CN113362286A - Natural resource element change detection method based on deep learning - Google Patents

Natural resource element change detection method based on deep learning Download PDF

Info

Publication number
CN113362286A
CN113362286A CN202110562234.5A CN202110562234A CN113362286A CN 113362286 A CN113362286 A CN 113362286A CN 202110562234 A CN202110562234 A CN 202110562234A CN 113362286 A CN113362286 A CN 113362286A
Authority
CN
China
Prior art keywords
natural resource
resource element
detected
region
segmentation
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.)
Granted
Application number
CN202110562234.5A
Other languages
Chinese (zh)
Other versions
CN113362286B (en
Inventor
王淑娟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jiangsu Xingyue Surveying And Mapping Technology Co ltd
Original Assignee
Jiangsu Xingyue Surveying And Mapping Technology Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Jiangsu Xingyue Surveying And Mapping Technology Co ltd filed Critical Jiangsu Xingyue Surveying And Mapping Technology Co ltd
Priority to CN202110562234.5A priority Critical patent/CN113362286B/en
Publication of CN113362286A publication Critical patent/CN113362286A/en
Application granted granted Critical
Publication of CN113362286B publication Critical patent/CN113362286B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
    • G06F18/24133—Distances to prototypes
    • G06F18/24137—Distances to cluster centroïds
    • G06F18/2414—Smoothing the distance, e.g. radial basis function networks [RBFN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • 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
    • 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
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • 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 natural resource element change detection method based on deep learning, which comprises the following steps: acquiring two images to be detected in the same area within different time; training a natural resource element recognition segmentation model; respectively identifying and segmenting natural resource elements in the image to be detected according to the natural resource element identification segmentation model to obtain a natural resource element segmentation graph; carrying out natural resource element merging processing on the natural resource element segmentation graph to obtain a natural resource element merging graph; performing element repeated detection and repeated element elimination on the natural resource elements in the natural resource element merged graph to obtain a natural resource element variation graph; wherein the natural resource element segmentation graph comprises a plurality of natural resource element regions; the natural resource element merged graph comprises the following steps: and the natural resource element segmentation graph upper layer and the natural resource element segmentation graph lower layer.

Description

Natural resource element change detection method based on deep learning
Technical Field
The invention relates to the technical field of image processing, in particular to a natural resource element change detection method based on deep learning.
Background
Natural resources are all tangible and intangible objects endowed by nature or reserved by predecessors and can be directly or indirectly used for meeting the needs of human beings, the natural resources can also promote the ecological industry of local development, provide ecological products, exert ecological benefits and provide substance products and economic benefits, so that the ecological advantages are converted into the economic advantages.
Disclosure of Invention
The invention provides a natural resource element change detection method based on deep learning, which is used for solving the problems of improving the detection precision and reducing a large amount of labor for manually identifying the natural resource element change when natural resource change detection is carried out on natural resources.
A natural resource element change detection method based on deep learning comprises the following steps:
acquiring two images to be detected in the same area within different time;
training a natural resource element recognition segmentation model;
respectively identifying and segmenting natural resource elements in the image to be detected according to the natural resource element identification segmentation model to obtain a natural resource element segmentation graph;
carrying out natural resource element merging processing on the natural resource element segmentation graph to obtain a natural resource element merging graph;
performing element repeated detection and repeated element elimination on the natural resource elements in the natural resource element merged graph to obtain a natural resource element variation graph;
wherein the natural resource element segmentation graph comprises a plurality of natural resource element regions;
the natural resource element merged graph comprises the following steps: and the natural resource element segmentation graph upper layer and the natural resource element segmentation graph lower layer.
As an embodiment of the present invention, the natural resource elements include agricultural resource elements, water resource elements, forest resource elements, and mineral resource elements.
As an embodiment of the present invention, training a natural resource element recognition segmentation model includes:
acquiring natural resource element category characteristic information and related scene characteristic information related to the natural resource element category characteristic information;
merging the natural resource element category characteristic information and the related scene characteristic information to obtain natural resource element merging characteristic information;
inputting the natural resource element combination characteristic information into a preset original generation countermeasure network to perform natural resource element image generation processing to obtain an original natural resource element image;
inputting the original natural resource element image into a preset original judgment network for true and false judgment to obtain an original natural resource element image judgment result;
inputting the original natural resource element image into a classification network of a preset original natural resource element image segmentation model to perform natural resource element image segmentation to obtain an original natural resource element image segmentation result;
and training a classification network of a preset original natural resource element image segmentation model based on the original natural resource element image discrimination result, the original natural resource element image segmentation result and the natural resource element category characteristic information to obtain a natural resource element identification segmentation model.
As an embodiment of the present invention, training a classification network of a preset original natural resource element image segmentation model based on an original natural resource element image discrimination result, an original natural resource element image segmentation result, and natural resource element category feature information to obtain a natural resource element recognition segmentation model, includes:
calculating a first recognition loss based on the discrimination result of the original natural resource element image and the authenticity of the original natural resource element image, wherein the calculation formula is as follows:
los1＝Ei～w(i)[log10(1-P(i))]+Ej～h(j)[log10P(j)]
wherein los is1For the first identification loss, i-w (i) is data generated by an original generation countermeasure network in an original natural resource element image, j-h (j) is real data in the original natural resource element image, P (i) is the probability that the discrimination result of the data generated by the original generation countermeasure network in the original natural resource element image is true, P (j) is the probability that the discrimination result of the real data in the original natural resource element image is true, Ei～w(i)[log10(1-P(i))]Is a function [ log ]10(1-P(i))]～[log10(1-P(w(i)))]Mathematical expectation of (1), Ej～h(j)[log10P(j)]Is a function [ log ]10P(j)]～[log10P(h(j))]A mathematical expectation of (d);
calculating a second segmentation loss based on the original natural resource element image segmentation result and the natural resource element category characteristic information, wherein the calculation formula is as follows:
Figure BDA0003079421810000031
wherein los is2For the second segmentation penalty, α and β are the length and width, k (σ), in the image size of the original natural resource element imagev,y) For the original natural resource element image segmentation result, tauv,yIs a preset AND k (sigma)v,y) A corresponding correct image segmentation result;
calculating the loss of the third natural resource element based on the first identification loss and the second segmentation loss, wherein the calculation formula is as follows:
los3＝los1+los2
if the loss of the third natural resource element is larger than a preset loss threshold value, updating a classification network of a preset original natural resource element image segmentation model, originally generating network parameters in a confrontation network and an original judgment network, and recalculating the loss of the third natural resource element;
and if the loss of the third natural resource element is less than or equal to a preset loss threshold value, taking a preset original natural resource element image segmentation model as a natural resource element identification segmentation model.
As an embodiment of the present invention, performing element repetition detection on a natural resource element in a natural resource element merged graph includes the following steps:
step 1: dividing the natural resource element merged graph into a plurality of detection areas, wherein each detection area only contains two natural resource element categories at most;
step 2: respectively acquiring natural resource element areas in an upper layer of a natural resource element segmentation map and a lower layer of the natural resource element segmentation map in a detection area;
and step 3: and if the natural resource element type in the upper layer of the natural resource element segmentation graph in the detection area is the same as the natural resource element type in the lower layer of the natural resource element segmentation graph in the detection area, judging that the natural resource elements in the current detection area are repeated.
As an embodiment of the present invention, the repeated element elimination of the natural resource elements in the natural resource element merged graph includes the following steps:
step 1: acquiring repeated natural resource element areas in an upper layer of a natural resource element segmentation map and a lower layer of the natural resource element segmentation map in the detection area;
step 2: respectively determining the range of the repeated natural resource element region in the upper layer of the natural resource element segmentation map and the lower layer of the natural resource element segmentation map in the detection region;
and step 3: and carrying out repeated element elimination on the natural resource elements in the range of the repeated natural resource element area in the upper layer of the natural resource element segmentation graph and the lower layer of the natural resource element segmentation graph in the detection area.
As an embodiment of the present invention, a natural resource element change detection method based on deep learning further includes:
carrying out image smoothing on an image to be detected to obtain a better image to be detected;
and carrying out image smoothing treatment on the natural resource element change graph to obtain a better natural resource element change graph.
As an embodiment of the present invention, a natural resource element change detection method based on deep learning further includes:
dividing an image to be detected into a plurality of regions to be detected according to a preset division rule;
acquiring actual change data of natural resource elements in the same region to be detected in two images to be detected, wherein the actual change data comprises: the shooting time of the two images to be detected and the variety change condition of natural resource elements are determined by taking one of the two images to be detected with longer shooting time as a standard diagram, taking any one of the regions to be detected in the standard diagram as a standard region to be detected, dividing the standard region to be detected into a central region to be detected, an east region to be detected, a west region to be detected, a south region to be detected and a north region to be detected according to a preset second division rule, taking one of the two images to be detected with shorter shooting time as a comparison diagram, taking the region to be detected corresponding to the standard region to be detected in the comparison diagram as a comparison region to be detected, and performing superposition comparison on the central region to be detected, the east region to be detected, the west region to be detected, the south region to be detected and the north region to be detected and the comparison region to be detected to obtain the central region range difference, the natural resource element number difference, The east to-be-detected region range difference, the west to-be-detected region range difference, the south to-be-detected region range difference and the north to-be-detected region range difference;
calculating actual change values of different natural resource elements in any to-be-detected region in different shooting intervals under the condition that the types of the natural resource elements are not changed based on actual change data, wherein the calculation formula is as follows:
Figure BDA0003079421810000051
wherein f ist,δ,εThe actual change value S of the epsilon-th natural resource element in the delta-th to-be-detected area in the interval time t is shot under the condition that the type of the natural resource element is not changedz,δThe central detected region range difference of the delta-th detected region, Sd,δEast to-be-detected region range difference of the δ -th to-be-detected region, Sn,δSouth to-be-detected region range difference S of the delta-th to-be-detected regionx,δWest zone difference, S, for the δ -th zoneb,δIs the north area-to-be-detected range difference mu of the delta area-to-be-detectedz,ε、μd,ε、μn,ε、μx,εAnd mub,εDifferent range error coefficients corresponding to the preset epsilon natural resource elements are respectively set;
if the actual change value of different natural resource elements in any to-be-detected area in different shooting intervals is larger than the ideal change value of the corresponding natural resource element in the preset corresponding shooting interval under the condition that the types of the natural resource elements are not changed, carrying out artificial intervention marking on any to-be-detected area corresponding to the situation that the actual change value of the different natural resource elements in any to-be-detected area in the different shooting intervals is larger than the ideal change value of the corresponding natural resource element in the preset corresponding shooting interval;
and if the actual change value of different natural resource elements in any to-be-detected area in different shooting intervals is smaller than or equal to the ideal change value of the corresponding natural resource element in the preset corresponding shooting interval under the condition that the types of the natural resource elements are not changed, marking the natural change of the corresponding to-be-detected area in the preset corresponding shooting interval when the actual change value of the different natural resource elements in any to-be-detected area in the different shooting intervals is larger than the ideal change value of the corresponding natural resource element in the preset corresponding shooting interval.
As an embodiment of the present invention, a natural resource element change detection method based on deep learning further includes:
when the type of the natural resource elements in any region to be detected changes, carrying out artificial intervention marking on the region to be detected corresponding to the change of the type of the natural resource elements;
wherein, the changing of the nature resource element comprises: the variety of the natural resource elements is changed, the natural resource elements are lost, and the natural resource elements are increased.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and drawings.
The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention.
In the drawings:
fig. 1 is a schematic diagram of a method for detecting natural resource element change based on deep learning in an embodiment of the present invention.
Detailed Description
The preferred embodiments of the present invention will be described in conjunction with the accompanying drawings, and it will be understood that they are described herein for the purpose of illustration and explanation and not limitation.
Referring to fig. 1, an embodiment of the present invention provides a method for detecting a change in natural resource elements based on deep learning, including:
acquiring two images to be detected in the same area within different time;
training a natural resource element recognition segmentation model;
respectively identifying and segmenting natural resource elements in the image to be detected according to the natural resource element identification segmentation model to obtain a natural resource element segmentation graph;
carrying out natural resource element merging processing on the natural resource element segmentation graph to obtain a natural resource element merging graph;
performing element repeated detection and repeated element elimination on the natural resource elements in the natural resource element merged graph to obtain a natural resource element variation graph;
wherein the natural resource element segmentation graph comprises a plurality of natural resource element regions;
the natural resource element merged graph comprises the following steps: the upper layer of the natural resource element segmentation graph and the lower layer of the natural resource element segmentation graph;
the working principle of the technical scheme is as follows: firstly, training a corresponding natural resource element recognition segmentation model according to the type of a natural resource element to be recognized and segmented, acquiring two images to be detected which are prepared for detection and are shot in the same region at different time, inputting the two images to be detected which are shot in the same region at different time into the natural resource element recognition segmentation model, recognizing and segmenting the natural resource element in the two images to be detected according to the training result in the model by the natural resource element recognition segmentation model, wherein the recognizing and segmenting refers to recognizing the natural resource element, the type of the natural resource element and the non-natural resource element in the images to be detected, segmenting the recognized natural resource element and the non-natural resource element to form two natural resource element segmentation maps, wherein the natural resource element segmentation maps except the natural resource element, the other parts are preferably covered by black, so that natural resource elements can be more clearly identified, the natural resource element segmentation maps are subjected to natural resource element combination processing after the natural resource element segmentation maps are obtained, preferably, the two natural resource element segmentation maps are subjected to translucency and then overlapped up and down, so that the difference of the natural resource elements in the upper image layer and the lower image layer of the natural resource element segmentation maps can be clearly distinguished in the obtained natural resource element combination maps, preferably, the natural resource element segmentation map formed by the image to be detected with longer shooting time is taken as the upper image layer of the natural resource element segmentation map, the natural resource element segmentation map formed by the other image to be detected is taken as the lower image layer of the natural resource element segmentation map, and then element repeated detection and repeated element elimination are carried out on the obtained natural resource element segmentation map to obtain a natural resource element change map, the element repetition detection is to judge whether there is an overlap between the types of natural resource elements corresponding to the upper layer of the natural resource element division diagram and the lower layer of the natural resource element division diagram, the preferable element repetition detection method is to judge that the types of natural resource elements overlap and is marked as 1 if there is a same type of natural resource element in a detection area of the upper layer and the lower layer, and is marked as 2 if there is a different type of natural resource element in a detection area of the upper layer and the lower layer, and is marked as 3 if there is a natural resource element in a detection area of the upper layer and a natural resource element does not exist in a detection area corresponding to the lower layer, and is to judge that the types of natural resource elements do not overlap and is marked as 3 if there is a natural resource element in a detection area of the lower layer, if no natural resource element exists in a certain detection area corresponding to the upper image layer, judging that the type of the natural resource element is not overlapped in the second type, marking the type as 4, removing overlapped parts of the natural resource element, which are overlapped in the natural resource element type in the upper image layer of the natural resource element segmentation image and the natural resource element type in the lower image layer of the natural resource element segmentation image, by repeated elements, and further obtaining a natural resource element change image, wherein the natural resource element change image shows the change condition of the natural resource element in the two images to be detected, preferably, the natural resource element type in the two images to be detected in the shooting interval is expanded or reduced and changed in the shooting interval, the natural resource element type in the two images to be detected in the shooting interval is marked as 2, and the natural resource element type in the two images to be detected in the shooting interval is marked as 3, marked as 4 is the newly added change of the natural resource element types in the two images to be detected in the period of the shooting interval, and the changes marked as 2, 3 and 4 are changes caused by human;
the beneficial effects of the above technical scheme are: the labor amount for manually identifying the change of the natural resource elements is reduced, the detection accuracy of the change of the natural resource elements is improved, and the change condition of the natural resource elements in the same region in different time is more accurately detected.
In one embodiment, the natural resource elements include agricultural resource elements, water resource elements, forest resource elements, mineral resource elements.
In one embodiment, training a natural resource element recognition segmentation model comprises:
acquiring natural resource element category characteristic information and related scene characteristic information related to the natural resource element category characteristic information;
merging the natural resource element category characteristic information and the related scene characteristic information to obtain natural resource element merging characteristic information;
inputting the natural resource element combination characteristic information into a preset original generation countermeasure network to perform natural resource element image generation processing to obtain an original natural resource element image;
inputting the original natural resource element image into a preset original judgment network for true and false judgment to obtain an original natural resource element image judgment result;
inputting the original natural resource element image into a classification network of a preset original natural resource element image segmentation model to perform natural resource element image segmentation to obtain an original natural resource element image segmentation result;
training a classification network of a preset original natural resource element image segmentation model based on the original natural resource element image discrimination result, the original natural resource element image segmentation result and the natural resource element category characteristic information to obtain a natural resource element identification segmentation model;
the working principle of the technical scheme is as follows: in training a natural resource element recognition segmentation model, first, feature information related to a natural resource element category and related scene feature information related to the feature information of the natural resource element category are obtained, for example, scene feature information related to water resource feature information and feature information of water resources, such as feature information of an ocean, feature information of a lake, and the like are obtained, and the collected feature information related to the natural resource element category and related scene feature information related to the feature information of the natural resource element category are input into a preset original generation countermeasure network to be subjected to natural resource element image generation processing, so as to obtain an original natural resource element image, wherein the preset original generation countermeasure network is preferably an initial generation countermeasure network which is not subjected to any training, and the original natural resource element image comprises input real natural resource element data and simulated natural resource element number generated by the original generation countermeasure network According to the method, after an original natural resource element image is obtained, the original natural resource element image is input into a preset original discrimination network to judge the authenticity of data in the original natural resource element image, namely the input real natural resource element data and simulated natural resource element data generated by an original generation countermeasure network are judged to judge the authenticity, finally, an original natural resource element image discrimination result is obtained, the same is preferably the initial discrimination network without any training, then, the original natural resource element image is input into a classification network of a preset original natural resource element image segmentation model to carry out natural resource element image segmentation, an original natural resource element image segmentation result is obtained, and finally, the original natural resource element image discrimination result, the original natural resource element image segmentation result and natural resource element category characteristic information are used for training the original natural resource element image to obtain the original natural resource element image segmentation result Obtaining a natural resource element identification segmentation model through a classification network of the image segmentation model;
the beneficial effects of the above technical scheme are: the method is beneficial to improving the recognition and segmentation accuracy of the natural resource element recognition and segmentation model.
In one embodiment, training a classification network of a preset original natural resource element image segmentation model based on an original natural resource element image discrimination result, an original natural resource element image segmentation result, and natural resource element category feature information to obtain a natural resource element recognition segmentation model, includes:
calculating a first recognition loss based on the discrimination result of the original natural resource element image and the authenticity of the original natural resource element image, wherein the calculation formula is as follows:
los1＝Ei～w(i)[log10(1-P(i))]+Ej～h(j)[log10P(j)]
wherein los is1For the first identification loss, i-w (i) is data generated by an original generation countermeasure network in an original natural resource element image, j-h (j) is real data in the original natural resource element image, P (i) is the probability that the discrimination result of the data generated by the original generation countermeasure network in the original natural resource element image is true, P (j) is the probability that the discrimination result of the real data in the original natural resource element image is true, Ei～w(i)[log10(1-P(i))]Is a function [ log ]10(1-P(i))]～[log10(1-P(w(i)))]Mathematical expectation of (1), Ej～h(j)[log10P(j)]Is a function [ log ]10P(j)]～[log10P(h(j))]A mathematical expectation of (d);
calculating a second segmentation loss based on the original natural resource element image segmentation result and the natural resource element category characteristic information, wherein the calculation formula is as follows:
Figure BDA0003079421810000121
wherein los is2For the second segmentation penalty, α and β are the length and width, k (σ), in the image size of the original natural resource element imagev,y) For the original natural resource element image segmentation result, tauv,yIs a preset AND k (sigma)v,y) A corresponding correct image segmentation result;
calculating the loss of the third natural resource element based on the first identification loss and the second segmentation loss, wherein the calculation formula is as follows:
los3＝los1+los2
if the loss of the third natural resource element is larger than a preset loss threshold value, updating a classification network of a preset original natural resource element image segmentation model, originally generating network parameters in a confrontation network and an original judgment network, and recalculating the loss of the third natural resource element;
if the loss of the third natural resource element is less than or equal to a preset loss threshold value, taking a preset original natural resource element image segmentation model as a natural resource element recognition segmentation model;
the working principle and the beneficial effects of the technical scheme are as follows: first, a first recognition loss is calculated from the original natural resource element image discrimination result and the authenticity of the original natural resource element image, the first recognition loss representing an error loss of the original generation countermeasure network and the original discrimination network when the original natural resource element image generation and the authenticity of the original natural resource element image are discriminated, then a second division loss is calculated from the original natural resource element image division result and the natural resource element category feature information, the second division loss representing an error loss of the classification network of the original natural resource element image division model between a result of the original natural resource element image division and a preset result of the natural resource element division based on the natural resource element category feature information, and a third natural resource element loss is calculated from the first recognition loss and the second division loss, when the loss of the third natural resource element is larger than a preset loss threshold value, updating the classification network of the original natural resource element image segmentation model, originally generating network parameters in the confrontation network and the original judgment network, and recalculating the loss of the third natural resource element; the updating refers to inputting the natural resource element merging feature information into a preset original generation countermeasure network again for natural resource element image generation processing to obtain an updated original natural resource element image, inputting the updated original natural resource element image into a preset original judgment network for true and false judgment to obtain an updated original natural resource element image judgment result, inputting the updated original natural resource element image into a classification network of a preset original natural resource element image segmentation model for natural resource element image segmentation to obtain an updated original natural resource element image segmentation result, and finally performing natural resource element image segmentation based on the updated original natural resource element image segmentation resultAnd then recalculating the third natural resource element loss according to the resource element image discrimination result, the updated original natural resource element image segmentation result and the natural resource element category characteristic information until the third natural resource element loss is less than or equal to a preset loss threshold value, and taking the original natural resource element image segmentation model as a natural resource element identification segmentation model to improve the identification and segmentation accuracy of the natural resource element identification segmentation model in element identification segmentation of the natural resource elements, wherein the preset sum is k (sigma)v,y) Corresponding correct image segmentation result tauv,yPreferably, the image segmentation is performed according to the natural resource element category feature information.
In one embodiment, the natural resource element merging processing is performed on the natural resource element partition map to obtain a natural resource element merging map, including:
merging a plurality of natural resource element areas in the two natural resource element segmentation graphs to obtain natural resource element merging graphs;
the working principle and the beneficial effects of the technical scheme are as follows: preferably, a natural resource element segmentation graph formed by an image to be detected with longer shooting time is used as an upper layer of the natural resource element segmentation graph, a natural resource element segmentation graph formed by another image to be detected is used as a lower layer of the natural resource element segmentation graph, wherein a plurality of natural resource element regions are respectively arranged on the upper layer of the natural resource element segmentation graph and the lower layer of the natural resource element segmentation graph, natural resource element merging processing is carried out on the upper layer of the natural resource element segmentation graph and the lower layer of the natural resource element segmentation graph, preferably, the two natural resource element segmentation graphs are semitransparent and then overlapped up and down, and thus, the obtained natural resource element merging graph can clearly distinguish the difference of natural resource elements in the upper layer of the natural resource element segmentation graph and the lower layer of the natural resource element segmentation graph.
In one embodiment, the element repetition detection of the natural resource elements in the natural resource element merged graph includes the following steps:
step 1: dividing the natural resource element merged graph into a plurality of detection areas, wherein each detection area only contains two natural resource element categories at most;
step 2: respectively acquiring natural resource element areas in an upper layer of a natural resource element segmentation map and a lower layer of the natural resource element segmentation map in a detection area;
and step 3: if the natural resource element type in the upper layer of the natural resource element segmentation graph in the detection area is the same as the natural resource element type in the lower layer of the natural resource element segmentation graph in the detection area, judging that the natural resource elements in the current detection area are repeated;
the working principle and the beneficial effects of the technical scheme are as follows: dividing the natural resource element merged map into a plurality of detection areas, preferably according to the positions of the natural resource elements in the upper layer of the natural resource element segmentation map in the natural resource element merged map, wherein each detection area contains at most two natural resource element categories, which means that the total natural resource element categories existing in the upper layer of the natural resource element segmentation map in the natural resource element merged map and the lower layer of the natural resource element segmentation map in the natural resource element merged map in the corresponding detection area, usually only one natural resource element category is contained in each detection area, for example, one water resource exists in a detection area of the upper layer, one water resource also exists in the detection area of the lower layer, or one water resource exists in a detection area of the upper layer, and no natural resource element exists in the detection area of the lower layer, or a water resource exists in a detection area of the lower layer, a natural resource element does not exist in the detection area of the upper layer, but a detection area containing two natural resource element types, such as returning forest, exists in the detection area of the upper layer, and a forest resource exists in the detection area of the lower layer, which is usually caused by human, the natural resource element areas in the upper layer of the natural resource element division map and the lower layer of the natural resource element division map in the detection area are respectively obtained, element repetition detection is performed on the natural resource elements of each detection area, namely, the judgment is performed according to whether the natural resource element types correspondingly existing in the upper layer of the natural resource element division map and the lower layer of the natural resource element division map overlap, and the preferable element repetition detection mode is that if the same type of natural resource elements simultaneously exist in a detection area of the upper layer and the lower layer, judging that the natural resource element types are overlapped and marked as 1, if the natural resource elements simultaneously existing in a certain detection area of an upper layer and a lower layer are different, judging that the natural resource element types are not overlapped in a first type and marked as 2, if the natural resource elements exist in a certain detection area of the upper layer and the natural resource elements do not exist in a certain detection area corresponding to the lower layer, judging that the natural resource element types are not overlapped in a second type and marked as 3, if the natural resource elements exist in a certain detection area of the lower layer and the natural resource elements do not exist in a certain detection area corresponding to the upper layer, judging that the natural resource element types are not overlapped in the second type and marked as 4, preferably, the natural resource element types marked as 1 are the expansion or reduction change of the natural resource elements in the two images to be detected in the period of the shooting interval, and the natural resource element types marked as 2 in the two images to be detected in the period of the shooting interval, the change marked as 3 is the disappearance change of the natural resource element types in the two images to be detected in the period of the shooting interval, the change marked as 4 is the new increase of the natural resource element types in the two images to be detected in the period of the shooting interval, the changes marked as 2, 3 and 4 are changes caused by human, the reason that the natural resource elements are repeated can be better distinguished, and the detection precision of the natural resource element change can be improved.
In one embodiment, the repeated element elimination of the natural resource elements in the natural resource element merged graph comprises the following steps:
step 1: acquiring repeated natural resource element areas in an upper layer of a natural resource element segmentation map and a lower layer of the natural resource element segmentation map in the detection area;
step 2: respectively determining the range of the repeated natural resource element region in the upper layer of the natural resource element segmentation map and the lower layer of the natural resource element segmentation map in the detection region;
and step 3: repeating element elimination is carried out on the natural resource elements in the range of the repeated natural resource element area in the upper layer of the natural resource element segmentation graph and the lower layer of the natural resource element segmentation graph in the detection area;
the working principle of the scheme is as follows: obtaining repeated natural resource element areas in an upper layer of a natural resource element segmentation map and a lower layer of the natural resource element segmentation map in a detection area through element repeated detection of natural resource elements, respectively determining the range of the repeated natural resource element areas in the upper layer of the natural resource element segmentation map and the lower layer of the natural resource element segmentation map in the detection area, and then performing repeated element rejection on the natural resource elements in the range of the repeated natural resource element areas in the upper layer of the natural resource element segmentation map and the lower layer of the natural resource element segmentation map in the detection area, which is beneficial to more intuitively feeling the change of the natural resource elements;
in one embodiment, a natural resource element change detection method based on deep learning further includes:
carrying out image smoothing on an image to be detected to obtain a better image to be detected;
carrying out image smoothing on the natural resource element change diagram to obtain a more optimal natural resource element change diagram;
the working principle and the beneficial effects of the technical scheme are as follows: the image smoothing processing is carried out on the image to be detected and the natural resource element change graph, so that the accuracy of the change degree of the subsequently observed natural resource elements is improved, and the image smoothing processing is preferably median filtering processing.
In one embodiment, a natural resource element change detection method based on deep learning further includes:
dividing an image to be detected into a plurality of regions to be detected according to a preset division rule;
acquiring actual change data of natural resource elements in the same region to be detected in two images to be detected, wherein the actual change data comprises: the shooting time of the two images to be detected and the variety change condition of natural resource elements are determined by taking one of the two images to be detected with longer shooting time as a standard diagram, taking any one of the regions to be detected in the standard diagram as a standard region to be detected, dividing the standard region to be detected into a central region to be detected, an east region to be detected, a west region to be detected, a south region to be detected and a north region to be detected according to a preset second division rule, taking one of the two images to be detected with shorter shooting time as a comparison diagram, taking the region to be detected corresponding to the standard region to be detected in the comparison diagram as a comparison region to be detected, and performing superposition comparison on the central region to be detected, the east region to be detected, the west region to be detected, the south region to be detected and the north region to be detected and the comparison region to be detected to obtain the central region range difference, the natural resource element number difference, The east to-be-detected region range difference, the west to-be-detected region range difference, the south to-be-detected region range difference and the north to-be-detected region range difference;
calculating actual change values of different natural resource elements in any to-be-detected region in different shooting intervals under the condition that the types of the natural resource elements are not changed based on actual change data, wherein the calculation formula is as follows:
Figure BDA0003079421810000171
wherein f ist,δ,εThe actual change value S of the epsilon-th natural resource element in the delta-th to-be-detected area in the interval time t is shot under the condition that the type of the natural resource element is not changedz,δThe central detected region range difference of the delta-th detected region, Sd,δEast to-be-detected region range difference of the δ -th to-be-detected region, Sn,δSouth to-be-detected region range difference S of the delta-th to-be-detected regionx,δWest zone difference, S, for the δ -th zoneb,δIs the north area-to-be-detected range difference mu of the delta area-to-be-detectedz,ε、μd,ε、μn,ε、μx,εAnd mub，εDifferent range error coefficients corresponding to the preset epsilon natural resource elements are respectively set;
if the actual change value of different natural resource elements in any to-be-detected area in different shooting intervals is larger than the ideal change value of the corresponding natural resource element in the preset corresponding shooting interval under the condition that the types of the natural resource elements are not changed, carrying out artificial intervention marking on any to-be-detected area corresponding to the situation that the actual change value of the different natural resource elements in any to-be-detected area in the different shooting intervals is larger than the ideal change value of the corresponding natural resource element in the preset corresponding shooting interval;
if the actual change value of different natural resource elements in any to-be-detected area in different shooting intervals is smaller than or equal to the ideal change value of the corresponding natural resource element in the preset corresponding shooting interval under the condition that the types of the natural resource elements are not changed, performing natural change marking on any to-be-detected area corresponding to the situation that the actual change value of the different natural resource elements in any to-be-detected area in the different shooting intervals is larger than the ideal change value of the corresponding natural resource element in the preset corresponding shooting intervals;
when the type of the natural resource elements in any region to be detected changes, carrying out artificial intervention marking on the region to be detected corresponding to the change of the type of the natural resource elements;
wherein, the changing of the nature resource element comprises: the variety of the natural resource elements is changed, the natural resource elements are lost, and the natural resource elements are increased.
The working principle of the technical scheme is as follows: dividing an image to be detected into a plurality of regions to be detected according to a preset division rule, preferably dividing the preset division rule according to the distribution condition of natural resource elements of two current images to be detected, wherein each region to be detected only comprises one natural resource element, and then acquiring actual change data of the natural resource elements in the same region to be detected in the two images to be detected, wherein the actual change data comprises: the shooting time of the two images to be detected, the change condition of the range of the natural resource element region in the same region to be detected and the change condition of the type of the natural resource element in the same region to be detected are specifically that one of the two images to be detected, which is shot for a longer time, is taken as a standard graph (the long shooting time here means that the image to be detected is older than the current time, for example, one image to be detected is shot in 2015, the other image to be detected is shot in 2019, the image to be detected shot in 2019 is taken as the image to be detected with a longer shooting time), any one of the regions to be detected in the standard graph is taken as the standard region to be detected, and the standard region to be detected is divided into a central region to be detected, an east region to be detected, a west region to be detected, a south region to be detected and a north region to be detected according to a preset second division rule, the preset second division rule is preferably to divide the standard region to be detected according to different directions, for example, there is a polygonal region to be detected, first, the largest square that can be taken from the region to be detected is selected as the central region to be detected, then, the four sides of the central region to be detected and the directions of east, south, west and north in the region to be detected divide the region to be detected in four directions except the central region to be detected in the standard region to be detected into the east region to be detected, the west region to be detected, the south region to be detected and the north region to be detected, the other of the two images to be detected is used as a comparison graph, the region to be detected corresponding to the standard region to be detected in the comparison graph is used as the comparison region to be detected, the method for selecting the corresponding region to be detected is preferably to overlap the two images to be detected, and the centers of all the regions to be detected in the comparison graph and the standard graph are marked, finding a central point in a standard area to be detected, selecting an area to be detected represented by a central point which is closest to the central point in the standard area to be detected in a comparison graph after the comparison graph is overlapped with the standard graph as an area to be detected corresponding to the standard area to be detected in the comparison graph, overlapping and comparing a central area to be detected, an east area to be detected, a west area to be detected, a south area to be detected and a north area to be detected with the area to be detected to obtain a central area range difference to be detected, an east area range difference to be detected, a west area range difference to be detected, a south area range difference to be detected and a north area range difference to be detected, calculating actual change values of different natural resource elements in any area to be detected in different shooting interval time under the condition that the types of the natural resource elements are not changed based on actual change data, wherein the actual change values represent that the natural resource elements in the area to be detected move to any direction in the shooting interval time To the expansion or contraction change condition and the change condition that the type changes or disappears, if the actual change value of different natural resource elements in any one to-be-detected region in different shooting intervals is larger than the ideal change value of the corresponding natural resource element in the preset corresponding shooting interval under the condition that the type of the natural resource elements is not changed, the artificial intervention marking is carried out on any to-be-detected region corresponding to the situation that the actual change value of different natural resource elements in any to-be-detected region in different shooting intervals is larger than the ideal change value of the corresponding natural resource element in the preset corresponding shooting interval, the artificial intervention marking is preferably red marked on the corresponding to-be-detected region, if the actual change value of different natural resource elements in any to-be-detected region in different shooting intervals is not larger than or equal to the actual change value of the corresponding natural resource elements in the preset corresponding shooting interval under the condition that the type of the natural resource elements is not changed When the ideal change value is obtained, the natural change mark is carried out on any corresponding region to be detected in the ideal change value of the corresponding natural resource element in the preset corresponding shooting interval time, wherein the actual change value of the different natural resource elements in any region to be detected in the different shooting interval time is greater than the corresponding region to be detected in the ideal change value of the corresponding natural resource element in the preset corresponding shooting interval time, the natural change mark is preferably a green mark drawn on the corresponding region to be detected, in addition, when the type of the natural resource element in any region to be detected is changed, the artificial intervention mark is carried out on the region to be detected corresponding to the change of the type of the natural resource element, wherein the change of the type of the natural resource element comprises the following steps: changing the types of natural resource elements, disappearing the natural resource elements and newly adding the natural resource elements;
the beneficial effects of the above technical scheme are: the change of the natural resource elements in the regions to be detected at the same position in different images to be detected is detected, and corresponding marks are made for the corresponding regions to be detected according to the detection result, so that when a user observes the change condition of the natural resource elements in a certain region, the user can visually see whether the natural resource elements are artificially changed or naturally changed, and the labor amount for artificially identifying the change of the natural resource elements is reduced. It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.

Claims (10)

1. A natural resource element change detection method based on deep learning is characterized by comprising the following steps:
acquiring two images to be detected in the same area within different time;
training a natural resource element recognition segmentation model;
respectively identifying and segmenting natural resource elements in the image to be detected according to the natural resource element identification segmentation model to obtain a natural resource element segmentation graph;
performing natural resource element merging processing on the natural resource element segmentation graph to obtain a natural resource element merging graph;
performing element repeated detection and repeated element elimination on the natural resource elements in the natural resource element merged graph to obtain a natural resource element variation graph;
wherein the natural resource element partition map comprises a plurality of natural resource element regions;
the natural resource element merged graph comprises: and the natural resource element segmentation graph upper layer and the natural resource element segmentation graph lower layer.
2. The method as claimed in claim 1, wherein the natural resource elements include agricultural resource elements, water resource elements, forest resource elements and mineral resource elements.
3. The method for detecting natural resource element change based on deep learning of claim 1, wherein the training of the natural resource element recognition segmentation model comprises:
acquiring natural resource element category characteristic information and related scene characteristic information related to the natural resource element category characteristic information;
merging the natural resource element category characteristic information and the related scene characteristic information to obtain natural resource element merging characteristic information;
inputting the natural resource element combination characteristic information into a preset original generation countermeasure network to perform natural resource element image generation processing to obtain an original natural resource element image;
inputting the original natural resource element image into a preset original judgment network for true and false judgment to obtain an original natural resource element image judgment result;
inputting the original natural resource element image into a classification network of a preset original natural resource element image segmentation model to perform natural resource element image segmentation to obtain an original natural resource element image segmentation result;
and training a classification network of the preset original natural resource element image segmentation model based on the original natural resource element image discrimination result, the original natural resource element image segmentation result and the natural resource element class characteristic information to obtain a natural resource element identification segmentation model.
4. The method as claimed in claim 3, wherein the training of the classification network of the preset original natural resource element image segmentation model based on the original natural resource element image discrimination result, the original natural resource element image segmentation result, and the natural resource element category feature information to obtain a natural resource element recognition segmentation model comprises:
calculating a first recognition loss based on the original natural resource element image discrimination result and the authenticity of the original natural resource element image, wherein a calculation formula is as follows:
los1＝Ei～w(i)[log10(1-P(i))]+Ej～h(j)[log10P(j)]
wherein los is1For the first identification loss, i-w (i) is data generated by an original generation countermeasure network in an original natural resource element image, j-h (j) is real data in the original natural resource element image, P (i) is the probability that the discrimination result of the data generated by the original generation countermeasure network in the original natural resource element image is true, P (j) is the probability that the discrimination result of the real data in the original natural resource element image is true, Ei～w(i)[log10(1-P(i))]Is a function [ log ]10(1-P(i))]～[log10(1-P(w(i)))]Mathematical expectation of (1), Ej～h(j)[log10P{j)]Is a function [ log ]10P(j)]～[log10P(h(j))]A mathematical expectation of (d);
calculating a second segmentation loss based on the original natural resource element image segmentation result and the natural resource element category feature information, wherein a calculation formula is as follows:
Figure FDA0003079421800000031
wherein los is2For the second segmentation penalty, α and β are the length and width, k (σ), in the image size of the original natural resource element imagev，y) For the original natural resource element image segmentation result, tauv，yIs a preset AND k (sigma)v，y) A corresponding correct image segmentation result;
calculating a third natural resource element loss based on the first recognition loss and the second segmentation loss, the calculation formula being as follows:
los3＝los1+los2
if the third natural resource element loss is larger than a preset loss threshold value, updating the classification network of the preset original natural resource element image segmentation model, the network parameters in the originally generated countermeasure network and the originally judged network, and recalculating the third natural resource element loss;
and if the loss of the third natural resource element is less than or equal to a preset loss threshold value, taking the preset original natural resource element image segmentation model as the natural resource element identification segmentation model.
5. The method for detecting natural resource element change based on deep learning according to claim 1, wherein the step of performing natural resource element merging processing on the natural resource element segmentation graph to obtain a natural resource element merging graph comprises:
and merging a plurality of natural resource element areas in the two natural resource element segmentation graphs to obtain a natural resource element merged graph.
6. The method for detecting natural resource element change based on deep learning according to claim 1, wherein the element repetition detection of the natural resource elements in the natural resource element merged map comprises the following steps:
step 1: dividing the natural resource element merged graph into a plurality of detection areas, wherein each detection area at most comprises two natural resource element categories;
step 2: respectively acquiring natural resource element areas in an upper layer of the natural resource element segmentation chart and a lower layer of the natural resource element segmentation chart in the detection area;
and step 3: and if the natural resource element type in the upper layer of the natural resource element segmentation graph in the detection area is the same as the natural resource element type in the lower layer of the natural resource element segmentation graph in the detection area, judging that the natural resource elements in the current detection area are repeated.
7. The method for detecting natural resource element change based on deep learning according to claim 1, wherein the repeated element elimination of the natural resource elements in the natural resource element merged graph includes the following steps:
step 1: acquiring repeated natural resource element areas in an upper layer of the natural resource element segmentation chart and a lower layer of the natural resource element segmentation chart in the detection area;
step 2: respectively determining the range of the repeated natural resource element region in the upper layer of the natural resource element segmentation map and the lower layer of the natural resource element segmentation map in the detection region;
and step 3: and repeating element elimination is carried out on the natural resource elements related to the range of the repeated natural resource element areas in the upper layer of the natural resource element segmentation graph and the lower layer of the natural resource element segmentation graph in the detection area.
8. The method for detecting natural resource element change based on deep learning according to claim 1, further comprising:
carrying out image smoothing treatment on the image to be detected to obtain a better image to be detected;
and carrying out image smoothing treatment on the natural resource element change graph to obtain a better natural resource element change graph.
9. The method for detecting natural resource element change based on deep learning according to claim 1, further comprising:
dividing the image to be detected into a plurality of areas to be detected according to a preset division rule;
acquiring actual change data of natural resource elements in the same to-be-detected region in the two to-be-detected images, wherein the actual change data comprises: the shooting time of the two images to be detected and the variety change condition of natural resource elements are determined by taking one of the two images to be detected with longer shooting time as a standard diagram, taking any one of the standard diagram as a standard region to be detected, dividing the standard region to be detected into a central region to be detected, an east region to be detected, a west region to be detected, a south region to be detected and a north region to be detected according to a preset second division rule, taking one of the two images with shorter shooting time as a comparison diagram, taking the region to be detected corresponding to the standard region to be detected in the comparison diagram as a comparison region to be detected, and performing superposition comparison on the central region to be detected, the east region to be detected, the west region to be detected, the south region to be detected and the north region to be detected and the comparison region to be detected, obtaining the range difference of a central region to be detected, the range difference of an east region to be detected, the range difference of a west region to be detected, the range difference of a south region to be detected and the range difference of a north region to be detected;
calculating actual change values of different natural resource elements in any to-be-detected region in different shooting intervals under the condition that the types of the natural resource elements are not changed based on the actual change data, wherein the calculation formula is as follows:
Figure FDA0003079421800000051
wherein f ist，δ，εThe actual change value S of the epsilon-th natural resource element in the delta-th to-be-detected area in the interval time t is shot under the condition that the type of the natural resource element is not changedz，δThe central detected region range difference of the delta-th detected region, Sd，δEast to-be-detected region range difference of the δ -th to-be-detected region, Sn，δSouth to-be-detected region range difference S of the delta-th to-be-detected regionx，δWest zone difference, S, for the δ -th zoneb，δIs the north area-to-be-detected range difference mu of the delta area-to-be-detectedz，ε、μd，ε、μn，ε、μx，εAnd mub，εDifferent range error coefficients corresponding to the preset epsilon natural resource elements are respectively set;
if the actual change value of different natural resource elements in any to-be-detected area in different shooting intervals is larger than the ideal change value of the corresponding natural resource element in the preset corresponding shooting interval under the condition that the types of the natural resource elements are not changed, carrying out artificial intervention marking on any to-be-detected area corresponding to the situation that the actual change value of the different natural resource elements in any to-be-detected area in the different shooting intervals is larger than the ideal change value of the corresponding natural resource element in the preset corresponding shooting intervals;
and if the actual change value of different natural resource elements in any to-be-detected area in different shooting intervals is smaller than or equal to the ideal change value of the corresponding natural resource element in the preset corresponding shooting interval under the condition that the types of the natural resource elements are not changed, performing natural change marking on any to-be-detected area corresponding to the situation that the actual change value of the different natural resource elements in any to-be-detected area in the different shooting intervals is larger than the ideal change value of the corresponding natural resource element in the preset corresponding shooting intervals.
10. The method for detecting natural resource element change based on deep learning according to claim 9, further comprising:
when the variety of the natural resource elements in any region to be detected changes, carrying out artificial intervention marking on the region to be detected corresponding to the change of the variety of the natural resource elements;
wherein the changing of the type of the natural resource element includes: the variety of the natural resource elements is changed, the natural resource elements are lost, and the natural resource elements are increased.
CN202110562234.5A 2021-05-24 2021-05-24 Natural resource element change detection method based on deep learning Active CN113362286B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110562234.5A CN113362286B (en) 2021-05-24 2021-05-24 Natural resource element change detection method based on deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110562234.5A CN113362286B (en) 2021-05-24 2021-05-24 Natural resource element change detection method based on deep learning

Publications (2)

Publication Number Publication Date
CN113362286A true CN113362286A (en) 2021-09-07
CN113362286B CN113362286B (en) 2022-02-01

Family

ID=77527307

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110562234.5A Active CN113362286B (en) 2021-05-24 2021-05-24 Natural resource element change detection method based on deep learning

Country Status (1)

Country Link
CN (1) CN113362286B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115620031A (en) * 2022-12-12 2023-01-17 广东新禾道信息科技有限公司 Method, system and equipment for processing natural resource right-confirming registration information

Citations (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101174299A (en) * 2007-11-16 2008-05-07 武汉大学 Method for detecting change of water body and settlement place based on aviation video
CN101694718A (en) * 2009-10-13 2010-04-14 西安电子科技大学 Method for detecting remote sensing image change based on interest areas
CN102169584A (en) * 2011-05-28 2011-08-31 西安电子科技大学 Remote sensing image change detection method based on watershed and treelet algorithms
CN103426158A (en) * 2012-05-17 2013-12-04 中国科学院电子学研究所 Method for detecting two-time-phase remote sensing image change
CN108492298A (en) * 2018-04-13 2018-09-04 西安电子科技大学 Based on the multispectral image change detecting method for generating confrontation network
US20190050625A1 (en) * 2017-08-08 2019-02-14 Spaceknow Inc. Systems, methods and computer program products for multi-resolution multi-spectral deep learning based change detection for satellite images
CN109871875A (en) * 2019-01-21 2019-06-11 大连理工大学 A kind of building change detecting method based on deep learning
CN111199550A (en) * 2020-04-09 2020-05-26 腾讯科技（深圳）有限公司 Training method, segmentation method, device and storage medium of image segmentation network
CN111598174A (en) * 2020-05-19 2020-08-28 中国科学院空天信息创新研究院 Training method of image ground feature element classification model, image analysis method and system
CN112085738A (en) * 2020-08-14 2020-12-15 南京邮电大学 Image segmentation method based on generation countermeasure network
CN112330685A (en) * 2020-12-28 2021-02-05 北京达佳互联信息技术有限公司 Image segmentation model training method, image segmentation device and electronic equipment
CN112818995A (en) * 2021-01-27 2021-05-18 北京达佳互联信息技术有限公司 Image classification method and device, electronic equipment and storage medium

Patent Citations (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101174299A (en) * 2007-11-16 2008-05-07 武汉大学 Method for detecting change of water body and settlement place based on aviation video
CN101694718A (en) * 2009-10-13 2010-04-14 西安电子科技大学 Method for detecting remote sensing image change based on interest areas
CN102169584A (en) * 2011-05-28 2011-08-31 西安电子科技大学 Remote sensing image change detection method based on watershed and treelet algorithms
CN103426158A (en) * 2012-05-17 2013-12-04 中国科学院电子学研究所 Method for detecting two-time-phase remote sensing image change
US20190050625A1 (en) * 2017-08-08 2019-02-14 Spaceknow Inc. Systems, methods and computer program products for multi-resolution multi-spectral deep learning based change detection for satellite images
CN108492298A (en) * 2018-04-13 2018-09-04 西安电子科技大学 Based on the multispectral image change detecting method for generating confrontation network
CN109871875A (en) * 2019-01-21 2019-06-11 大连理工大学 A kind of building change detecting method based on deep learning
CN111199550A (en) * 2020-04-09 2020-05-26 腾讯科技（深圳）有限公司 Training method, segmentation method, device and storage medium of image segmentation network
CN111598174A (en) * 2020-05-19 2020-08-28 中国科学院空天信息创新研究院 Training method of image ground feature element classification model, image analysis method and system
CN112085738A (en) * 2020-08-14 2020-12-15 南京邮电大学 Image segmentation method based on generation countermeasure network
CN112330685A (en) * 2020-12-28 2021-02-05 北京达佳互联信息技术有限公司 Image segmentation model training method, image segmentation device and electronic equipment
CN112818995A (en) * 2021-01-27 2021-05-18 北京达佳互联信息技术有限公司 Image classification method and device, electronic equipment and storage medium

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
杨希 等: ""基于多时相遥感图像灰度差值法的地表变化检侧"", 《四川测绘》 *
陈璐 等: ""基于深度学习的城市高分遥感图像变化检测方法的研究"", 《计算机应用研究》 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115620031A (en) * 2022-12-12 2023-01-17 广东新禾道信息科技有限公司 Method, system and equipment for processing natural resource right-confirming registration information

Also Published As

Publication number Publication date
CN113362286B (en) 2022-02-01

Similar Documents

Publication Publication Date Title
CN104182985B (en) Remote sensing image change detection method
CN105574550A (en) Vehicle identification method and device
CN111723732A (en) Optical remote sensing image change detection method, storage medium and computing device
San et al. Building extraction from high resolution satellite images using Hough transform
CN108492298B (en) Multispectral image change detection method based on generation countermeasure network
CN107507170A (en) A kind of airfield runway crack detection method based on multi-scale image information fusion
CN112489054A (en) Remote sensing image semantic segmentation method based on deep learning
CN106295502A (en) A kind of method for detecting human face and device
CN113689445B (en) High-resolution remote sensing building extraction method combining semantic segmentation and edge detection
CN108171119B (en) SAR image change detection method based on residual error network
CN110807485B (en) Method for fusing two-classification semantic segmentation maps into multi-classification semantic map based on high-resolution remote sensing image
CN106096555A (en) The method and apparatus of three dimensional face detection
CN109544468A (en) A kind of image data amplification method
CN116206185A (en) Lightweight small target detection method based on improved YOLOv7
CN105389799A (en) SAR image target detection method based on sketch and low-rank decomposition
CN113362286B (en) Natural resource element change detection method based on deep learning
CN113870254B (en) Target object detection method and device, electronic equipment and storage medium
CN113516060B (en) Building space fusion discrimination method and device based on multi-temporal high-resolution data
WO2018042208A1 (en) Street asset mapping
CN112800934B (en) Behavior recognition method and device for multi-class engineering vehicle
CN109726637A (en) The parking site detecting system and method in parking stall region are indicated with resilient polygonal
CN113378905A (en) Small target detection method based on distribution distance
CN104112145A (en) Facial expression recognition method based on PWLD and D-S evidence theory
CN104463896A (en) Image corner point detection method and system based on kernel similar region distribution characteristics
CN110458023A (en) A kind of training method of road detection model, road detection method and device

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
CB02 Change of applicant information

Address after: 224000 room 209, building 1, R & D building Park, No. 69, Donghuan South Road, Yancheng Economic and Technological Development Zone, Jiangsu Province

Applicant after: JIANGSU XINGYUE SURVEYING AND MAPPING TECHNOLOGY CO.,LTD.

Address before: 224000 floors 9-10, building 23, Hai * building, No. 68, hope Avenue Middle Road, Tinghu District, Yancheng City, Jiangsu Province

Applicant before: JIANGSU XINGYUE SURVEYING AND MAPPING TECHNOLOGY CO.,LTD.

CB02 Change of applicant information
GR01 Patent grant
GR01 Patent grant