CN111047551B - Remote sensing image change detection method and system based on U-net improved algorithm - Google Patents
Remote sensing image change detection method and system based on U-net improved algorithm Download PDFInfo
- Publication number
- CN111047551B CN111047551B CN201911073929.6A CN201911073929A CN111047551B CN 111047551 B CN111047551 B CN 111047551B CN 201911073929 A CN201911073929 A CN 201911073929A CN 111047551 B CN111047551 B CN 111047551B
- Authority
- CN
- China
- Prior art keywords
- remote sensing
- image
- network
- input image
- change detection
- 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.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 59
- 230000008859 change Effects 0.000 title claims abstract description 57
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 25
- 238000000034 method Methods 0.000 claims abstract description 37
- 238000004458 analytical method Methods 0.000 claims abstract description 24
- 230000006872 improvement Effects 0.000 claims abstract description 14
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 230000002708 enhancing effect Effects 0.000 claims abstract description 7
- 101100295091 Arabidopsis thaliana NUDT14 gene Proteins 0.000 claims abstract 7
- 238000012545 processing Methods 0.000 claims description 38
- 238000000605 extraction Methods 0.000 claims description 9
- 238000011176 pooling Methods 0.000 claims description 8
- 238000005070 sampling Methods 0.000 claims description 7
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 claims description 6
- 238000013461 design Methods 0.000 claims description 6
- 230000004927 fusion Effects 0.000 claims description 6
- 230000010354 integration Effects 0.000 claims description 3
- 230000007246 mechanism Effects 0.000 abstract description 3
- 230000011218 segmentation Effects 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 9
- 238000013135 deep learning Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 8
- 238000004590 computer program Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 238000004880 explosion Methods 0.000 description 4
- 230000009467 reduction Effects 0.000 description 4
- 238000012549 training Methods 0.000 description 4
- 230000001419 dependent effect Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000008034 disappearance Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013529 biological neural network Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 230000036039 immunity Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- 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
- G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
- G06T3/4038—Scaling the whole image or part thereof for image mosaicing, i.e. plane images composed of plane sub-images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration by the use of more than one image, e.g. averaging, subtraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
Abstract
The invention provides a remote sensing image change detection method and a remote sensing image change detection system based on a U-net improvement algorithm, wherein the method comprises the following steps: preprocessing the remote sensing images, and integrating the preprocessed remote sensing images with different time phases to obtain an input image; based on the improved U-net network, carrying out downsampling encoding and upsampling decoding on the input image; and finally, outputting a binary image of whether the image is changed or not according to the analysis result. According to the invention, the network structure in the semantic segmentation direction is applied to the field of change detection, and a residual error learning mechanism is introduced, so that the encoder can quickly converge and deepen the network layer number, and meanwhile, the ASPP is used for enhancing the perceptibility of the network to the image characteristics, so that the algorithm can be kept at a higher level in terms of precision and efficiency, and meanwhile, the method has stronger robustness. The method is applicable to the field of change detection of remote sensing images, can be popularized to other fields, and has important significance.
Description
Technical Field
The invention relates to the technical field of computer vision, in particular to a remote sensing image change detection method and system based on a U-net improved algorithm.
Background
The change detection work of the remote sensing image is to quantitatively analyze the remote sensing image of different time phases of the same area by the pointer, thereby determining the form and the characteristics of the earth surface in the image and detecting the changed part and the unchanged part of the area at different times. The remote sensing image change detection has a wide role in the land resource control direction, and is widely applied to the detection of the direction of building violations, natural disaster assessment, urban construction planning and the like.
With the continuous development of satellite technology, the remote sensing image at the present stage has the characteristics of short cruising period, huge data volume and the like, and the method for detecting the regional variation by using manual contrast analysis cannot meet the requirements. How to extract the change information of the corresponding area from the remote sensing image rapidly and accurately is one of the hot spot problems in the application field of the remote sensing image at the present stage.
On the other hand, with the rapid development of machine computing capability and the widespread improvement of data quality in recent years, deep learning is becoming a popular direction in the field of computer machine learning, and has achieved very good results in the directions of computer vision, natural language processing and the like. The deep learning principle is to simulate a biological neural network and build a network structure consisting of neurons connected with each other. When inputting data such as images and texts to the network, the network can coordinate among network layers, extract deep features in the data for analysis, and finally feed back interpretation of the data.
The method is characterized in that the method is superior to the deep learning method in computer vision, the remote sensing image change detection is also started from the traditional method of constructing differential images and then carrying out threshold division, the method is combined with the deep learning algorithm, the nonlinear change characteristics in different time-phase satellite images are learned by using a deep neural network, and the final change detection result is determined. At present, the deep learning idea provides a new idea for satellite image processing, and a change detection algorithm based on deep learning becomes a new hot spot in the field of satellite image change detection.
The main current change detection methods at the present stage mainly comprise two types: (1) the pixel-oriented change detection is characterized in that the pixel is taken as a basic unit, the pixel is analyzed according to the spectrum and the field characteristics of the pixel, and the change detection problem of the remote sensing image is converted into the two classification problems of target change and target unchanged. The method uses each pixel as a unit to operate, so that a lot of noise information is easy to obtain, and the accuracy is influenced. (2) Object-oriented change detection is improved in a pixel-by-pixel method by first clustering a plurality of pixels to form an object and then classifying the object. And then comparing the classification conditions of the images with different time phases to obtain a detection result. Such methods have a cleaner picture and less noise than the first method, and have the disadvantage of being too dependent on the classification accuracy of the object class.
One such scheme is based on the detection of changes in the stack noise reduction auto-encoder network. The scheme belongs to a pixel-oriented change detection technology, and comprises the steps of firstly extracting field characteristics of a remote sensing image, inputting the field characteristics of each pixel point into a stack noise reduction automatic encoder, training layer by layer, and performing supervised fine adjustment until training is completed. In the existing change detection technology, a proper neighborhood range is selected, and the method can achieve a good change detection effect on the basis of reducing noise. However, the method is very dependent on selection of a neighborhood range in practical application, and if the selection is improper, false leakage detection is easy to occur. Meanwhile, the network structure is relatively simple, the number of layers is small, so that the effect is better than that of the traditional method, but a new breakthrough is difficult.
In summary, in the existing mainstream methods for detecting the change of the remote sensing image, the two methods mainly include classification, comparison and analysis, where the former corresponds to the object-oriented change detection, and the latter corresponds to the pixel-oriented change detection. The existing remote sensing image change detection methods have some defects.
The object-oriented change detection method can take texture features, shape features, neighborhood relations among objects and other information of the objects as basis of change detection, but classifies the pixel points after clustering, so that the accuracy of ground classification greatly influences the accuracy of change detection, and serious error accumulation problems exist. Meanwhile, the method has higher requirements on the quality of the remote sensing image, and the method can meet the condition that the invariable ground type presents similar characteristics and the invariable ground type presents different characteristics as far as possible. The problem faced by the pixel-oriented variation detection method is that the processing result is noisy and is excessively dependent on threshold selection.
Disclosure of Invention
The invention aims to solve the technical problems of low detection precision, poor noise resistance and the like of the existing change detection method by providing a remote sensing image change detection method and a remote sensing image change detection system based on a U-net improved algorithm.
In order to solve the technical problems, the invention provides a remote sensing image change detection method based on a U-net improvement algorithm, which comprises the following steps:
preprocessing a remote sensing image to be detected, and integrating the preprocessed remote sensing images in different time phases to obtain an input image;
based on an improved U-net network, carrying out downsampling encoding and upsampling decoding on the input image to finish analysis processing on the input image; the improved U-net network adopts a Resnet-101 network to extract data characteristics in an encoder part, and a hole space pyramid pooling design ASPP is added between the encoder and the decoder for extracting data multi-scale context information;
based on the analysis processing result of the input image, outputting a binary image of whether the remote sensing image to be detected changes or not.
Further, the preprocessing of the remote sensing image to be detected includes:
performing enhancement operation on the remote sensing images to be detected, and simultaneously processing the remote sensing images with different time phases by using histogram matching; wherein the enhancing operation includes flipping, zooming in or zooming out.
Further, the integrating the preprocessed remote sensing images in different time phases includes:
splicing the preprocessed two remote sensing images with different time phases to form an input image with 8+2 channels; wherein the 8 channels are four-channel images of two time phases, and the 2 channels are label channels which are changed or not.
Further, the processing method for performing downsampling encoding and upsampling decoding on the input image based on the improved U-net network to complete analysis processing on the input image comprises the following steps:
downsampling feature extraction is carried out on the input image by utilizing a Resnet-101 network;
carrying out feature coding on the feature information extracted by the Resnet-101 network by using ASPP;
up-sampling the encoded data while connecting the decoder portion for scale fusion.
Further, the Resnet-101 network includes a plurality of residual blocks, each residual block including a three-layer convolution.
Further, the ASPP includes one 1*1 convolution kernel and three 3*3 convolution kernels, as well as image level features.
Further, the decoder partial loss function in the improved U-net network is as follows:
wherein LT is the probability of label accuracy, PT is the probability of prediction accuracy, w is the network weight, and the value range of Loss function of Loss is [0,1 ].
Correspondingly, in order to solve the technical problems, the invention also provides a remote sensing image change detection system based on the U-net improvement algorithm, which comprises:
the image processing module is used for preprocessing the remote sensing image to be detected, and integrating the preprocessed remote sensing images in different time phases to obtain an input image;
the analysis processing module is used for carrying out downsampling encoding and upsampling decoding on the input image based on the improved U-net network to finish analysis processing on the input image; the improved U-net network encoder adopts a Resnet-101 network to extract data characteristics, and a hole space pyramid pooling design ASPP is added between the encoder and the decoder for extracting data multi-scale context information;
and the result output module is used for outputting a binary image of whether the remote sensing image to be detected changes or not based on the analysis processing result of the input image.
Further, the image processing module includes:
the image preprocessing unit is used for performing enhancement operation on the remote sensing images to be detected and processing the remote sensing images with different time phases by using histogram matching; wherein the enhancing operation includes flipping, zooming in or zooming out;
the image integration unit is used for splicing the preprocessed two remote sensing images with different time phases to form an 8+2 channel input image; wherein the 8 channels are four-channel images of two time phases, and the 2 channels are label channels which are changed or not.
Further, the analysis processing module includes:
a Resnet-101 network unit, configured to perform downsampling feature extraction on the input image by using a Resnet-101 network;
the ASPP unit is used for carrying out feature coding on the feature information extracted by the Resnet-101 network unit by using ASPP;
and the up-sampling decoding unit is used for up-sampling the encoded data and is connected with the decoder part for scale fusion.
The technical scheme of the invention has the following beneficial effects:
according to the invention, the network structure in the semantic segmentation direction is applied to the field of change detection, and a residual error learning mechanism is introduced, so that the encoder can quickly converge and deepen the network layer number, the feature extraction speed is accelerated, and the problems of gradient disappearance and gradient explosion are solved; meanwhile, ASPP is used for enhancing the perceptibility of the network to the image characteristics, so that the accuracy of the algorithm is ensured, and the efficiency is higher than that of other deep learning methods; the algorithm can be kept at a higher level in precision and efficiency, and has higher precision and stronger robustness. The method is suitable for the field of change detection of remote sensing images, can be popularized to other fields, and has important significance.
Drawings
FIG. 1 is a flowchart of a remote sensing image change detection method based on a U-net improvement algorithm of the present invention;
FIG. 2 is a network structure diagram of the remote sensing image change detection method of the present invention;
fig. 3 is a network structure diagram of a network of Resnet-101.
Detailed Description
In order to make the technical problems, technical solutions and advantages to be solved more apparent, the following detailed description will be given with reference to the accompanying drawings and specific embodiments.
First embodiment
Referring to fig. 1 to 3, the present embodiment provides a remote sensing image change detection method based on a U-net improvement algorithm, as shown in fig. 1, the remote sensing image change detection method includes:
s101, preprocessing a remote sensing image to be detected, and integrating the preprocessed remote sensing images in different time phases to obtain an input image;
it should be noted that, the process of preprocessing the remote sensing image to be detected in this embodiment includes:
performing enhancement operation on the remote sensing images to be detected, and simultaneously processing the remote sensing images with different time phases by using histogram matching; the enhancement operation includes flipping, zooming in or zooming out, and the like.
The process for integrating the preprocessed remote sensing images in different time phases comprises the following steps:
splicing the preprocessed two remote sensing images with different time phases to form an input image with 8+2 channels; wherein the 8 channels are four-channel images of two time phases, and the 2 channels are label channels which are changed or not. That is, the present embodiment integrates two images of different phases, forming an eight-channel image as an input of the algorithm. And meanwhile, carrying out one-hot coding treatment on the label to enable the label to become label data of two channels.
S102, performing downsampling encoding and upsampling decoding on the input image based on an improved U-net network to finish analysis processing on the input image;
it should be noted that, in order to overcome the problems of simple network structure and shallow hierarchy of the stack noise reduction automatic encoder network in the present stage, the embodiment deepens the network structure as much as possible while ensuring the efficiency, thereby improving the change detection precision; the embodiment improves the U-net network, the U-net network framework can combine the advantages of a plurality of networks at present, so that a network with reasonable depth and deep features can be built, the U-net network framework is mainly divided into an encoder and a decoder, the encoder is used for extracting the features of data, and the decoder is used for analyzing the extracted features.
The embodiment improves the U-net network, so that the U-net network can be suitable for the problem of remote sensing image change detection, the accuracy of remote sensing image change detection is effectively improved while the calculation efficiency is ensured, and the structure is shown in fig. 2.
Specifically, after the shortcomings of the existing method are analyzed, the number of network layers is selected to be deepened in order to improve the accuracy, in order to avoid the problems of gradient elimination and gradient explosion caused by deepening of the number of network layers, the Resnet-101 network is adopted to replace an encoder part in the U-net network to perform data feature extraction, the Resnet-101 network depth is reasonable, the thought of residual error learning is introduced, the problems of gradient explosion or incapability of training and the like caused by over-deep network can be avoided, and training of the neural network can be accelerated very quickly. The Resnet-101 network is composed of a plurality of residual blocks, each of which is composed of three layers of convolutions, the structure of which is shown in FIG. 3. Inputting an image, completing convolution and ReLU operation, and obtaining residual mapping definition as follows:
F(x,{W i })＝W 2 δ(W 1 x i ) (1)
in equation (1), x i Is the input vector of the residual block, delta represents the ReLU operation, W i Is a connection matrix of matching size. At the same time, inputting pictures into the block connection to obtain an identity map W s x i . Inputting the remaining mapping and identity mapping to element addition and obtaining the remaining block is defined as:
x i+1 ＝F(x,{W i })+W s x i (2)
in equation (2), when x i When the size is inconsistent with F, W is used s To match, wherein F (x, { W) i }) may represent multiple convolutional layers.
Meanwhile, considering that each pixel neighborhood information has great influence on change detection, the embodiment adds a spatial pyramid pooling design (ASPP) of hole convolution between the encoder and the decoder for extracting the data multi-scale context information, wherein the ASPP can solve the problem of resolution reduction caused by DCNNs, does not increase new learning parameters, and reduces the loss of detail information while expanding receptive fields. Among them, ASPP includes one 1*1 convolution kernel and three 3*3 convolution kernels, as well as features at the image level. The expansion coefficients of the three 3*3 convolution kernels are 6, 12 and 18, respectively.
In this embodiment, the decoder partial loss function is as follows:
in equation (3), LT is the probability of label accuracy, PT is the probability of prediction accuracy, and w is the network weight. The value range of the Loss function of Loss is [0,1 ].
The U-net improved network of the embodiment has proper depth in the encoder part, can accurately extract the characteristics of an input image, has strong noise immunity, and can avoid the condition of more noise information of pixel-oriented change detection to a great extent. And the decoder employs hole space pyramid pooling (ASPP, atrous Spatial Pyramid Pooling) for extracting multi-scale context information. The rest part of the decoder adopts a U-net network with simple structure. The algorithm has stronger robustness, so the scheme of the embodiment not only can obtain better effect in the change detection, but also has higher efficiency.
Based on the above, in this embodiment, the step S102 specifically includes the following steps:
downsampling feature extraction is carried out on the input image by utilizing a Resnet-101 network;
carrying out feature coding on the feature information extracted by the Resnet-101 network by using ASPP;
upsampling the encoded data while connecting the decoder portion for scale fusion; the decoder is connected here in a jump connection;
s103, based on the analysis processing result of the input image, outputting a binary image of whether the remote sensing image to be detected changes.
It should be noted that, in this embodiment, the output of S103 is specifically an image with a 0-1 label, so as to indicate whether the remote sensing image to be detected changes, and complete the change detection of the remote sensing image.
According to the embodiment, the network structure in the semantic segmentation direction is applied to the field of change detection, and a residual error learning mechanism is introduced, so that the encoder can quickly converge and deepen the number of network layers, the feature extraction speed is increased, and the problems of gradient disappearance and gradient explosion are solved; meanwhile, ASPP is used for enhancing the perceptibility of the network to the image characteristics, so that the accuracy of the algorithm is ensured, and the efficiency is higher than that of other deep learning methods; the algorithm can be kept at a higher level in precision and efficiency, and has higher precision and stronger robustness. The method is suitable for the field of change detection of remote sensing images, can be popularized to other fields, and has important significance.
Second embodiment
The embodiment provides a remote sensing image change detection system based on a U-net improvement algorithm, which comprises:
the image processing module is used for preprocessing the remote sensing image to be detected, and integrating the preprocessed remote sensing images in different time phases to obtain an input image;
the analysis processing module is used for carrying out downsampling encoding and upsampling decoding on the input image based on the improved U-net network to finish analysis processing on the input image; the improved U-net network encoder adopts a Resnet-101 network to extract data characteristics, and a hole space pyramid pooling design ASPP is added between the encoder and the decoder for extracting data multi-scale context information;
and the result output module is used for outputting a binary image of whether the remote sensing image to be detected changes or not based on the analysis processing result of the input image.
Further, the image processing module includes:
the image preprocessing unit is used for performing enhancement operations such as overturning, zooming in and out and the like on the remote sensing image to be detected, and processing the remote sensing images with different time phases by using histogram matching;
the image integration unit is used for splicing the preprocessed two remote sensing images with different time phases to form an 8+2 channel input image; wherein the 8 channels are four-channel images of two time phases, and the 2 channels are label channels which are changed or not.
Further, the analysis processing module includes:
a Resnet-101 network unit, configured to perform downsampling feature extraction on an input image by using a Resnet-101 network;
the ASPP unit is used for carrying out feature coding on the feature information extracted by the Resnet-101 network unit by using ASPP;
and the up-sampling decoding unit is used for up-sampling the encoded data and is connected with the decoder part for scale fusion.
The remote sensing image change detection system based on the U-net improvement algorithm of the embodiment corresponds to the remote sensing image change detection method based on the U-net improvement algorithm of the first embodiment; the functions realized by the functional modules in the remote sensing image change detection system based on the U-net improvement algorithm are in one-to-one correspondence with the flow steps in the remote sensing image change detection method based on the U-net improvement algorithm in the first embodiment; therefore, the description is omitted here.
Furthermore, it should be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, apparatus, or computer program product. Accordingly, embodiments of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, embodiments of the invention may take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
Embodiments of the present invention are described with reference to flowchart illustrations and/or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It should also be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or terminal. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or terminal device comprising the element.
While the preferred embodiments of the present invention have been described, it should be noted that once the basic inventive concepts of the present invention are known, those skilled in the art can make several improvements and modifications without departing from the principles of the present invention, which are also considered as the scope of the present invention; that is, the appended claims are intended to be construed to include preferred embodiments and all such variations and modifications as fall within the scope of the embodiments of the invention.
Claims (2)
1. The remote sensing image change detection method based on the U-net improvement algorithm is characterized by comprising the following steps of:
preprocessing a remote sensing image to be detected, and integrating the preprocessed remote sensing images in different time phases to obtain an input image;
based on an improved U-net network, carrying out downsampling encoding and upsampling decoding on the input image to finish analysis processing on the input image; the improved U-net network adopts a Resnet-101 network to extract data characteristics in an encoder part, and a hole space pyramid pooling design ASPP is added between the encoder and the decoder for extracting data multi-scale context information;
based on the analysis processing result of the input image, outputting a binary image of whether the remote sensing image to be detected changes or not;
the preprocessing of the remote sensing image to be detected comprises the following steps:
performing enhancement operation on the remote sensing images to be detected, and simultaneously processing the remote sensing images with different time phases by using histogram matching; wherein the enhancing operation includes flipping, zooming in or zooming out;
the integrating the preprocessed remote sensing images in different time phases comprises the following steps:
splicing the preprocessed two remote sensing images with different time phases to form an input image with 8+2 channels; wherein the 8 channels are four-channel images of two time phases, and the 2 channels are label channels which are changed or not;
the improved U-net network-based downsampling encoding and upsampling decoding are carried out on the input image to complete the analysis processing of the input image, and the method comprises the following steps:
downsampling feature extraction is carried out on the input image by utilizing a Resnet-101 network;
carrying out feature coding on the feature information extracted by the Resnet-101 network by using ASPP;
upsampling the encoded data while connecting the decoder portion for scale fusion;
the Resnet-101 network comprises a plurality of residual blocks, each residual block comprising a three-layer convolution;
the ASPP includes one 1*1 convolution kernel and three 3*3 convolution kernels, as well as image level features;
the decoder partial loss function in the improved U-net network is as follows:
wherein LT is the probability of label accuracy, PT is the probability of prediction accuracy, w is the network weight, and the value range of Loss function of Loss is [0,1 ].
2. The remote sensing image change detection system based on the U-net improvement algorithm is characterized by comprising:
the image processing module is used for preprocessing the remote sensing image to be detected, and integrating the preprocessed remote sensing images in different time phases to obtain an input image;
the analysis processing module is used for carrying out downsampling encoding and upsampling decoding on the input image based on the improved U-net network to finish analysis processing on the input image; the improved U-net network encoder adopts a Resnet-101 network to extract data characteristics, and a hole space pyramid pooling design ASPP is added between the encoder and the decoder for extracting data multi-scale context information;
the result output module is used for outputting a binary image of whether the remote sensing image to be detected changes or not based on the analysis processing result of the input image;
the image processing module comprises:
the image preprocessing unit is used for performing enhancement operation on the remote sensing images to be detected and processing the remote sensing images with different time phases by using histogram matching; wherein the enhancing operation includes flipping, zooming in or zooming out;
the image integration unit is used for splicing the preprocessed two remote sensing images with different time phases to form an 8+2 channel input image; wherein the 8 channels are four-channel images of two time phases, and the 2 channels are label channels which are changed or not;
the analysis processing module comprises:
a Resnet-101 network unit, configured to perform downsampling feature extraction on the input image by using a Resnet-101 network;
the ASPP unit is used for carrying out feature coding on the feature information extracted by the Resnet-101 network unit by using ASPP;
and the up-sampling decoding unit is used for up-sampling the encoded data and is connected with the decoder part for scale fusion.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911073929.6A CN111047551B (en) | 2019-11-06 | 2019-11-06 | Remote sensing image change detection method and system based on U-net improved algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911073929.6A CN111047551B (en) | 2019-11-06 | 2019-11-06 | Remote sensing image change detection method and system based on U-net improved algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111047551A CN111047551A (en) | 2020-04-21 |
| CN111047551B true CN111047551B (en) | 2023-10-31 |
Family
ID=70231899
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911073929.6A Active CN111047551B (en) | 2019-11-06 | 2019-11-06 | Remote sensing image change detection method and system based on U-net improved algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111047551B (en) |
Families Citing this family (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111563906A (en) * | 2020-05-07 | 2020-08-21 | 南开大学 | Knee joint magnetic resonance image automatic segmentation method based on deep convolutional neural network |
| CN111611968B (en) * | 2020-05-29 | 2022-02-01 | 中国科学院西北生态环境资源研究院 | Processing method of remote sensing image and remote sensing image processing model |
| CN111986099B (en) * | 2020-06-30 | 2022-05-13 | 武汉大学 | Tillage monitoring method and system based on convolutional neural network with residual error correction fused |
| CN111860233B (en) * | 2020-07-06 | 2021-05-18 | 中国科学院空天信息创新研究院 | SAR image complex building extraction method and system based on attention network selection |
| CN111915571A (en) * | 2020-07-10 | 2020-11-10 | 云南电网有限责任公司带电作业分公司 | Image change detection method, device, storage medium and equipment fusing residual error network and U-Net network |
| CN111932553B (en) * | 2020-07-27 | 2022-09-06 | 北京航空航天大学 | Remote sensing image semantic segmentation method based on area description self-attention mechanism |
| CN112070779B (en) * | 2020-08-04 | 2022-03-15 | 武汉大学 | Remote sensing image road segmentation method based on convolutional neural network weak supervised learning |
| CN111931684B (en) * | 2020-08-26 | 2021-04-06 | 北京建筑大学 | Weak and small target detection method based on video satellite data identification features |
| CN112396594B (en) * | 2020-11-27 | 2024-03-29 | 广东电网有限责任公司肇庆供电局 | Method and device for acquiring change detection model, change detection method, computer equipment and readable storage medium |
| CN112381172B (en) * | 2020-11-28 | 2022-09-16 | 桂林电子科技大学 | InSAR interference image phase unwrapping method based on U-net |
| CN112668494A (en) * | 2020-12-31 | 2021-04-16 | 西安电子科技大学 | Small sample change detection method based on multi-scale feature extraction |
| CN112767413B (en) * | 2021-01-06 | 2022-03-15 | 武汉大学 | Remote sensing image depth semantic segmentation method integrating region communication and symbiotic knowledge constraints |
| CN112733777B (en) * | 2021-01-18 | 2023-12-05 | 华侨大学 | Road extraction method, device and equipment of remote sensing image and storage medium |
| CN113096208B (en) * | 2021-03-16 | 2022-11-18 | 天津大学 | Reconstruction method of neural network magnetic resonance image based on double-domain alternating convolution |
| CN113065551B (en) * | 2021-03-19 | 2023-08-08 | 杭州迪英加科技有限公司 | Method for performing image segmentation using deep neural network model |
| CN112990041B (en) * | 2021-03-25 | 2022-10-18 | 武汉大学 | Remote sensing image building extraction method based on improved U-net |
| CN113033454B (en) * | 2021-04-07 | 2023-04-25 | 桂林电子科技大学 | Method for detecting building change in urban video shooting |
| CN113204608A (en) * | 2021-05-27 | 2021-08-03 | 广州大学 | Automatic map updating method, storage medium and system based on remote sensing image |
| CN113628180B (en) * | 2021-07-30 | 2023-10-27 | 北京科技大学 | Remote sensing building detection method and system based on semantic segmentation network |
| CN113901877A (en) * | 2021-09-13 | 2022-01-07 | 广州市城市规划勘测设计研究院 | Method, device and equipment for detecting change of remote sensing image building and storage medium |
| CN113688956A (en) * | 2021-10-26 | 2021-11-23 | 西南石油大学 | Sandstone slice segmentation and identification method based on depth feature fusion network |
| CN114549972B (en) * | 2022-01-17 | 2023-01-03 | 中国矿业大学(北京) | Strip mine stope extraction method, device, equipment and medium |
| CN115019186B (en) * | 2022-08-08 | 2022-11-22 | 中科星图测控技术(合肥)有限公司 | Method and system for detecting remote sensing change |
| CN115205710B (en) * | 2022-09-16 | 2022-12-09 | 北京理工大学 | Double-time-phase remote sensing image change detection method combined with color correction |
| 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 |
| CN116664599A (en) * | 2023-06-01 | 2023-08-29 | 云南大学 | Image steganalysis method based on steganography area prediction |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018214195A1 (en) * | 2017-05-25 | 2018-11-29 | 中国矿业大学 | Remote sensing imaging bridge detection method based on convolutional neural network |
| CN109447994A (en) * | 2018-11-05 | 2019-03-08 | 陕西师范大学 | In conjunction with the remote sensing image segmentation method of complete residual error and Fusion Features |
| CN109919206A (en) * | 2019-02-25 | 2019-06-21 | 武汉大学 | A kind of remote sensing image ground mulching classification method based on complete empty convolutional neural networks |
| CN109934154A (en) * | 2019-03-08 | 2019-06-25 | 北京科技大学 | A kind of remote sensing image variation detection method and detection device |
| CN109934200A (en) * | 2019-03-22 | 2019-06-25 | 南京信息工程大学 | A kind of RGB color remote sensing images cloud detection method of optic and system based on improvement M-Net |
-
2019
- 2019-11-06 CN CN201911073929.6A patent/CN111047551B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018214195A1 (en) * | 2017-05-25 | 2018-11-29 | 中国矿业大学 | Remote sensing imaging bridge detection method based on convolutional neural network |
| CN109447994A (en) * | 2018-11-05 | 2019-03-08 | 陕西师范大学 | In conjunction with the remote sensing image segmentation method of complete residual error and Fusion Features |
| CN109919206A (en) * | 2019-02-25 | 2019-06-21 | 武汉大学 | A kind of remote sensing image ground mulching classification method based on complete empty convolutional neural networks |
| CN109934154A (en) * | 2019-03-08 | 2019-06-25 | 北京科技大学 | A kind of remote sensing image variation detection method and detection device |
| CN109934200A (en) * | 2019-03-22 | 2019-06-25 | 南京信息工程大学 | A kind of RGB color remote sensing images cloud detection method of optic and system based on improvement M-Net |
Non-Patent Citations (1)
| Title |
|---|
| Yuhao Wang 等.Dense Semantic Labeling with Atrous Spatial Pyramid Pooling and Decoder for High-Resolution Remote Sensing Imagery.Remote Sensing.2018,第Volume 11卷第1-18页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111047551A (en) | 2020-04-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111047551B (en) | Remote sensing image change detection method and system based on U-net improved algorithm | |
| CN108509978B (en) | Multi-class target detection method and model based on CNN (CNN) multi-level feature fusion | |
| CN111612807B (en) | Small target image segmentation method based on scale and edge information | |
| CN111612008B (en) | Image segmentation method based on convolution network | |
| CN111814794B (en) | Text detection method and device, electronic equipment and storage medium | |
| Li et al. | Multiscale features supported DeepLabV3+ optimization scheme for accurate water semantic segmentation | |
| CN112541503A (en) | Real-time semantic segmentation method based on context attention mechanism and information fusion | |
| CN109657551B (en) | Face detection method based on context information enhancement | |
| CN112950477B (en) | Dual-path processing-based high-resolution salient target detection method | |
| CN113505792B (en) | Multi-scale semantic segmentation method and model for unbalanced remote sensing image | |
| CN113487618B (en) | Portrait segmentation method, portrait segmentation device, electronic equipment and storage medium | |
| CN112906706A (en) | Improved image semantic segmentation method based on coder-decoder | |
| CN113807355A (en) | Image semantic segmentation method based on coding and decoding structure | |
| CN111353544B (en) | Improved Mixed Pooling-YOLOV 3-based target detection method | |
| CN113888547A (en) | Non-supervision domain self-adaptive remote sensing road semantic segmentation method based on GAN network | |
| CN111797841B (en) | Visual saliency detection method based on depth residual error network | |
| CN110852327A (en) | Image processing method, image processing device, electronic equipment and storage medium | |
| CN111882620A (en) | Road drivable area segmentation method based on multi-scale information | |
| CN114724155A (en) | Scene text detection method, system and equipment based on deep convolutional neural network | |
| CN112700418A (en) | Crack detection method based on improved coding and decoding network model | |
| CN114155371A (en) | Semantic segmentation method based on channel attention and pyramid convolution fusion | |
| CN114120148B (en) | Method for detecting changing area of remote sensing image building | |
| CN115830449A (en) | Remote sensing target detection method with explicit contour guidance and spatial variation context enhancement | |
| CN113139544A (en) | Saliency target detection method based on multi-scale feature dynamic fusion | |
| CN115082778B (en) | Multi-branch learning-based homestead identification method and system |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |