CN115908425B - Edge detection-based rock-fill grading information detection method - Google Patents
Edge detection-based rock-fill grading information detection method Download PDFInfo
- Publication number
- CN115908425B CN115908425B CN202310109315.9A CN202310109315A CN115908425B CN 115908425 B CN115908425 B CN 115908425B CN 202310109315 A CN202310109315 A CN 202310109315A CN 115908425 B CN115908425 B CN 115908425B
- Authority
- CN
- China
- Prior art keywords
- rock
- fill
- point
- point cloud
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Abstract
The invention discloses a rockfill grading information detection method based on edge detection, which comprises four steps of segmentation model generation and rockfill grading acquisition, wherein the segmentation model generation comprises training data set construction, segmentation model training and segmentation model debugging, and the rockfill grading acquisition comprises four steps of rockfill point cloud acquisition, point cloud data preprocessing, segmentation model application and rockfill grading calculation. The beneficial effects of the invention are as follows: the method comprises the steps of generating a segmentation model and acquiring the grading of the rock-fill material, and directly and quickly completing the segmentation of the point cloud of the rock-fill material and the detection of the grading information by utilizing the constructed network model.
Description
Technical Field
The invention relates to the technical field of construction quality detection, in particular to a rockfill grading information detection method based on edge detection.
Background
In the hydraulic engineering construction process, the grading of the dam material rock-fill is one of important indexes for construction quality detection. In the current engineering practice, a time-consuming and labor-consuming screening method is generally adopted to detect the rock-fill grading. The screening method is to screen the rock-fill by using screens with different screen diameters, count the percentage of the residual stone mass on each screen to the total mass, and draw a corresponding grading curve. Although many automatic screening machines replace manual screening in the current engineering, the automatic screening machines are still difficult to apply to site detection of a stock ground, the process is complex, and the increasing demands of China on intelligent construction are not met.
In recent years, a plurality of students acquire vehicle-mounted earth and stone images by using a camera, generate binary images of photos by adopting a maximum inter-class difference method (Otus algorithm) and a local double-window threshold optimization method, and then divide image gradients by using a watershed method to acquire block stone boundaries so as to acquire rock-fill grading; or an industrial camera is used for acquiring an image of the flint on the conveyor belt, after image preprocessing processes such as filtering and denoising, a watershed particle segmentation algorithm based on multi-scale morphological improvement is applied to segment the flint from the background, and after research comparison, the accuracy of the watershed method segmentation boundary is found to reach 71%. However, since a large amount of information is often lost in the two-dimensional image, the detection accuracy of the method in the verification set is low, and the method is very sensitive to noise. Although the three-dimensional point cloud reconstruction technology based on the three-dimensional laser scanner has been widely applied to hydraulic engineering construction monitoring due to the advantages of high precision and no contact, the processing of the point cloud data is very difficult due to the disorder, neighborhood and affine invariance of the three-dimensional point cloud, so that a rockfill grading detection method based on the three-dimensional point cloud is not available at present.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a rockfill grading information detection method based on edge detection.
The aim of the invention is achieved by the following technical scheme: the rock-fill grading information detection method based on edge detection comprises four steps of segmentation model generation and rock-fill grading acquisition, wherein the segmentation model generation comprises training data set construction, segmentation model training and segmentation model debugging, and the rock-fill grading acquisition comprises four steps of rock-fill point cloud acquisition, point cloud data preprocessing, segmentation model application and rock-fill grading calculation.
Preferably, the training data set construction includes a rock-fill training point cloud set generation and a rock-fill training label set generation, the rock-fill training point cloud set generation includes the following steps:
s1.0: randomly generating a plurality of three-dimensional coordinate control points in a unit space, and generating corresponding convex hull polyhedral triangular plates by using the control points after random stretching;
s1.1: establishing a block stone sphere model by using a rigid sphere algorithm, and converting the convex hull polyhedral triangular plates into a point cloud format by using a PCL (PCL) calculation library to obtain point cloud data of a single block stone;
s1.2: forming accumulation of the random number of block stone sphere models in a gravity deposition mode in open source discrete element software yade, recording the coordinates of the center points of the balls before and after accumulation to calculate a conversion matrix, and moving the point cloud coordinates of the corresponding block stones according to the obtained conversion matrix to obtain input training point cloud data;
the rock-fill training label set is generated by a normal vector difference method.
Preferably, the segmentation model comprises a deep learning model and a block stone generation model, the deep learning model is used for detecting block stone boundary edges of a rock-fill point cloud, the deep learning model comprises a feature extraction layer and a feature propagation layer, and the feature extraction layer comprises a furthest point sampling layer, a ball query adjacent point combination layer and a feature calculation layer; the block stone generation model is used for identifying block stone boundary edge information, dividing single block stones from the rock fill, and comprises edge elimination, region growing and label mapping.
Preferably, the segmentation model training is used for optimizing training parameters and super parameters in the deep learning model, wherein the training parameters of the deep learning model are weights and offsets in a convolution layer, the weights are randomly generated during model initialization, and the offsets are set to 0 during model initialization; the super parameters of the deep learning model comprise a learning rate and a regularization coefficient, wherein the learning rate is used for controlling the magnitude of a parameter gradient update value in each iteration process, and the regularization coefficient is used for controlling a penalty value of the parameter gradient update in each iteration process.
Preferably, the segmentation model debugging is used for optimizing parameters in the block stone generation model, and the segmentation model debugging comprises edge elimination, the number of adjacent points mapped by the labels, and curvature threshold and distance threshold of region growth.
Preferably, the rock-fill point cloud acquisition needs to set a scanning network in advance.
Preferably, the point cloud data preprocessing includes denoising, downsampling, registration, and smoothing.
Preferably, the application of the segmentation model is to input the trained and debugged segmentation model, the preprocessed rock-fill point cloud coordinates as a model, obtain rock-fill boundary edge points after deep learning the model, and obtain the point cloud coordinates of each rock in the rock-fill after generating the model by the rock-fill.
Preferably, the rock-fill grading calculation further comprises the following steps:
s2.0: converting each rock point cloud in the rock-fill into a triangular plate form, calculating the volume V of the corresponding rock, and taking the diameter of a sphere with the same volume as the volume V as the particle size of the rock;
s2.1: and counting the percentage of the total volume of the block stones with the grain diameters smaller than the set values in the rock stacking to the total volume of the rock stacking, and drawing a corresponding grading curve.
The invention has the following advantages: according to the invention, the two parts of segmentation model generation and rock-fill grading acquisition are adopted, and the segmentation of the rock-fill material point cloud and the detection of grading information can be directly and rapidly completed by utilizing the constructed network model.
Drawings
FIG. 1 is a schematic diagram of the structure of the present invention;
FIG. 2 is a schematic structural diagram of a deep learning model framework;
FIG. 3 is a schematic diagram of a deep learning model training process;
FIG. 4 is a schematic diagram of a structure of a segmentation model debugging effect;
FIG. 5 is a schematic diagram of a structure established by a scanning network of a three-dimensional laser scanner;
fig. 6 is a schematic structural diagram of a processed stockyard rock-fill point cloud.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, based on the embodiments of the invention, which are apparent to those of ordinary skill in the art without inventive faculty, are intended to be within the scope of the invention.
In addition, the embodiments of the present invention and the features of the embodiments may be combined with each other without collision.
It should be noted that: like reference numerals and letters denote like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures.
In the description of the present invention, it should be noted that, directions or positional relationships indicated by terms such as "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., are directions or positional relationships based on those shown in the drawings, or are directions or positional relationships conventionally put in use of the inventive product, or are directions or positional relationships conventionally understood by those skilled in the art, are merely for convenience of describing the present invention and for simplifying the description, and are not to indicate or imply that the apparatus or element to be referred to must have a specific direction, be constructed and operated in a specific direction, and thus should not be construed as limiting the present invention. Furthermore, the terms "first," "second," and the like, are used merely to distinguish between descriptions and should not be construed as indicating or implying relative importance.
In the description of the present invention, it should also be noted that, unless explicitly specified and limited otherwise, the terms "disposed," "mounted," "connected," and "connected" are to be construed broadly, and may be, for example, fixedly connected, detachably connected, or integrally connected; can be mechanically or electrically connected; can be directly connected or indirectly connected through an intermediate medium, and can be communication between two elements. The specific meaning of the above terms in the present invention will be understood in specific cases by those of ordinary skill in the art.
In this embodiment, as shown in fig. 1, the method for detecting the rock-fill grading information based on edge detection includes four steps of generating a segmentation model and obtaining the rock-fill grading, wherein the generation of the segmentation model includes four steps of training data set construction, segmentation model training and segmentation model debugging, and the obtaining of the rock-fill grading includes four steps of rock-fill point cloud obtaining, point cloud data preprocessing, segmentation model application and rock-fill grading calculation. The method comprises the steps of generating a segmentation model and acquiring the grading of the rock-fill material, and directly and quickly completing the segmentation of the point cloud of the rock-fill material and the detection of the grading information by utilizing the constructed network model.
Further, the training data set construction comprises a rock-fill training point cloud set generation and a rock-fill training label set generation, wherein the rock-fill training point cloud set generation comprises the following steps:
s1.0: randomly generating a plurality of three-dimensional coordinate control points in a unit space, and generating corresponding convex hull polyhedral triangular plates by using the control points after random stretching; preferably, 100-500 three-dimensional coordinate control points are randomly generated in a unit space.
S1.1: establishing a block stone sphere model by using a rigid sphere algorithm, and converting the convex hull polyhedral triangular plates into a point cloud format by using a PCL (PCL) calculation library to obtain point cloud data of a single block stone;
s1.2: forming accumulation of the random number of block stone sphere models in a gravity deposition mode in open source discrete element software yade, recording the coordinates of the center points of the balls before and after accumulation to calculate a conversion matrix, and moving the point cloud coordinates of the corresponding block stones according to the obtained conversion matrix to obtain input training point cloud data;
the rock-fill training label set is generated by a normal vector difference method. Specifically, the concrete steps of generating the rock-fill training label set by a normal vector difference method are as follows: firstly, 8 adjacent points near a target point are selected through a K-nearest neighbor method, then 8 adjacent points are fitted into a plane through a PCA algorithm, the normal vector of the plane is used as the normal vector of the target point, and when the point near the target point changes, the corresponding normal vector changes. Therefore, when the rock blocks are put into the space in sequence, the normal vector of the newly generated point cloud is calculated, and the point where the normal vector changes in each step is selected as the boundary edge point of the rock blocks, so that the generation of the rock-fill training label set is completed. In this example 4000 sets of stacked stones were co-produced. The rigid sphere algorithm, the PCL calculation library, the open source discrete meta-software and the PCA algorithm mentioned therein are all prior art and are not improved, and will not be described herein.
Still further, as shown in fig. 2, the segmentation model includes a deep learning model and a block stone generation model, the deep learning model is used for detecting a block stone boundary edge of a rock-fill point cloud, the deep learning model includes a feature extraction layer and a feature propagation layer, and the feature extraction layer includes a furthest point sampling layer, a ball query adjacent point combination layer and a feature calculation layer; the block stone generation model is used for identifying block stone boundary edge information, dividing single block stones from the rock fill, and comprises edge elimination, region growing and label mapping. Specifically, the deep learning model uses a PointNet++ level neural network model architecture, and the most distant point sampling layer specifically comprises the following steps: randomly select a point
As seed point, select the point farthest from the point in space +.>
As a second seed point, the distance point set is then selected from the space +.>
Furthest point->
As a third seed point, the above procedure is cycled until a preset number of seed points are selected, and in this embodiment, 6 furthest sampling layers are used, and the number of selected seed points is respectively: 768. 512, 256, 128, 64, 32; the ball inquiry adjacent point combination layer specifically comprises: for a certain point in the space, the point is taken as a sphere center, a ball is made according to a set radius, a preset number of points contained in the ball are selected as adjacent points of the point and combined into a point set, in the embodiment, 6 ball query adjacent point combination layers are adopted in total, and the number of selected points is as follows: 256. 128, 64, 32, 16, 8, with radii set to 0.1, 0.2, 0.3, 0.4, 0.5; the specific steps of the feature calculation layer are as follows: firstly, extracting the characteristics of an input point set through convolution operation, and reusing a maximum poolThe maximum value in each dimension of the feature is reserved in a chemical mode, and then the feature of the output point set is extracted through convolution operation again. The feature propagation layer firstly selects points in 3 high-level point clouds from each point in the low-level point clouds, and features of each point in the bottom level are obtained through feature shearing and convolution, so that feature propagation is realized. The edge elimination is specifically to select 32 adjacent points of each block stone boundary edge according to a K adjacent method, and delete all block stone boundary edge points and adjacent points from a rock-fill point cloud; the specific steps of the region growth are as follows:
s3.0: firstly, calculating the curvature and normal vector of each point in the point cloud by using a PCA algorithm, and arranging each point as a global seed set according to the order of the curvature from small to large;
s3.1: selecting a first point in the global seed set as a seed point;
s3.2: initializing a region subset and a region seed subset to be empty;
s3.3: 8 adjacent points near the seed point are selected by using a K adjacent method and are used as points to be fixed for judgment in sequence;
s3.4: if the included angle and the distance between the to-be-fixed point and the normal vector of the seed point are smaller than the set threshold value, adding the to-be-fixed point into the region subset; if the curvature of the to-be-positioned point is smaller than the set curvature threshold value, adding the to-be-positioned point into the regional seed set, and deleting the to-be-positioned point in the global seed set;
s3.5: when all the points to be determined are judged, if the regional seed subsets are not empty, selecting a first point in the regional seed sets as a seed point, and repeating the steps S3.3-S3.5; if the regional seed subset is empty, the current regional subset is saved, and then the steps S3.1-S3.5 are repeated until the global seed subset is empty;
s3.6: after the steps, the rock-fill point cloud after adjacent point removal is divided into a plurality of regional subsets, each regional subset represents an independent block stone, and only regional subsets with the points more than 30 are used for eliminating errors. Points in the subsets of the areas are given the same labels, and block stone segmentation is completed. The label mapping is specifically to obtain 8 adjacent points nearby a point without a label by using a K adjacent algorithm, taking a label value with the largest label occurrence frequency in the adjacent points as a label of the point, and adding the label value into a corresponding region subset. In this embodiment, the K-nearest neighbor algorithm is a prior art, and is not improved, and will not be described here again.
In this embodiment, the segmentation model training is used to optimize training parameters and superparameters in the deep learning model, that is, to achieve better recognition of the boundary edges of the stone blocks. The training parameters of the deep learning model are weight and bias in a convolution layer, the weight is randomly generated when the model is initialized, and the bias is set to 0 when the model is initialized; the super parameters of the deep learning model comprise a learning rate and a regularization coefficient, wherein the learning rate is used for controlling the magnitude of a parameter gradient update value in each iteration process, and the regularization coefficient is used for controlling a penalty value of the parameter gradient update in each iteration process, so that the possibility of overfitting the model is reduced. Specifically, the preferred procedure for training parameters is: the training rock-fill point cloud set is used as a deep learning model input, the training label set is used as a deep learning model training reference value, a cross entropy loss function is used for calculating a loss value between an output value of the deep learning model and the training reference value, a back propagation algorithm is used for calculating gradient update values of all training parameters according to the loss value, one-time updating of the training parameters is completed, the process is repeated until the gradient update values converge or the loss value reaches a preset threshold value, and training parameter optimization is completed; the preferred procedure for the super parameters is: setting different super-parameter combinations to finish the optimization of training parameters corresponding to each combination, selecting the super-parameter combination with the minimum loss value after the optimization is finished as the super-parameter of the deep learning model, and taking the corresponding training parameter as the training parameter of the deep learning model. Through a super-parameter optimization experiment, the optimal learning rate is 0.0001, the optimal regularization coefficient is 0.04, and the training process is shown in fig. 3. In this embodiment, the cross entropy loss function and the back propagation algorithm are prior art, and are not improved, and will not be described in detail here.
Further, the segmentation model debugging is used for optimizing parameters in the block stone generation model, and the segmentation model debugging comprises edge elimination, the number of adjacent points mapped by the labels, a curvature threshold value for region growth and a distance threshold value. Specifically, the segmentation model debugging is to set different parameter combinations, select the combination with the best segmentation effect in the parameter combinations as the parameter of the segmentation model, and preferably obtain the best effect when the number of selected edge elimination neighboring points is 32, the number of label mapping points is 8, the region growing curvature threshold is 0.02, and the distance threshold is 0.2, as shown in fig. 4.
In this embodiment, as shown in fig. 5, the scanning net needs to be set in advance for the rock-fill point cloud acquisition. Specifically, the main function of setting the scanning network in advance is to acquire multi-view point cloud data, and keep the complete three-dimensional information of the rockfill material, and the specific setting position is referred to fig. 5.
Further, the point cloud data preprocessing includes denoising, downsampling, registration, and smoothing. Specifically, denoising, downsampling, and registration can be quickly accomplished in Riscan Pro software; and removing textures of the surface of the block stone point cloud by using a mobile least square algorithm in the PCL algorithm library smoothly, and increasing the effect of block stone boundary edge detection. In this embodiment, both Riscan Pro software and mobile least squares are prior art, and are not improved, and will not be described in detail here.
Still further, as shown in fig. 6, the application of the segmentation model is to input the trained and debugged segmentation model, the preprocessed rock-fill point cloud coordinates as a model, obtain boundary edge points of the rock blocks after deep learning the model, and obtain the point cloud coordinates of each rock block in the rock-fill after generating the model by the rock blocks.
In this embodiment, the rock-fill grading calculation further includes the steps of:
s2.0: converting each rock point cloud in the rock-fill into a triangular plate form, calculating the volume V of the corresponding rock, and taking the diameter of a sphere with the same volume as the volume V as the particle size of the rock;
s2.1: and counting the percentage of the total volume of the block stones with the grain diameters smaller than the set values in the rock stacking to the total volume of the rock stacking, and drawing a corresponding grading curve.
Although the present invention has been described with reference to the foregoing embodiments, it will be apparent to those skilled in the art that modifications may be made to the embodiments described, or equivalents may be substituted for elements thereof, and any modifications, equivalents, improvements and changes may be made without departing from the spirit and principles of the present invention.
Claims (7)
1. A rock-fill grading information detection method based on edge detection is characterized in that: the method comprises the steps of segmentation model generation and rock-fill grading acquisition, wherein the segmentation model generation comprises four steps of training data set construction, segmentation model training and segmentation model debugging, and the rock-fill grading acquisition comprises four steps of rock-fill point cloud acquisition, point cloud data preprocessing, segmentation model application and rock-fill grading calculation;
the training data set construction comprises a rock-fill training point cloud set generation and a rock-fill training label set generation, wherein the rock-fill training point cloud set generation comprises the following steps:
s1.0: randomly generating a plurality of three-dimensional coordinate control points in a unit space, and generating corresponding convex hull polyhedral triangular plates by using the control points after random stretching;
s1.1: establishing a block stone sphere model by using a rigid sphere algorithm, and converting the convex hull polyhedral triangular plates into a point cloud format by using a PCL (PCL) calculation library to obtain point cloud data of a single block stone;
s1.2: forming accumulation of the random number of block stone sphere models in a gravity deposition mode in open source discrete element software yade, recording the coordinates of the center points of the balls before and after accumulation to calculate a conversion matrix, and moving the point cloud coordinates of the corresponding block stones according to the obtained conversion matrix to obtain input training point cloud data;
the method comprises the steps that a rock-fill training label set is generated through a normal vector difference method, 8 adjacent points of a target point are selected through a K adjacent method, the 8 adjacent points are fitted into a plane through a PCA algorithm, the normal vector of the plane is used as the normal vector of the target point, when the adjacent points are changed, the corresponding normal vector is also changed, when a rock block is put into a space in sequence, the normal vector of a newly generated point cloud is calculated, the point with the changed normal vector in each step is selected as a boundary edge point of the rock block, and the generation of the rock-fill training label set is completed;
the segmentation model comprises a deep learning model and a block stone generation model, wherein the deep learning model is used for detecting block stone boundary edges of a rock-fill point cloud, the deep learning model comprises a feature extraction layer and a feature propagation layer, and the feature extraction layer comprises a furthest point sampling layer, a ball query adjacent point combination layer and a feature calculation layer; the block stone generation model is used for identifying boundary edge information of the block stone, dividing single block stone from the rock pile, and comprises edge elimination, region growth and label mapping;
the specific steps of the region growth are as follows:
s3.0: calculating the curvature and normal vector of each point in the point cloud by using a PCA algorithm, and arranging each point as a global seed subset according to the order of the curvature from small to large;
s3.1: selecting a first point in the global seed set as a seed point;
s3.2: initializing a region subset and a region seed subset to be empty;
s3.3: 8 adjacent points of the seed points are selected by using a K adjacent method and are used as points to be fixed for judgment in sequence;
s3.4: when the included angle and the distance between the to-be-fixed point and the normal vector of the seed point are smaller than the set threshold value, adding the to-be-fixed point into the region subset; when the curvature of the to-be-positioned point is smaller than the set curvature threshold value, adding the to-be-positioned point into the regional seed set, and deleting the to-be-positioned point in the global seed set;
s3.5: when all the points to be determined are judged, if the regional seed subsets are not empty, selecting a first point in the regional seed sets as a seed point, and repeating the steps S3.3-S3.5; if the regional seed subset is empty, the current regional subset is saved, and then the steps S3.1-S3.5 are repeated until the global seed subset is empty;
s3.6: after the steps, the rock-fill point cloud after the adjacent points are removed is divided into a plurality of regional subsets, and each regional subset represents an independent rock block.
2. The method for detecting rock-fill gradation information based on edge detection according to claim 1, wherein: the segmentation model training is used for optimizing training parameters and super parameters in a deep learning model, the training parameters of the deep learning model are weights and offsets in a convolution layer, the weights are randomly generated during model initialization, and the offsets are set to be 0 during model initialization; the super parameters of the deep learning model comprise a learning rate and a regularization coefficient, wherein the learning rate is used for controlling the magnitude of a parameter gradient update value in each iteration process, and the regularization coefficient is used for controlling a penalty value of the parameter gradient update in each iteration process.
3. The method for detecting rock-fill gradation information based on edge detection according to claim 2, wherein: the segmentation model debugging is used for optimizing parameters in the block stone generation model, and comprises edge elimination, the number of adjacent points mapped by the labels, a curvature threshold value for region growth and a distance threshold value.
4. The method for detecting rock-fill gradation information based on edge detection according to claim 1, wherein: and the rock-fill point cloud acquisition needs to set a scanning network in advance.
5. The method for detecting rock-fill gradation information based on edge detection according to claim 4, wherein: the point cloud data preprocessing includes denoising, downsampling, registration and smoothing.
6. The method for detecting rock-fill gradation information based on edge detection according to claim 5, wherein: the application of the segmentation model is to input the trained and debugged segmentation model, the preprocessed rock-fill point cloud coordinates as a model, obtain rock-fill boundary edge points after deep learning of the model, and obtain the point cloud coordinates of each rock in the rock-fill after rock generation of the model.
7. The method for detecting rock-fill gradation information based on edge detection according to claim 6, wherein: the rock-fill grading calculation further comprises the following steps:
s2.0: converting each rock point cloud in the rock-fill into a triangular plate form, calculating the volume V of the corresponding rock, and taking the diameter of a sphere with the same volume as the volume V as the particle size of the rock;
s2.1: and counting the percentage of the total volume of the block stones with the grain diameters smaller than the set values in the rock stacking to the total volume of the rock stacking, and drawing a corresponding grading curve.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310109315.9A CN115908425B (en) | 2023-02-14 | 2023-02-14 | Edge detection-based rock-fill grading information detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310109315.9A CN115908425B (en) | 2023-02-14 | 2023-02-14 | Edge detection-based rock-fill grading information detection method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115908425A CN115908425A (en) | 2023-04-04 |
| CN115908425B true CN115908425B (en) | 2023-06-30 |
Family
ID=86488406
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310109315.9A Active CN115908425B (en) | 2023-02-14 | 2023-02-14 | Edge detection-based rock-fill grading information detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115908425B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107066654A (en) * | 2016-12-15 | 2017-08-18 | 广东工业大学 | Towards the TRAJECTORY CONTROL dot picking method and apparatus in the enveloped box face of point cloud model |
| CA2962334A1 (en) * | 2016-03-29 | 2017-09-29 | Robert Lindsay Vanderbeck | Tunnel convergence detection apparatus and method |
| CN110232736A (en) * | 2019-06-18 | 2019-09-13 | 中国矿业大学 | A kind of down-hole combined mining working three-dimensional scenic fast construction method |
| CN112330699A (en) * | 2020-11-14 | 2021-02-05 | 重庆邮电大学 | Three-dimensional point cloud segmentation method based on overlapping region alignment |
| CN114022474A (en) * | 2021-11-23 | 2022-02-08 | 浙江宁海抽水蓄能有限公司 | Particle grading rapid detection method based on YOLO-V4 |
| CN115439762A (en) * | 2022-07-29 | 2022-12-06 | 中国科学院武汉岩土力学研究所 | High and steep slope dangerous rock mass rapid identification method based on unmanned aerial vehicle LiDAR (light detection and ranging) simulated land flight |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106228536B (en) * | 2016-07-12 | 2018-12-18 | 四川大学 | The earth and rockfill dam dam material grain composition method of inspection based on Digital Image Processing |
| CN109102109A (en) * | 2018-07-16 | 2018-12-28 | 四川大学 | A kind of piece of exploitation of stone explosion gradation prediction technique |
| CN112287939A (en) * | 2020-10-29 | 2021-01-29 | 平安科技（深圳）有限公司 | Three-dimensional point cloud semantic segmentation method, device, equipment and medium |
| CN114758043A (en) * | 2020-12-29 | 2022-07-15 | 中冶长天国际工程有限责任公司 | Intelligent ore block size identification method and system based on 3D cloud picture |
| CN113012161B (en) * | 2021-03-05 | 2023-04-28 | 华南理工大学 | Stacked scattered target point cloud segmentation method based on convex region growth |
| CN115439655A (en) * | 2022-03-29 | 2022-12-06 | 葫芦岛农函大玄宇食用菌野驯繁育有限公司 | Pleurotus ostreatus growth monitoring method based on point cloud deep learning |
| CN115115839A (en) * | 2022-07-28 | 2022-09-27 | 南京工业大学 | Building indoor point cloud segmentation method based on local feature enhanced PointNet + + network |
-
2023
- 2023-02-14 CN CN202310109315.9A patent/CN115908425B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2962334A1 (en) * | 2016-03-29 | 2017-09-29 | Robert Lindsay Vanderbeck | Tunnel convergence detection apparatus and method |
| CN107066654A (en) * | 2016-12-15 | 2017-08-18 | 广东工业大学 | Towards the TRAJECTORY CONTROL dot picking method and apparatus in the enveloped box face of point cloud model |
| CN110232736A (en) * | 2019-06-18 | 2019-09-13 | 中国矿业大学 | A kind of down-hole combined mining working three-dimensional scenic fast construction method |
| CN112330699A (en) * | 2020-11-14 | 2021-02-05 | 重庆邮电大学 | Three-dimensional point cloud segmentation method based on overlapping region alignment |
| CN114022474A (en) * | 2021-11-23 | 2022-02-08 | 浙江宁海抽水蓄能有限公司 | Particle grading rapid detection method based on YOLO-V4 |
| CN115439762A (en) * | 2022-07-29 | 2022-12-06 | 中国科学院武汉岩土力学研究所 | High and steep slope dangerous rock mass rapid identification method based on unmanned aerial vehicle LiDAR (light detection and ranging) simulated land flight |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115908425A (en) | 2023-04-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109409263B (en) | Method for detecting urban ground feature change of remote sensing image based on Siamese convolutional network | |
| CN109614985B (en) | Target detection method based on densely connected feature pyramid network | |
| CN109029381B (en) | Tunnel crack detection method and system and terminal equipment | |
| CN109919206B (en) | Remote sensing image earth surface coverage classification method based on full-cavity convolutional neural network | |
| CN111914795B (en) | Method for detecting rotating target in aerial image | |
| CN111527467A (en) | Method and apparatus for automatically defining computer-aided design files using machine learning, image analysis, and/or computer vision | |
| EP3614308A1 (en) | Joint deep learning for land cover and land use classification | |
| CN111738165B (en) | Method for extracting individual plant canopy from high-resolution unmanned aerial vehicle visible light remote sensing image | |
| CN107067405B (en) | Remote sensing image segmentation method based on scale optimization | |
| CN111339947A (en) | Method and system for extracting remote sensing image fuzzy boundary ground object, storage medium and equipment | |
| Chen et al. | Photogrammetric point cloud segmentation and object information extraction for creating virtual environments and simulations | |
| CN110991444A (en) | Complex scene-oriented license plate recognition method and device | |
| Jiang et al. | Determining ground elevations covered by vegetation on construction sites using drone-based orthoimage and convolutional neural network | |
| Xia et al. | Deep extraction of cropland parcels from very high-resolution remotely sensed imagery | |
| CN113963262A (en) | Mining area land coverage classification method based on depth feature fusion model | |
| Salah et al. | Evaluation of the self‐organizing map classifier for building detection from lidar data and multispectral aerial images | |
| CN111611960B (en) | Large-area ground surface coverage classification method based on multilayer perceptive neural network | |
| CN107271367B (en) | A kind of identifying water boy method and device | |
| CN115147401A (en) | Digital image processing-based intelligent detection method for dam material qualification of earth and rockfill dam | |
| EP4174792A1 (en) | Method for scene understanding and semantic analysis of objects | |
| Wahbi et al. | A deep learning classification approach using high spatial satellite images for detection of built-up areas in rural zones: Case study of Souss-Massa region-Morocco | |
| CN105069459A (en) | Surface feature type extracting method for high-resolution SAR image | |
| CN114742864A (en) | Belt deviation detection method and device | |
| CN115908425B (en) | Edge detection-based rock-fill grading information detection method | |
| CN116091706B (en) | Three-dimensional reconstruction method for multi-mode remote sensing image deep learning matching |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |