CN113191271A - Method for detecting surface particulate matter of weir dam based on deep learning - Google Patents
Method for detecting surface particulate matter of weir dam based on deep learning Download PDFInfo
- Publication number
- CN113191271A CN113191271A CN202110484329.XA CN202110484329A CN113191271A CN 113191271 A CN113191271 A CN 113191271A CN 202110484329 A CN202110484329 A CN 202110484329A CN 113191271 A CN113191271 A CN 113191271A
- Authority
- CN
- China
- Prior art keywords
- deep learning
- weir dam
- particulate matter
- image data
- data set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000013135 deep learning Methods 0.000 title claims abstract description 32
- 238000000034 method Methods 0.000 title claims abstract description 32
- 239000013618 particulate matter Substances 0.000 title claims abstract description 32
- 238000012549 training Methods 0.000 claims abstract description 35
- 239000002245 particle Substances 0.000 claims abstract description 28
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 27
- 239000002344 surface layer Substances 0.000 claims abstract description 26
- 238000004364 calculation method Methods 0.000 claims abstract description 23
- 239000000126 substance Substances 0.000 claims abstract description 19
- 238000005259 measurement Methods 0.000 claims abstract description 14
- 238000012360 testing method Methods 0.000 claims abstract description 12
- 238000013136 deep learning model Methods 0.000 claims abstract description 11
- 238000007781 pre-processing Methods 0.000 claims abstract description 6
- 230000003321 amplification Effects 0.000 claims description 6
- 238000002372 labelling Methods 0.000 claims description 6
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 6
- 238000013461 design Methods 0.000 claims description 5
- 238000005286 illumination Methods 0.000 claims description 4
- 238000009499 grossing Methods 0.000 claims description 3
- 238000010008 shearing Methods 0.000 claims description 3
- 238000013519 translation Methods 0.000 claims description 3
- 230000000007 visual effect Effects 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 230000002708 enhancing effect Effects 0.000 claims description 2
- 238000012545 processing Methods 0.000 claims description 2
- 238000001514 detection method Methods 0.000 abstract description 19
- 230000008569 process Effects 0.000 abstract description 7
- 238000013527 convolutional neural network Methods 0.000 description 9
- 239000010410 layer Substances 0.000 description 4
- 230000006872 improvement Effects 0.000 description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 2
- 238000003491 array Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 239000004575 stone Substances 0.000 description 2
- 238000012916 structural analysis Methods 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 239000002689 soil Substances 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- CCEKAJIANROZEO-UHFFFAOYSA-N sulfluramid Chemical group CCNS(=O)(=O)C(F)(F)C(F)(F)C(F)(F)C(F)(F)C(F)(F)C(F)(F)C(F)(F)C(F)(F)F CCEKAJIANROZEO-UHFFFAOYSA-N 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000009827 uniform distribution Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- 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
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30242—Counting objects in image
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Multimedia (AREA)
- Geometry (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a method for detecting surface-layer particulate matter of a weir dam based on deep learning, which comprises the following steps: acquiring a color image of the surface granular substance of the weir dam, and establishing a training image data set and a test image data set; preprocessing a training image data set to form a standard training image data set; marking the surface granular substance of the damming dam in the standard training image data set to generate a marked file set; training the standard training image data set and the labeled file set based on a deep learning algorithm to generate a deep learning model; carrying out target identification on the surface particulate matter of the damming dam in the test image data set by using a deep learning model; carrying out particle size measurement and calculation on the identification target based on a three-dimensional reconstruction algorithm; and evaluating the model by adopting the identification precision and the particle size measuring and calculating precision. The method provided by the invention is used for realizing automatic detection of the surface granular substance of the weir dam, and has the characteristics of simple process, reliable calculation, high identification speed, high accuracy and strong robustness.
Description
Technical Field
The invention relates to the technical field of emergency rescue of a damming dam, in particular to a method for detecting surface-layer particulate matters of the damming dam based on deep learning.
Background
The damming dam is a special dam body formed by water storage after intercepting valleys and rivers due to landslide, collapse, debris flow, volcanic karst flow, tillite and the like caused by rainfall, earthquake, volcano, glacier activity and the like, and has the characteristics of large dam body square amount, large water storage amount, large safety threat and the like. The structural characteristics of the formed damming dam are important data bases for stability analysis and emergency rescue. The structure of the damming dam depends on the material source, the landform shape of the river valley and the motion accumulation process, and the material composition of the damming dam generally has the characteristics of uneven soil and stone distribution, uneven particle size of the lump stones, uneven compactness and the like, so that the difficulty in effectively detecting the damming dam is increased.
At present, the detection of the surface granular substances of the weir dam mainly adopts a manual mode, the automation degree is low, the timeliness is poor, and the requirement of emergency rescue is difficult to meet. Therefore, a deep learning-based method for detecting the particulate matter on the surface layer of the weir dam needs to be researched. In recent years, with the development of convolutional neural networks and the improvement of hardware computing power, deep learning-based target detection has made a breakthrough progress and has been widely applied in various fields of computer vision, and many advanced target detection algorithms use deep learning networks as backbone networks and detection networks thereof to extract features from input images or videos. The target detection method based on deep learning comprises two types: one is a two-stage target detection method, i.e., candidate regions are generated first, and then the candidate regions are classified and regressed, such as FasterR-CNN; the other is a one-stage target detection method, i.e. the object type and coordinates are directly regressed from the image without generating a candidate frame, and representative methods are YOLO, SSD, and the like.
Disclosure of Invention
The invention aims to provide a method for detecting the surface granular substance of the weir dam based on deep learning, which is used for realizing the automatic detection of the surface granular substance of the weir dam and has the characteristics of simple process, reliable calculation, high identification speed, high accuracy and strong robustness.
In order to achieve the purpose, the invention provides the following scheme:
a method for detecting surface particulate matter of a weir dam based on deep learning comprises the following steps:
s1) a remote sensing device is used for carrying a digital camera to obtain color images of the surface layer particulate matter of the weir dam under different conditions, and a training image data set and a test image data set are established;
s2) preprocessing the training image data set to form a standard training image data set;
s3) manually labeling the surface layer particulate matter of the weir dam in the standard training image data set according to the selected label to generate a labeled file set;
s4) training the standard training image data set obtained in the step S2) and the labeled file set obtained in the step S3) based on a deep learning algorithm, and generating a deep learning model after training is completed;
s5) carrying out target identification on the surface particulate matter of the weir dam in the test image data set obtained in the step S1) by using a deep learning model, and calculating identification precision;
s6) carrying out particle size measurement and calculation on the identification target based on a three-dimensional reconstruction algorithm, and calculating the particle size measurement and calculation precision;
s7) using the two indices of the recognition accuracy calculated in step S5) and the particle diameter measurement accuracy calculated in step S6), the deep learning model is evaluated.
Optionally, in step S1), the remote sensing device is a satellite, a manned machine, an unmanned aerial vehicle or a hot air balloon; the different conditions comprise different illumination conditions, meteorological conditions, navigation height conditions, shooting angles, landform backgrounds and weir dam forming reasons; there is no duplication or intersection between the training image dataset and the test image dataset.
Optionally, the preprocessing in step S2) includes:
enhancing the image to improve the visual effect of the image;
data amplification is carried out to increase the number of samples of the image;
the size specification is carried out to unify the size of the image;
the format is standardized, and the format of the image is unified.
Optionally, the image enhancement adopts one or more of smoothing, sharpening, contrast enhancement, brightness enhancement, saturation enhancement and color conversion processing; the data amplification adopts one or more of rotation, translation, shearing, zooming and overturning; the size specification takes 1024 pixels by 1024 pixels or 512 pixels by 512 pixels; the format specification adopts one of JPG, PNG, GIF and TIF.
Optionally, the selected label in step S3) classifies the particulate matter on the surface of the weir dam into one category, or designs multiple categories according to different conditions in step S1), or performs the categories according to certain pixel criteria.
Optionally, in step S4), the deep learning algorithm is one of two stages of fast R-CNN, Mask R-CNN, Cascade R-CNN and TridentNet algorithms or one of one stage of YOLO, SSD, DSSD, RetinaNet, CornerNet and CenterNet algorithms.
Optionally, the identification precision in step S7) includes two indexes, i.e., an accuracy P and a recall R, and the calculation formulas are respectively:
wherein TP is the number of correctly detected surface layer particulate matters of the weir dam, TPFP is the number of all detected surface layer particulate matters of the weir dam, and TPFN is the number of surface layer particulate matters of the weir dam which should be detected;
the particle size measurement and calculation precision comprises an error x of the particle sizeiAverage value ofAnd standard deviation sigma, the calculation formula is respectively as follows:
xi=Lm-Lt,
wherein L ismFor measuring and calculating the particle size, L, of a certain object to be identifiedtThe number n is the number of the targets to be identified.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects: the method for detecting the surface granular substances of the weir dam based on deep learning provided by the invention designs a complete set of complete technical process and scheme, can realize identification, positioning and particle size measurement and calculation of the surface granular substances of the weir dam under complex environmental conditions, has the characteristics of simple process, reliable calculation, high identification speed, high accuracy and strong robustness, reduces the workload of a manual detection mode commonly adopted in the past, and provides good data basis and decision basis for structural analysis and emergency rescue and disaster relief of the subsequent weir dam.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
Fig. 1 is a flowchart of a method for detecting particulate matter on the surface of a weir dam based on deep learning according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a method for detecting the surface granular substance of the weir dam based on deep learning, which is used for realizing the automatic detection of the surface granular substance of the weir dam and has the characteristics of simple process, reliable calculation, high identification speed, high accuracy and strong robustness.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Fig. 1 is a flowchart of a method for detecting particulate matter on a surface layer of a weir dam based on deep learning according to an embodiment of the present invention, and as shown in fig. 1, the method for detecting particulate matter on a surface layer of a weir dam based on deep learning according to an embodiment of the present invention includes the following steps:
s1) carrying a digital camera by using a remote sensing device, acquiring color images of the surface granular substances of the weir dam under different conditions under the condition of natural illumination, and establishing a training image data set and a test image data set; the remote sensing device adopts a satellite, a man-machine, an unmanned aerial vehicle or a hot air balloon, and can also directly adopt a hand-held camera mode; when an image is collected, image samples under different conditions are selected as much as possible to enrich the types of the samples and improve the quality of model training, wherein the different conditions include but are not limited to different illumination conditions (morning, noon, evening and the like), meteorological conditions (clear, cloudy, foggy and the like), navigation conditions (set according to the precision of a detection target and the ground size GSD represented by one pixel), shooting angles, landform backgrounds and weir dam forming reasons (types of mountain landslides, earthquakes, lava and the like); no repetition or intersection exists between the training image data set and the testing image data set;
s2) preprocessing the training image data set to form a standard training image data set so as to enhance identifiability of related information, weaken unrelated information in the image and improve adaptability of the model to data; the pretreatment comprises the following steps:
image enhancement to improve the visual effect of images, usually applied to the case of poor visibility conditions; the image enhancement adopts but is not limited to smoothing (neighborhood averaging, median filtering, multi-image averaging, and the like), sharpening, contrast enhancement, brightness enhancement, saturation enhancement, and color transformation;
data amplification is carried out to increase the number of samples of the image, effectively relieve the overfitting condition of the model and bring stronger generalization capability to the model; the data amplification adopts but is not limited to rotation, translation, shearing, zooming and overturning;
the size specification is carried out to unify the size of the image; the size specification employs, but is not limited to, 1024 pixels by 1024 pixels or 512 pixels by 512 pixels;
the format is standard, and the formats of the images are unified; the format specification adopts but is not limited to JPG, PNG, GIF, TIF;
s3) manually labeling the surface layer particulate matter of the weir dam in the standard training image data set according to the selected label to generate a labeled file set; annotation software such as, but not limited to, LabelImg, LabelMe, annotation form such as, but not limited to, rectangle, polygon, circle; the selected label classifies the surface granular substances of the weir dam into one class, or designs multi-classification according to different conditions in the step S1), or can classify according to certain pixel standard; for example, two types of particulate matter are defined for the surface of a weir dam: a large target, and a small target, where a large target is defined as a particle size pixel of the particulate matter greater than α, and a small target is defined as a particle size pixel of α or less, and the parameter α may be an absolute value (such as, but not limited to, 32 pixels) or a relative value (such as, but not limited to, 0.1 of the width or height of the image size); when labeling, a certain number of negative samples are generally set, and the types of the negative samples are such as, but not limited to, water bodies, roads and vegetation; after each training image is labeled, a labeling file (such as but not limited to XML and JSON formats) is generated, and the area coordinates of each label type in the image are recorded;
s4) training the standard training image data set obtained in the step S2) and the labeled file set obtained in the step S3) based on a deep learning algorithm, and generating a deep learning model after training is completed; the deep learning algorithm is one of two-stage algorithms of fast R-CNN, Mask R-CNN, Cascade R-CNN, TridentNet and the like or one of one-stage algorithms of YOLO, SSD, DSSD, RetinaNet, CornerNet, CenterNet and the like, and the target detection of the surface granular substances of the weir dam by utilizing the algorithms and the improvement thereof is within the protection range of the invention and is not repeated herein;
in the embodiment of the invention, the SSD algorithm is taken as an example to explain the improvement condition of the deep learning algorithm; the SSD algorithm uses a multi-scale feature layer for detection by using the idea of a YOLO grid and an anchor mechanism of FasterR-CNN for reference, so that the SSD can quickly predict and relatively accurately acquire the position of a target; the anchor arrays of the feature layers are increased from small to large so as to detect targets with different scales; the front characteristic layer is used for detecting a small target, and the rear characteristic layer is used for detecting a large target;
the anchors can be set in a proportional and uniform distribution manner, taking the width of the training image as an example, assuming that the width is W (unit: pixel), the minimum anchor is set as W0.1, the maximum anchor is set as W0.95, and the intermediate anchors are uniformly distributed between W0.1-W0.95; the anchor array can also be set by statistical analysis. Selecting a certain number of images, recording the pixel size of the granular substances on the surface layer of the weir dam to form a set M, taking the width of a rectangular frame as an example, and then calculating an anchor array suitable for the granular substances on the surface layer of the weir dam by adopting a clustering algorithm in the field of data mining; here, the K-means algorithm is taken as an example, and the steps are as follows:
(1) setting the number of anchor arrays as N, namely the number of clusters;
(2) sorting the set M from small to large, wherein the minimum value is MminMaximum value of MmaxAt MminAnd MmaxRandomly selecting N values as a clustering center, thus forming N clusters;
(3) for each sample value s in the set M, respectively calculating the distance from the sample value s to N clustering centers, and dividing s into clusters where the clustering centers closest to the sample value s are located;
(4) recalculating the center of each cluster, and adopting an average value method;
(5) returning to the step (2) for iterative operation until the clustering center of each cluster tends to converge;
after the iterative operation, an anchor array which is relatively suitable for the surface granular substance of the weir dam can be obtained, and the quality and the migration capability of the training model can be improved;
s5) carrying out target identification on the surface particulate matter of the weir dam in the test image data set obtained in the step S1) by using a deep learning model, and calculating identification precision; performing frame selection marking on the target according to the set reliability, wherein the frame selection area is the detected surface layer particulate matter of the weir dam, and recording the pixel coordinates of the detected surface layer particulate matter;
s6) carrying out particle size measurement and calculation on the identification target based on a three-dimensional reconstruction algorithm, and calculating the particle size measurement and calculation precision; on the basis of the known image position and attitude parameters of the unmanned aerial vehicle (or other remote sensing devices), the three-dimensional reconstruction algorithm in computer vision or digital photogrammetry is utilized, the target identification result (namely the frame-selected pixel coordinates) in the step S5) is used as input, the solid dimension of the identification target is calculated, and the particle size information (such as width and height) of the surface layer particulate matter of the weir dam can be obtained; wherein, the three-dimensional reconstruction algorithm is such as but not limited to a front intersection algorithm, a bundle adjustment, SFM;
s7) evaluating the deep learning model by adopting two indexes of the identification precision calculated in the step S5) and the particle size measuring and calculating precision calculated in the step S6); the identification precision comprises two indexes of accuracy rate P (precision) and recall rate R (recall), and the calculation formulas are respectively as follows:
wherein TP is the number of correctly detected surface layer particulate matters of the weir dam, TPFP is the number of all detected surface layer particulate matters of the weir dam, and TPFN is the number of surface layer particulate matters of the weir dam which should be detected; the value range of the accuracy P is 0-1, the false alarm condition in the detection result is reflected, and the higher the value is, the higher the accuracy is, the less the false alarm is; the recall ratio R ranges from 0 to 1, the report missing condition in the detection result is reflected, and the higher the value is, the less the report missing is;
the particle size measurement and calculation precision comprises an error x of the particle sizeiAverage value ofAnd standard deviation sigma, the calculation formula is respectively as follows:
xi=Lm-Lt,
wherein L ismFor measuring and calculating the particle size, L, of a certain object to be identifiedtThe number n is the number of the targets to be identified; the average x reflects the average level of the particle size error, and the labeling difference σ reflects the dispersion degree of the particle size error.
The embodiment of the invention provides a method for detecting surface particulate matter of a weir dam based on deep learning, which comprises the steps of selecting 1500 training images with the size of 1024 x 1024 pixels, marking 10 detection targets on each image on average, selecting 6 test images for experiment, and obtaining a detection result: the image recognition accuracy of the model reaches 95.67%, the recall rate reaches 89%, and the recognition precision reaches 2 cm.
The method for detecting the surface granular substances of the weir dam based on deep learning provided by the invention designs a complete set of complete technical process and scheme, can realize identification, positioning and particle size measurement and calculation of the surface granular substances of the weir dam under complex environmental conditions, has the characteristics of simple process, reliable calculation, high identification speed, high accuracy and strong robustness, reduces the workload of a manual detection mode commonly adopted in the past, and provides good data basis and decision basis for structural analysis and emergency rescue and disaster relief of the subsequent weir dam.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (7)
1. A method for detecting surface particulate matter of a weir dam based on deep learning is characterized by comprising the following steps:
s1) a remote sensing device is used for carrying a digital camera to obtain color images of the surface layer particulate matter of the weir dam under different conditions, and a training image data set and a test image data set are established;
s2) preprocessing the training image data set to form a standard training image data set;
s3) manually labeling the surface layer particulate matter of the weir dam in the standard training image data set according to the selected label to generate a labeled file set;
s4) training the standard training image data set obtained in the step S2) and the labeled file set obtained in the step S3) based on a deep learning algorithm, and generating a deep learning model after training is completed;
s5) carrying out target identification on the surface particulate matter of the weir dam in the test image data set obtained in the step S1) by using a deep learning model, and calculating identification precision;
s6) carrying out particle size measurement and calculation on the identification target based on a three-dimensional reconstruction algorithm, and calculating the particle size measurement and calculation precision;
s7) using the two indices of the recognition accuracy calculated in step S5) and the particle diameter measurement accuracy calculated in step S6), the deep learning model is evaluated.
2. The method for detecting the granular substances on the surface layer of the weir dam based on the deep learning of claim 1, wherein in the step S1), the remote sensing device is a satellite, a manned machine, an unmanned aerial vehicle or a hot air balloon; the different conditions comprise different illumination conditions, meteorological conditions, navigation height conditions, shooting angles, landform backgrounds and weir dam forming reasons; there is no duplication or intersection between the training image dataset and the test image dataset.
3. The method for detecting particulate matter on the surface of a weir dam based on deep learning of claim 1, wherein the preprocessing in step S2) comprises:
enhancing the image to improve the visual effect of the image;
data amplification is carried out to increase the number of samples of the image;
the size specification is carried out to unify the size of the image;
the format is standardized, and the format of the image is unified.
4. The method for detecting the particulate matter on the surface layer of the weir dam based on the deep learning of claim 3, wherein the image enhancement adopts one or more of smoothing, sharpening, contrast enhancement, brightness enhancement, saturation enhancement and color conversion processing; the data amplification adopts one or more of rotation, translation, shearing, zooming and overturning; the size specification takes 1024 pixels by 1024 pixels or 512 pixels by 512 pixels; the format specification adopts one of JPG, PNG, GIF and TIF.
5. The method for detecting particulate matter on the surface of a weir dam based on deep learning of claim 1, wherein the selected label in step S3) classifies the particulate matter on the surface of the weir dam into one class, or designs multiple classes according to different situations in step S1), or classifies the particulate matter according to certain pixel criteria.
6. The method for detecting particulate matter on the surface of a weir dam based on deep learning of claim 1, wherein the deep learning algorithm in step S4) is one of two-stage fast R-CNN, Mask R-CNN, Cascade R-CNN and TridentNet algorithms or one of one-stage YOLO, SSD, DSSD, RetinaNet, CornerNet and cenerten algorithms.
7. The method for detecting particulate matter on the surface of a weir dam based on deep learning of claim 1, wherein the identification precision in step S7) includes two indexes of accuracy P and recall R, and the calculation formulas are respectively:
wherein TP is the number of correctly detected surface layer particulate matters of the weir dam, TPFP is the number of all detected surface layer particulate matters of the weir dam, and TPFN is the number of surface layer particulate matters of the weir dam which should be detected;
the particle size measurement and calculation precision comprises an error x of the particle sizeiAverage value ofAnd standard deviation sigma, the calculation formula is respectively as follows:
xi=Lm-Lt,
wherein L ismFor measuring and calculating the particle size, L, of a certain object to be identifiedtThe number n is the number of the targets to be identified.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110484329.XA CN113191271A (en) | 2021-04-30 | 2021-04-30 | Method for detecting surface particulate matter of weir dam based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110484329.XA CN113191271A (en) | 2021-04-30 | 2021-04-30 | Method for detecting surface particulate matter of weir dam based on deep learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113191271A true CN113191271A (en) | 2021-07-30 |
Family
ID=76983689
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110484329.XA Pending CN113191271A (en) | 2021-04-30 | 2021-04-30 | Method for detecting surface particulate matter of weir dam based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113191271A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114331160A (en) * | 2021-12-30 | 2022-04-12 | 四川大学 | Damming dam burst disaster chain mode identification method based on landslide and river plugging form |
CN115346114A (en) * | 2022-07-21 | 2022-11-15 | 中铁二院工程集团有限责任公司 | Method and equipment for identifying and positioning bad geologic body by railway tunnel aviation electromagnetic method |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109801282A (en) * | 2019-01-24 | 2019-05-24 | 湖北大学 | Pavement behavior detection method, processing method, apparatus and system |
CN110047072A (en) * | 2019-04-30 | 2019-07-23 | 福建南方路面机械有限公司 | A kind of gravel size identification processing system and processing method based on mobile interchange |
CN110070537A (en) * | 2019-04-25 | 2019-07-30 | 清华大学 | The granularity of still image particle and the intelligent identification Method of sphericity and device |
CN110211173A (en) * | 2019-04-03 | 2019-09-06 | 中国地质调查局发展研究中心 | A kind of paleontological fossil positioning and recognition methods based on deep learning |
CN110390691A (en) * | 2019-06-12 | 2019-10-29 | 合肥合工安驰智能科技有限公司 | A kind of ore scale measurement method and application system based on deep learning |
CN111079847A (en) * | 2019-12-20 | 2020-04-28 | 郑州大学 | Remote sensing image automatic labeling method based on deep learning |
CN111161292A (en) * | 2019-11-21 | 2020-05-15 | 合肥合工安驰智能科技有限公司 | Ore size measurement method and application system |
-
2021
- 2021-04-30 CN CN202110484329.XA patent/CN113191271A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109801282A (en) * | 2019-01-24 | 2019-05-24 | 湖北大学 | Pavement behavior detection method, processing method, apparatus and system |
CN110211173A (en) * | 2019-04-03 | 2019-09-06 | 中国地质调查局发展研究中心 | A kind of paleontological fossil positioning and recognition methods based on deep learning |
CN110070537A (en) * | 2019-04-25 | 2019-07-30 | 清华大学 | The granularity of still image particle and the intelligent identification Method of sphericity and device |
CN110047072A (en) * | 2019-04-30 | 2019-07-23 | 福建南方路面机械有限公司 | A kind of gravel size identification processing system and processing method based on mobile interchange |
CN110390691A (en) * | 2019-06-12 | 2019-10-29 | 合肥合工安驰智能科技有限公司 | A kind of ore scale measurement method and application system based on deep learning |
CN111161292A (en) * | 2019-11-21 | 2020-05-15 | 合肥合工安驰智能科技有限公司 | Ore size measurement method and application system |
CN111079847A (en) * | 2019-12-20 | 2020-04-28 | 郑州大学 | Remote sensing image automatic labeling method based on deep learning |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114331160A (en) * | 2021-12-30 | 2022-04-12 | 四川大学 | Damming dam burst disaster chain mode identification method based on landslide and river plugging form |
CN114331160B (en) * | 2021-12-30 | 2023-04-28 | 四川大学 | Dam blocking and dam bursting disaster chain mode identification method based on landslide river blocking form |
CN115346114A (en) * | 2022-07-21 | 2022-11-15 | 中铁二院工程集团有限责任公司 | Method and equipment for identifying and positioning bad geologic body by railway tunnel aviation electromagnetic method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Yu et al. | A real-time detection approach for bridge cracks based on YOLOv4-FPM | |
US11561092B2 (en) | Method for measuring antenna downtilt angle based on multi-scale deep semantic segmentation network | |
Zhang et al. | Fusion of images and point clouds for the semantic segmentation of large-scale 3D scenes based on deep learning | |
Yang et al. | Spatial and spectral hybrid image classification for rice lodging assessment through UAV imagery | |
Iqbal et al. | How computer vision can facilitate flood management: A systematic review | |
CN113191271A (en) | Method for detecting surface particulate matter of weir dam based on deep learning | |
Bai et al. | An intelligent water level monitoring method based on SSD algorithm | |
Mohammadi et al. | Deep learning-based damage detection from aerial SfM point clouds | |
CN112766417A (en) | Method and system for recognizing current land type of land utilization of target land block by using field photo | |
Yousef et al. | Shoreline extraction from light detection and ranging digital elevation model data and aerial images | |
CN116665153A (en) | Road scene segmentation method based on improved deep bv3+ network model | |
CN110853034A (en) | Crack detection method, crack detection device, electronic equipment and computer-readable storage medium | |
CN113469097B (en) | Multi-camera real-time detection method for water surface floaters based on SSD network | |
Weidner et al. | An algorithm for measuring landslide deformation in terrestrial lidar point clouds using trees | |
CN113553949A (en) | Tailing pond semantic segmentation method based on photogrammetric data | |
CN112085725A (en) | Residual film residual quantity detection method and early warning system based on heuristic iterative algorithm | |
CN112581301A (en) | Detection and early warning method and system for residual film quantity of farmland based on deep learning | |
Kim et al. | Surface sediment classification using a deep learning model and unmanned aerial vehicle data of tidal flats | |
Said et al. | Desertification Detection in Makkah Region based on Aerial Images Classification. | |
Liu et al. | Identification of Damaged Building Regions from High-Resolution Images Using Superpixel-Based Gradient and Autocorrelation Analysis | |
Li et al. | Mapping Water Bodies Using SAR Imagery-an Application Over the Spiritwood Valley Aquifier, Manitoba | |
CN117456287B (en) | Method for observing population number of wild animals by using remote sensing image | |
Albalawi | Comparing pixel-based to object-based image classifications for assessing lulc change in an arid environment of northern west saudi arabia | |
So et al. | Deep-learning-based Automatic Detection and Classification of Traffic Signs Using Images Collected by Mobile Mapping Systems. | |
Kumari et al. | Comparative Performance of Maximum Likelihood and Minimum Distance Classifiers on Land Use and Land Cover Analysis of Varanasi District (India) |
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 |