration and a change mark, and constructing a change detection data set with the same resolution;
and (3) carrying out double-thrice down-sampling on the image of the later time phase of the double-time-phase image in the change detection data set with the same resolution ratio by N times to simulate a low-resolution image, so as to obtain a multi-resolution change detection data set.
Further, the lightweight arbitrary scale change detection model includes: the super-resolution module is used for resampling the low-resolution image to the size of a high-resolution image; and the change detection module adopts a weight-sharing feature extractor to extract multi-level features and carries out change detection.
Further, the pre-training of the light-weight arbitrary scale change detection model by using the change detection data set based on the high-resolution remote sensing image is a specific process of pre-training the change detection module of the light-weight arbitrary scale change detection model by using the change detection data set based on the high-resolution remote sensing image, which comprises the following steps:
adjusting training parameters, the training parameters including: training times, batch size, an optimizer, initial learning rate setting and a loss function;
and training a change detection module of the model based on the high-resolution remote sensing image change detection data set, and storing parameters of the model with the optimal precision.
Further, the specific steps of training the whole model constructed in step S2 by using the multiresolution change detection data set to obtain a trained lightweight arbitrary scale change detection model are as follows:
loading parameters of a pre-trained change detection module for initializing the model;
adjusting training parameters, wherein the training parameters comprise training times, batch size, an optimizer, initial learning rate setting and a loss function, training the whole model by using a multi-resolution change detection data set with N times of resolution difference, the super-resolution module is used for resampling a low-resolution image to a high-resolution image, and the super-resolution module and the change detection module set different initial learning rates;
and obtaining the light-weight arbitrary scale change detection model.
Further, the specific process of step S5 is: inputting the double time phase images with different resolutions of the region to be detected into the trained model to obtain a change result graph of the region to be detected, and combining the change result graphs to obtain a complete region change detection graph.
The invention also provides a light-weight arbitrary-scale double-time-phase image change detection system, which comprises: a memory and a processor, wherein the memory includes a lightweight arbitrary-scale double-temporal image change detection method program, and the lightweight arbitrary-scale double-temporal image change detection method program, when executed by the processor, implements the steps of:
s1: acquiring a change detection data set based on the existing high-resolution remote sensing image; simultaneously acquiring an image of a region to be detected and constructing a multi-resolution change detection data set;
s2: constructing a light-weight arbitrary scale change detection model;
s3: pre-training a light-weight random scale change detection model by using the change detection data set based on the high-resolution remote sensing image;
s4: training the whole model constructed in the step S2 by using a multi-resolution change detection data set to obtain a trained light-weight arbitrary scale change detection model;
s5: and inputting the double time phase images with different resolutions of the region to be detected into the trained model to obtain a change result graph of the region to be detected.
Further, the specific process of acquiring the existing change detection dataset based on the high-resolution remote sensing image in step S1 is as follows:
s101: determining a data source of a change detection data set of the high-resolution remote sensing image;
s102: and selecting a data set meeting preset conditions according to the determined data source for downloading.
The invention also provides a computer readable storage medium, which includes a lightweight arbitrary-scale double-temporal-phase image change detection method program, and when the lightweight arbitrary-scale double-temporal-phase image change detection method program is executed by a processor, the steps of the lightweight arbitrary-scale double-temporal-phase image change detection method are realized.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that:
the invention provides a light-weight arbitrary scale change detection method, which comprises the steps of constructing a light-weight arbitrary scale change detection model, overcoming the resolution difference between double-time-phase images by using a super-resolution module in the model, recovering a low-resolution image to the size of a high-resolution image, extracting multi-level features in the double-time-phase images by using a change detection module, and further realizing the change detection of the double-time images.
Drawings
Fig. 1 is a flow chart of a lightweight arbitrary-scale double-temporal-phase image change detection method according to the present invention.
FIG. 2 is a schematic diagram of change detection with 3.0 times of resolution difference in the embodiment of the present invention.
FIG. 3 is a schematic diagram of change detection with 4.0 times of resolution difference in the embodiment of the present invention.
FIG. 4 is a schematic diagram of change detection with 5.0 times of resolution difference in the embodiment of the present invention.
Detailed Description
In order that the above objects, features and advantages of the present invention can be more clearly understood, a more particular description of the invention will be rendered by reference to the appended drawings. It should be noted that the embodiments and features of the embodiments of the present application may be combined with each other without conflict.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, however, the present invention may be practiced in other ways than those specifically described herein, and therefore the scope of the present invention is not limited by the specific embodiments disclosed below.
Example 1
As shown in fig. 1, a lightweight arbitrary-scale two-time phase image change detection method includes the following steps:
s1: acquiring a change detection data set based on the existing high-resolution remote sensing image;
the specific process of step S1 is:
s101: determining a data source of a change detection data set of the high-resolution remote sensing image;
s102: and selecting a data set meeting preset conditions according to the determined data source for downloading.
It should be noted that, in the embodiment of the present invention, data sources may be arranged through existing data sites, such as ieee xplore, Github, and the like, and a data set with a relatively large data volume, a better data quality, and an image type closer to an image of a detection task is selected for downloading for pre-training of a subsequent model. In one embodiment, a CCD change detection data set comprising 16000 256-256 image pairs comprising 10000 training samples, 3000 verification samples and 3000 test samples with an image resolution of 0.03-1 m can be used. The method not only provides the change information of common objects such as buildings, roads, forests and the like, but also provides the change information of a plurality of detailed objects such as automobiles, tanks and the like.
Acquiring a double-temporal image of a region to be detected and constructing a multi-resolution change detection data set;
the method comprises the following specific steps:
selecting a sample region in a region to be detected, and acquiring a clear and cloudless double-time-phase image with the same resolution in the sample region;
preprocessing and geographically registering the double-time images according to the acquired data description document of the double-time images;
marking the area with changed land coverage on the image of the sample area after geographic registration by visual interpretation, converting label vector data into raster data after marking is finished, collecting samples from the image of the sample area after geographic registration and a change mark, and constructing a change detection data set with the same resolution;
and (3) carrying out double-thrice down-sampling on the image of the later time phase of the double-time-phase image in the change detection data set with the same resolution ratio by N times to simulate a low-resolution image, so as to obtain a multi-resolution change detection data set.
It should be noted that, firstly, remote sensing image data of an area to be predicted is acquired on an open website, if the resolution of the obtained image of the area to be predicted, which meets the time and area range required by the detection task, is inconsistent due to the limitation of the cloud amount and the re-returning period, that is, there is a resolution difference of N times, therefore, a sufficient sample area is selected in the area to be predicted, and a double-time-phase image with the same resolution and a clear and non-cloud sample area is downloaded on the website. Then, preprocessing and geographic registration are carried out on the double-time images according to the acquired data description documents of the double-time images;
the processing dimension of the pretreatment comprises: DN value of data, projection coordinate system, radiometric calibration, radiometric correction, mosaic, clipping, histogram matching, etc. Marking the area with changed land coverage on the image of the sample area after geographic registration by visual interpretation, converting label vector data into raster data after marking is finished, collecting samples from the image of the sample area after geographic registration and a change mark, and constructing a change detection data set with the same resolution; and (3) carrying out double-thrice down-sampling on the image of the later time phase of the double-time-phase image in the change detection data set with the same resolution ratio by N times to simulate a low-resolution image, so as to obtain a multi-resolution change detection data set. It should be noted that the constructed multiresolution change detection data set is used for model training, and simultaneously, the double time phase images of the to-be-detected region with the resolution difference of N times in the multiresolution change detection data set are cut for later use for subsequent detection.
In a specific embodiment, the multiresolution change detection dataset is constructed using a google earth image-based change detection dataset comprising 19 pairs of images with a spatial resolution of 0.55 meters, ranging in size from 1006 x 1168 to 4936 x 5224. All images were taken between 2006 and 2019. The data set provides a labeling of building changes between the image pairs. We first cut the dataset into 1118 pairs of 256 × 256 sized non-overlapping samples, as per 6: 2: and 2, dividing the proportion into a training set, a verification set and a test set, and randomly rotating the training set to realize data amplification to avoid overfitting, so that a double-time phase change detection data set with the same resolution ratio is obtained.
In order to verify the validity of the proposed model, a multi-resolution change detection data set needs to be further constructed. Therefore, for the obtained double time phase change detection data set, double-cubic downsampling can be adopted to process the image of the second time phase to obtain various low-resolution images with different scales, so that the problem of multi-scale change detection in the practical application process is simulated. In a specific implementation process, the down-sampling can be performed by setting three resolution differences of 3.0, 4.0 and 5.0.
S2: constructing a light-weight arbitrary scale change detection model;
it should be noted that, in the embodiment of the present invention, the lightweight arbitrary scale change detection model includes: the super-resolution module is used for resampling the low-resolution image to the size of a high-resolution image; more specifically, the super-resolution module includes: the method comprises two parts, namely a lightweight feature extractor and an over-scale reconstruction module (OSM). The lightweight feature extractor consists of a 3 x 3 convolutional layer and 3 sense Group modules. Each Dense Group comprises 3 residual modules which are fused with an SE attention mechanism, and the effective multiplexing of information is forced through a recursive structure, so that more effective characteristics are obtained. Based on the extracted features, the over-scale reconstruction module obtains an over-scale feature map by using two convolution layers of 3 multiplied by 3 and 5 Pixel Shuffle layers, and then obtains a high-resolution image with a target size through bicubic interpolation.
The over-scale reconstruction module is used for generating over-scale features so as to realize super-resolution of any scale and obtain a high-quality super-resolution image; the change detection module adopts a weight-sharing feature extractor to extract multi-level features and detect changes, and more specifically, the change detection module constructs a feature extractor based on ResNet-18, and is used for extracting multi-level features of an input image and learning a change map based on a feature vector of a depth measurement learning double-temporal image.
S3: pre-training a light-weight random scale change detection model by using the change detection data set based on the high-resolution remote sensing image;
the specific process of pre-training the change detection module of the lightweight arbitrary scale change detection model by using the change detection data set based on the high-resolution remote sensing image is as follows:
adjusting training parameters, the training parameters including: training times, batch size, an optimizer, initial learning rate setting and a loss function;
and training a change detection module of the model based on the high-resolution remote sensing image change detection data set, and storing parameters of the model with the optimal precision.
It should be noted that, in a specific implementation process, as it takes much time and labor cost to construct a new multi-resolution data set, the number of samples of the region to be predicted is often less, and direct training is likely to cause model overfitting. Therefore, the change detection module of the constructed model is pre-trained by using the change detection data set based on the high-resolution remote sensing image, so that the change information in the existing data set is fully utilized. In the pre-training, the following training parameters may be set, the number of training times is 200, the optimizer is Adam, the initial learning rate is 0.0001, and the training batch size is 8.
S4: training the whole model constructed in the step S2 by using a multi-resolution change detection data set to obtain a trained light-weight arbitrary scale change detection model;
the specific training process is as follows:
loading parameters of a pre-trained change detection module for initializing the model;
adjusting training parameters, wherein the training parameters comprise training times, batch size, an optimizer, initial learning rate setting and a loss function, training the whole model by using a multi-resolution change detection data set with N times of resolution difference, the super-resolution module is used for resampling a low-resolution image to a high-resolution image, and the super-resolution module and the change detection module set different initial learning rates;
and obtaining the light-weight arbitrary scale change detection model.
In a specific implementation, the training parameters are set as follows: the set training times is 100, the optimizer is Adam, the training batch size is 8, and the two modules are optimized by different loss functions separately to achieve a faster convergence effect. The initial learning rate of the super-resolution module is 0.001, and the change detection module is used for fine adjustment by adopting the learning rate of 0.0001 due to the fact that the pre-training parameters are loaded.
S5: and inputting the double time phase images with different resolutions of the region to be detected into the trained model to obtain a change result graph of the region to be detected.
It should be noted that, in the embodiment of the present invention, the two time-phase images with different resolutions of the region to be detected are input into the trained model to obtain the change result graph of the region to be detected, and the change result graphs are combined to obtain the complete region change detection graph. Fig. 2 is a schematic diagram showing a change detection with a resolution difference of 3.0 times, where (a) represents a time T1 image, (b) represents a time T2 image, (c) represents a time T2 low-resolution image, (d) represents a true value image, (e) represents a super-resolution image, and (f) represents a detection result.
Fig. 3 is a schematic diagram showing a change detection with a resolution difference of 4.0 times, in which (a) shows a time T1 image, (b) shows a time T2 image, (c) shows a time T2 low-resolution image, (d) shows a true value map, (e) shows a super-resolution image, and (f) shows a detection result.
Fig. 4 is a schematic diagram showing a change detection with a 5.0-fold difference in resolution, in which (a) shows a time T1 image, (b) shows a time T2 image, (c) shows a time T2 low-resolution image, (d) shows a true value map, (e) shows a super-resolution image, and (f) shows a detection result.
The invention also provides a light-weight arbitrary-scale double-time-phase image change detection system, which comprises: a memory and a processor, wherein the memory includes a lightweight arbitrary-scale double-temporal image change detection method program, and the lightweight arbitrary-scale double-temporal image change detection method program, when executed by the processor, implements the steps of:
s1: acquiring a change detection data set based on the existing high-resolution remote sensing image; simultaneously acquiring double-temporal images of a region to be detected and constructing a multi-resolution change detection data set;
s2: constructing a light-weight arbitrary scale change detection model;
s3: pre-training a light-weight random scale change detection model by using the change detection data set based on the high-resolution remote sensing image;
s4: training the whole model constructed in the step S3 by using a multi-resolution change detection data set to obtain a trained light-weight arbitrary scale change detection model;
s5: and inputting the double time phase images with different resolutions of the region to be detected into the trained model to obtain a change result graph of the region to be detected.
Further, the specific process of acquiring the existing change detection dataset based on the high-resolution remote sensing image in step S1 is as follows:
s101: determining a data source of a change detection data set of the high-resolution remote sensing image;
s102: and selecting a data set meeting preset conditions according to the determined data source for downloading.
The invention also provides a computer readable storage medium, which includes a lightweight arbitrary-scale double-temporal-phase image change detection method program, and when the lightweight arbitrary-scale double-temporal-phase image change detection method program is executed by a processor, the steps of the lightweight arbitrary-scale double-temporal-phase image change detection method are realized.
It should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.
Claims (10)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111335452.1A CN114067205A (en) | 2021-11-11 | 2021-11-11 | Light-weight arbitrary-scale double-time-phase image change detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111335452.1A CN114067205A (en) | 2021-11-11 | 2021-11-11 | Light-weight arbitrary-scale double-time-phase image change detection method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114067205A true CN114067205A (en) | 2022-02-18 |
Family
ID=80275277
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111335452.1A Pending CN114067205A (en) | 2021-11-11 | 2021-11-11 | Light-weight arbitrary-scale double-time-phase image change detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114067205A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115457259A (en) * | 2022-09-14 | 2022-12-09 | 华洋通信科技股份有限公司 | Image rapid saliency detection method based on multi-channel activation optimization |
| CN116051519A (en) * | 2023-02-02 | 2023-05-02 | 广东国地规划科技股份有限公司 | Method, device, equipment and storage medium for detecting double-time-phase image building change |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111666836A (en) * | 2020-05-22 | 2020-09-15 | 北京工业大学 | High-resolution remote sensing image target detection method of M-F-Y type lightweight convolutional neural network |
| CN112906638A (en) * | 2021-03-19 | 2021-06-04 | 中山大学 | Remote sensing change detection method based on multi-level supervision and depth measurement learning |
| CN113569760A (en) * | 2021-07-29 | 2021-10-29 | 重庆交通大学 | 3D Change Detection Method Based on Multimodal Deep Learning |
-
2021
- 2021-11-11 CN CN202111335452.1A patent/CN114067205A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111666836A (en) * | 2020-05-22 | 2020-09-15 | 北京工业大学 | High-resolution remote sensing image target detection method of M-F-Y type lightweight convolutional neural network |
| CN112906638A (en) * | 2021-03-19 | 2021-06-04 | 中山大学 | Remote sensing change detection method based on multi-level supervision and depth measurement learning |
| CN113569760A (en) * | 2021-07-29 | 2021-10-29 | 重庆交通大学 | 3D Change Detection Method Based on Multimodal Deep Learning |
Non-Patent Citations (1)
| Title |
|---|
| PARICHEHR BEHJATI: "OverNet: Lightweight Multi-Scale Super-Resolution with Overscaling Network", ARXIV, 9 November 2020 (2020-11-09), pages 3 - 4 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115457259A (en) * | 2022-09-14 | 2022-12-09 | 华洋通信科技股份有限公司 | Image rapid saliency detection method based on multi-channel activation optimization |
| CN115457259B (en) * | 2022-09-14 | 2023-10-31 | 华洋通信科技股份有限公司 | Image rapid saliency detection method based on multichannel activation optimization |
| CN116051519A (en) * | 2023-02-02 | 2023-05-02 | 广东国地规划科技股份有限公司 | Method, device, equipment and storage medium for detecting double-time-phase image building change |
| CN116051519B (en) * | 2023-02-02 | 2023-08-22 | 广东国地规划科技股份有限公司 | Method, device, equipment and storage medium for detecting double-time-phase image building change |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112949549B (en) | Super-resolution-based change detection method for multi-resolution remote sensing image | |
| US11403846B2 (en) | Crop boundary detection in images | |
| CN109840553B (en) | Extraction method and system of cultivated land crop type, storage medium and electronic equipment | |
| Cheng et al. | A spatial and temporal nonlocal filter-based data fusion method | |
| Samal et al. | Monitoring land use changes associated with urbanization: An object based image analysis approach | |
| Jin et al. | A super-resolution mapping method using local indicator variograms | |
| CN110929607A (en) | Remote sensing identification method and system for urban building construction progress | |
| Li et al. | Geospatial technology for earth observation | |
| Xu et al. | A sub-pixel mapping method based on an attraction model for multiple shifted remotely sensed images | |
| CN112906638A (en) | Remote sensing change detection method based on multi-level supervision and depth measurement learning | |
| Mercer et al. | Ultrahigh‐resolution mapping of peatland microform using ground‐based structure from motion with multiview stereo | |
| CN110263716B (en) | A method for super-resolution land cover mapping of remote sensing images based on street view images | |
| CN114067205A (en) | Light-weight arbitrary-scale double-time-phase image change detection method | |
| CN111666909A (en) | Suspected contaminated site space identification method based on object-oriented and deep learning | |
| Lauko et al. | Local color and morphological image feature based vegetation identification and its application to human environment street view vegetation mapping, or how green is our county? | |
| CN112906662A (en) | Method, device and equipment for detecting change of remote sensing image and storage medium | |
| CN113378642B (en) | A method for detecting illegally occupied buildings in rural areas | |
| Liu et al. | Robust multi-scale image matching for deriving ice surface velocity field from sequential satellite images | |
| CN117788822A (en) | Farmland boundary positioning information extraction method based on UAV low-altitude remote sensing images | |
| Xiong et al. | Fusing Landsat-7, Landsat-8 and Sentinel-2 surface reflectance to generate dense time series images with 10m spatial resolution | |
| He et al. | Deep learning approaches to spatial downscaling of GRACE terrestrial water storage products using EALCO model over Canada | |
| Crivellari et al. | Super-resolution GANs for upscaling unplanned urban settlements from remote sensing satellite imagery–the case of Chinese urban village detection | |
| Price et al. | Global GDP prediction with night-lights and transfer learning | |
| López-Caloca et al. | Analyzing short term spatial and temporal dynamics of water presence at a basin-scale in Mexico using SAR data | |
| Wang et al. | Investigating the influence of registration errors on the patch-based spatio-temporal fusion method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |