CN113740216B - Air-ground integrated detection method for wide-gradation mixed aggregate - Google Patents
Air-ground integrated detection method for wide-gradation mixed aggregate Download PDFInfo
- Publication number
- CN113740216B CN113740216B CN202111058008.XA CN202111058008A CN113740216B CN 113740216 B CN113740216 B CN 113740216B CN 202111058008 A CN202111058008 A CN 202111058008A CN 113740216 B CN113740216 B CN 113740216B
- Authority
- CN
- China
- Prior art keywords
- aggregate
- wide
- detection
- conveyor belt
- curve
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 53
- 238000005516 engineering process Methods 0.000 claims abstract description 35
- 238000000034 method Methods 0.000 claims abstract description 29
- 230000011218 segmentation Effects 0.000 claims abstract description 29
- 239000002245 particle Substances 0.000 claims description 36
- 238000012216 screening Methods 0.000 claims description 22
- 238000010276 construction Methods 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 11
- 239000000463 material Substances 0.000 claims description 7
- 238000005070 sampling Methods 0.000 claims description 7
- 239000002344 surface layer Substances 0.000 claims description 7
- 238000013145 classification model Methods 0.000 claims description 6
- 230000010354 integration Effects 0.000 claims description 6
- 238000010801 machine learning Methods 0.000 claims description 6
- 239000002184 metal Substances 0.000 claims description 3
- 238000005259 measurement Methods 0.000 abstract description 3
- 238000013135 deep learning Methods 0.000 abstract description 2
- 238000013527 convolutional neural network Methods 0.000 description 13
- 230000006870 function Effects 0.000 description 7
- 230000007547 defect Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 239000004576 sand Substances 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 239000000203 mixture Substances 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 239000002689 soil Substances 0.000 description 3
- 239000004575 stone Substances 0.000 description 3
- 206010063385 Intellectualisation Diseases 0.000 description 2
- 239000010426 asphalt Substances 0.000 description 2
- 239000004567 concrete Substances 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 239000010410 layer Substances 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000004566 building material Substances 0.000 description 1
- 239000011362 coarse particle Substances 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000005056 compaction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000000945 filler Substances 0.000 description 1
- 239000008187 granular material Substances 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000017105 transposition Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N15/00—Investigating characteristics of particles; Investigating permeability, pore-volume or surface-area of porous materials
- G01N15/02—Investigating particle size or size distribution
- G01N15/0205—Investigating particle size or size distribution by optical means
- G01N15/0227—Investigating particle size or size distribution by optical means using imaging; using holography
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/86—Combinations of lidar systems with systems other than lidar, radar or sonar, e.g. with direction finders
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/88—Lidar systems specially adapted for specific applications
- G01S17/89—Lidar systems specially adapted for specific applications for mapping or imaging
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- General Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Electromagnetism (AREA)
- Chemical & Material Sciences (AREA)
- Dispersion Chemistry (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a wide-gradation mixed Aggregate space-ground integrated detection method, which is based on a deep learning method and a laser radar measurement technology under a big data background and provides an Aggregate rapid intelligent detection segmentation algorithm Aggregate Net model based on an example segmentation framework.
Description
Technical Field
The invention relates to the technical field of aggregate detection, in particular to a method for detecting air-ground integrated gradation of wide-gradation mixed aggregate.
Background
The construction engineering and the hydraulic engineering are used as important infrastructures for national economy construction and social development in China, and safety of the construction engineering and the hydraulic engineering is improved insignificantly. The wide-graded mixed aggregate is a main building material for water conservancy, highway, railway and building engineering. The reasonable aggregate mixing proportion is an important factor influencing the durability of an engineering structure and guaranteeing the operation safety of the engineering. In order to ensure the compaction quality of filling engineering and improve the deformation resistance and the impermeability, the filler with reasonable gradation is particularly important. Wherein, the volume ratio of the aggregate in the concrete and the cemented sand gravel particle material is more than 50 percent to 70 percent, and the aggregate gradation is an important factor for measuring the mechanical property of the wide-gradation mixed material. How to rapidly and accurately detect and judge the rationality of the gradation of the filled soil stones is always a focus of attention in the engineering field. At present, the aggregate grading detection method mainly depends on a traditional manual or mechanical screening method, and a grading curve of the wide-graded mixed aggregate is obtained through statistical calculation. Although the traditional screening method is mature in technology, the number of samples calculated each time is limited, manpower is consumed, efficiency is low, construction progress is influenced, and the development requirements of digitalization, automation and intellectualization of the current construction technology cannot be met. Therefore, it is necessary to develop a method for rapidly detecting the particle size of the aggregate in a non-contact intelligent manner, so as to achieve rapid and dynamic acquisition of the wide-gradation mixed aggregate gradation.
At present, domestic and foreign researches mainly focus on the aspects of ore granularity, asphalt mixture uniformity, particle size distribution, aggregate particle characteristics and the like, and scholars detect the aggregate particle size based on digital image processing technology and analysis theory. Doctor dayakar penumadu uses an automatic test machine to quickly determine the particle size distribution of unbound aggregate. Lee J.R.J., U.S. utilizes laser triangulation to develop a system for collecting three-dimensional data of analytical particles from the surface of coarse particles. Sulaiman M S et al analyzed the particle size distribution characteristics using automatic image processing techniques with riverbed sand as the sample. Maerz designs a system for single-source visual multi-view collection and evaluation of particle morphology. The Hyoungkwan Kim carries out experiments and evaluation on various specific parameters based on a laser aggregate analysis system, and verifies the reliability of the system. Zhou Jian Hua et al studied the aggregate collection device to obtain various parameter indices when the aggregates were dispersed. Chua changes poor et al and utilizes CCD camera subassembly to gather the granule image, monitors the ore granularity based on image processing method. The Sundong slope is based on an image recognition technology, and the sand transportation rate of the migrated silt is determined. Pengyong researches a method for quantitatively describing the uniformity of asphalt mixtures by using a digital image processing technology and provides index parameters for describing the uniformity of the mixtures. The image recognition technology is emerging and tried in some industries, the soil and stone materials have geometric feature similarity with ores, silt and concrete aggregates, and the detection of the particle size grading of the soil and stone materials by using the image recognition technology has technical feasibility. However, the problem of grading identification and detection of mixed stacked aggregates cannot be solved by a simple image identification technology, and therefore, an efficient and convenient method applied to grading detection of the particle size of a large amount of wide-graded mixed aggregates needs to be developed urgently.
Disclosure of Invention
Aiming at the defects in the prior art, the method for detecting the air-ground integrated gradation of the wide-gradation mixed aggregate solves the problems that the traditional screening method has the defects of limited number of samples calculated each time, labor consumption and low efficiency and influences the construction progress.
In order to achieve the purpose of the invention, the invention adopts the technical scheme that: a method for detecting the air-ground integrated gradation of wide-gradation mixed aggregates comprises the following steps:
s1, selecting a parent metal excavated from a river bed or a construction site as a detection aggregate, stacking all the detection aggregates in a primary storage pile, airing for a period of time, carrying out grading detection on the surface layer stacked aggregates by adopting an unmanned aerial vehicle LiDAR and unmanned aerial vehicle photogrammetry technology, and establishing a first grading curve;
s2, randomly sampling the wide-graded mixed aggregate on the river bed or the primary storage pile excavated on the construction site in a large range on site by adopting a manual screening method, and establishing a second grading curve;
s3, performing curve fitting on the first grading curve and the second grading curve to establish a third grading curve;
s4, grabbing a certain amount of wide-gradation mixed aggregate from the primary storage pile to serve as aggregate to be detected, collecting images of the aggregate to be detected on the conveyor belt through a combined device of a photogrammetry technology and a laser radar technology, and establishing a fourth gradation curve;
s5, randomly sampling the wide-gradation mixed aggregate in the step S4 to obtain a manual screening detection sample, and performing field manual screening on the manual screening detection sample to establish a fifth gradation curve;
s6, performing curve fitting on the fourth grading curve and the fifth grading curve to establish a sixth grading curve;
and S7, fitting the third grading curve and the sixth grading curve by adopting a least square method to establish a seventh grading curve, namely the final grading curve based on the wide-grading mixed aggregate 'air-ground integration' detection method.
Further, step S1 includes the following substeps:
s11, selecting a parent material excavated from a river bed or a construction site as a detection aggregate, stacking all the detection aggregates in a primary storage pile for airing for a period of time, and acquiring surface aggregate images of the surface stacked aggregates by adopting an unmanned aerial vehicle LiDAR and an unmanned aerial vehicle photogrammetry technology;
s12, performing example segmentation and recognition detection on the surface Aggregate image by adopting a pre-trained Aggregate Net model based on machine learning to obtain a predicted frame and a mask of each surface Aggregate;
s13, assuming the shape of aggregate particles as an ellipsoid, and calculating the volume of the ellipsoid of the wide-graded mixed aggregate according to the predicted frame and the mask of each surface aggregate;
s14, converting the screening mass ratio of the wide-graded mixed aggregate into a volume ratio according to the same density of the same aggregate particles, obtaining the mass of each particle group of the wide-graded mixed aggregate in the total mass percentage, and constructing a first grading curve of the wide-graded mixed aggregate.
Further, step S4 includes the following substeps:
s41, grabbing a certain amount of wide-graded mixed aggregate from the primary storage pile to serve as aggregate to be detected, and placing the aggregate to be detected on a conveyor belt;
s42, erecting a combined device of photogrammetry technology and laser radar technology on the conveyor belt, and collecting the aggregate image to be detected on the conveyor belt for the aggregate to be detected on the conveyor belt;
s43, carrying out example segmentation and identification detection on the Aggregate image to be detected on the conveyor belt by using a pre-trained Aggregate Net model based on machine learning to obtain a prediction frame and a mask of each surface Aggregate on the conveyor belt;
s44, assuming the shape of aggregate particles as an ellipsoid, calculating the area of an ellipsoid of the wide-graded mixed aggregate on the conveyor belt according to the predicted frame and the mask of each surface aggregate on the conveyor belt, and obtaining the thickness of the aggregate by a laser radar technology to obtain the volume of the ellipsoid approximate to the volume of the aggregate;
s45, converting the screening mass ratio of the wide-graded mixed aggregate into a volume ratio according to the same density of the same aggregate particles, obtaining the mass of each particle group of the wide-graded mixed aggregate on the conveyor belt accounting for the total mass percentage, and constructing a fourth grading curve of the wide-graded mixed aggregate.
Further, the calculation formula of the approximate ellipsoid volume in step S44 is:
wherein, a is the predicted major diameter, b is the predicted minor diameter size, and h is the height of each aggregate on the surface layer obtained by the laser radar.
Further, the Aggregate Net model comprises: the method comprises the following steps of (1) providing a region suggestion network, a scene classification model and a semantic segmentation model;
the area suggestion network generates candidate areas, and each anchor point generates 9 candidate areas with 3 sizes and 3 length-width ratios;
the scene classification model is used for classifying and regressing the candidate regions, namely classifying the candidate regions into pebble and non-pebble, and the regression is to adjust the positions of the frame marks to be closer to the actual positions;
the semantic segmentation model is responsible for segmenting pebble boundaries in each candidate frame, a scale A obtained by unmanned aerial vehicle photogrammetry and a scale B obtained by an industrial camera erected above a conveying ground are respectively obtained through the flight height of the unmanned aerial vehicle and the height of two camera sensors erected above a conveying belt, the predicted major diameter a and the predicted minor diameter B of each aggregate are calculated, and then the predicted particle diameter d of each aggregate is calculated.
Further, the calculation formula of the predicted particle diameter d of each aggregate is as follows:
in conclusion, the beneficial effects of the invention are as follows:
(1) the invention provides a wide-gradation mixed Aggregate space-ground integrated gradation detection method, based on a deep learning method and a laser radar measurement technology under a big data background, an Aggregate rapid intelligent detection and segmentation algorithm Aggregate Net model based on an example segmentation frame is provided, the model has strong parallel target detection and example segmentation capabilities, the predicted long diameter, the predicted short diameter and the Aggregate particle size of each Aggregate can be obtained, the thickness of the Aggregate is obtained by the laser radar measurement technology, so that the equivalent volume of each Aggregate is accurately calculated, and finally, data fitting is carried out on the Aggregate and an artificial sampling and screening result, so that a gradation curve result with higher precision is obtained.
(2) The aggregate proportioning detection and calculation method based on the artificial intelligence technology realizes the aggregate proportioning analysis, removes the manual work, improves the working efficiency, realizes the dynamic non-contact detection, can make up the defects of the traditional aggregate grading detection method in the hydraulic engineering and other buildings in China, and makes up the defects that the existing damming aggregate grading detection efficiency is low, a professional detection instrument cannot be used on site in a large range, and short plates such as the morphology characteristics of stacked aggregates cannot be dynamically analyzed. The research of the invention has important significance for the promotion and innovation of aggregate grading detection theory and technology, and further improves the technology level of digitization, automation and intellectualization in the traditional industry.
Drawings
FIG. 1 is a flow chart of a method for detecting the empty and ground integration of wide-graded mixed aggregates;
FIG. 2 is a flowchart of the Aggregate Net model algorithm.
Detailed Description
The following description of the embodiments of the present invention is provided to facilitate the understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and it will be apparent to those skilled in the art that various changes may be made without departing from the spirit and scope of the invention as defined and defined in the appended claims, and all matters produced by the invention using the inventive concept are protected.
As shown in fig. 1, a method for detecting the empty and ground integration of wide-gradation mixed aggregate comprises the following steps:
s1, selecting a parent metal excavated from a river bed or a construction site as a detection aggregate, stacking all the detection aggregates in a primary storage pile, airing for a period of time, carrying out grading detection on the surface layer stacked aggregates by adopting an unmanned aerial vehicle LiDAR and unmanned aerial vehicle photogrammetry technology, and establishing a first grading curve;
step S1 includes the following substeps:
s11, selecting a parent material excavated from a river bed or a construction site as a detection aggregate, stacking all the detection aggregates in a primary storage pile for airing for a period of time, and acquiring surface aggregate images of the surface stacked aggregates by adopting an unmanned aerial vehicle LiDAR and an unmanned aerial vehicle photogrammetry technology;
s12, performing example segmentation and recognition detection on the surface Aggregate image by adopting a pre-trained Aggregate Net model based on machine learning to obtain a predicted frame and a mask of each surface Aggregate;
s13, assuming the shape of aggregate particles as an ellipsoid, and calculating the volume of the ellipsoid of the wide-graded mixed aggregate according to the predicted frame and the mask of each surface aggregate;
the calculation formula of the ellipsoid volume of the wide-graded mixed aggregate is as follows:
wherein, a is the predicted major diameter, b is the predicted minor diameter size, and h is the height of each aggregate on the surface layer obtained by the laser radar.
S14, converting the screening mass ratio of the wide-graded mixed aggregate into a volume ratio according to the same density of the same aggregate particles, obtaining the mass of each particle group of the wide-graded mixed aggregate in the total mass percentage, and constructing a first grading curve of the wide-graded mixed aggregate.
S2, randomly sampling the wide-graded mixed aggregate on the river bed or the primary storage pile excavated on the construction site in a large range on site by adopting a manual screening method, and establishing a second grading curve;
s3, performing curve fitting on the first grading curve and the second grading curve to establish a third grading curve;
s4, grabbing a certain amount of wide-gradation mixed aggregate from the primary storage pile to serve as aggregate to be detected, collecting images of the aggregate to be detected on the conveyor belt through a combined device of a photogrammetry technology and a laser radar technology, and establishing a fourth gradation curve;
step S4 includes the following substeps:
s41, grabbing a certain amount of wide-graded mixed aggregate from the primary storage pile to serve as aggregate to be detected, and placing the aggregate to be detected on a conveyor belt;
s42, erecting a combined device of photogrammetry technology and laser radar technology on the conveyor belt, and collecting the aggregate image to be detected on the conveyor belt for the aggregate to be detected on the conveyor belt;
s43, performing instance segmentation and identification detection on the Aggregate image to be detected on the conveyor belt by using a pre-trained Aggregate Net model based on machine learning to obtain a predicted frame and a mask of each surface Aggregate on the conveyor belt;
s44, assuming the shape of aggregate particles as an ellipsoid, calculating the area of an ellipsoid of the wide-graded mixed aggregate on the conveyor belt according to the predicted frame and the mask of each surface aggregate on the conveyor belt, and obtaining the thickness of the aggregate by a laser radar technology to obtain the volume of the ellipsoid approximate to the volume of the aggregate;
the calculation formula of the approximate ellipsoid volume in step S44 is:
wherein, a is the predicted major diameter, b is the predicted minor diameter size, and h is the height of each aggregate on the surface layer obtained by the laser radar.
S45, converting the screening mass ratio of the wide-graded mixed aggregate into a volume ratio according to the same density of the same aggregate particles, obtaining the mass of each particle group of the wide-graded mixed aggregate on the conveyor belt accounting for the total mass percentage, and constructing a fourth grading curve of the wide-graded mixed aggregate.
S5, randomly sampling the wide-gradation mixed aggregate in the step S4 to obtain a manual screening detection sample, and performing field manual screening on the manual screening detection sample to establish a fifth gradation curve;
s6, performing curve fitting on the fourth grading curve and the fifth grading curve to establish a sixth grading curve;
s7, fitting the third grading curve and the sixth grading curve by adopting a least square method curve, and establishing a seventh grading curve, namely the final grading curve based on the wide-graded mixed aggregate 'air-ground integrated' grading detection method.
The Aggregate Net model comprises: the method comprises the following steps of (1) providing a region suggestion network, a scene classification model and a semantic segmentation model;
the area suggestion network generates candidate areas, and each anchor point generates 9 candidate areas with 3 sizes and 3 length-width ratios;
the scene classification model is used for classifying and regressing the candidate regions, namely classifying the candidate regions into pebble and non-pebble, and the regression is to adjust the positions of the frame marks to be closer to the actual positions;
the semantic segmentation model is responsible for segmenting pebble boundaries in each candidate frame, a scale A obtained by unmanned aerial vehicle photogrammetry and a scale B obtained by an industrial camera erected above a conveying ground are respectively obtained through the flight height of the unmanned aerial vehicle and the height of two camera sensors erected above a conveying belt, the predicted major diameter a and the predicted minor diameter B of each aggregate are calculated, and then the predicted particle diameter d of each aggregate is calculated.
The calculation formula of the predicted particle size d of each aggregate is as follows:
the invention establishes an Aggregate Net network model on the basis of a Mask R-CNN network structure, wherein the Mask R-CNN is an example segmentation algorithm and can be used for target detection, target example segmentation, target key point detection and the like, a prediction frame is generated by a target detection model, and a semantic segmentation model is segmented to obtain a specific example boundary. The Mask R-CNN network structure is based on a fast-RCNN framework, a fully-connected segmentation sub-network is added behind a basic feature network, and the original two tasks of classification and regression are changed into three tasks of classification, regression and segmentation. The Mask R-CNN network structure is a two-stage framework, the first stage of scanning images and generating Proposals (promusals, i.e., areas that may contain an object), the second stage of classifying Proposals and generating bounding boxes and masks. The Mask R-CNN performs pixel level segmentation by adding a branch to the Faster R-CNN, which is a full convolution network based on convolutional neural network feature mapping, and then the branch outputs a binary Mask to indicate whether a given pixel is part of the target object. Once these masks are generated, MaskR-CNN combines RoIAlign with the classification and bounding box from Faster R-CNN for accurate segmentation. And a binary mask is independently predicted for each target by using the average binary cross entropy loss, so that the introduction of inter-class competition is avoided, and the segmentation performance is greatly improved.
The Mask R-CNN example segmentation network defines a multitask loss function, which comprises 3 parts and has the formula:
L＝Lcls+Lbox+Lmax
wherein L is a loss function, LclsTo classify errors, LboxTo detect errors, LmaxIs the loss of semantically split branches. L isclsAnd LboxAnd (4) predicting the belonged category and the regression frame coordinate value of each ROI by utilizing full-connection layer processing. L ismaxThe target for each ROI is segmented and given a mask representation. Inputting each feature map of mask branches, and outputting a k multiplied by m feature map after a series of convolution and transposition convolution operations, wherein k represents the dimension of output and is the total category number, and each dimension isCorresponding to one category, the competition between the categories can be effectively avoided, and m x m represents the size of the characteristic diagram, LmaxTo average the binary cross entropy function, the function classifies each pixel. And each dimension is subjected to secondary classification by using a sigmoid function, and whether the dimension is the category is judged.
For RGB visible light image recognition of the obtained Aggregate, the invention uses an improved Mask R-CNN convolutional neural network model-Aggregate Net model to classify and recognize the wide-gradation mixed pebble Aggregate. The Aggregate Net model has strong parallel target detection and instance segmentation capabilities, and compared with other algorithms, the algorithm can detect more detailed characteristics of visible light images, and finally achieves segmentation detection of the on-site stacked sand-containing water-containing wide-graded mixed Aggregate. Each target in the invention is each pebble region in the stacked aggregates, a binary mask is finally predicted for each pebble, and the minimum circumscribed rectangle is calculated as the long diameter and short diameter of each Aggregate according to the mask.
The invention establishes an automatically constructed Aggregate Net network model, and for each ROI, LclsAnd the method is responsible for predicting the category of the target aggregate, if the ROI is predicted to be the aggregate, only the relative entropy error of the aggregate is used as error calculation when the loss of the ROI is segmented, and other categories do not participate in the loss function. Convolution calculation of the Aggregate Net network model to obtain a plurality of regions of interest, LclsPredict each region of interest class as aggregate, LboxPredicting position coordinates, L, of regression boxes of a plurality of targetsmaxAnd independently predicting a binary mask for the positions of a plurality of regression boxes by using average binary cross entropy loss and a Sigmoid function, wherein the contour information of the divided objects is placed at a plurality of different depths of layers, and different objects are represented by masks with different colors in a division map. For a larger number of wide-graded mixed Aggregate images, the Aggregate Net model generates a corresponding binary mask for each Aggregate segmentation.
The invention establishes an Aggregate Net network model on the basis of a Mask R-CNN network structure, utilizes ENVI software to preprocess image data, marks multi-angle stacked Aggregate images as a training set of the model at the early stage, uses ResNet50 as a feature extractor, adopts a leave-one method to verify and optimize the model, and finally determines the network parameters and the model of the better segmented stacked wide-gradation mixed Aggregate image.
Claims (4)
1. A method for detecting the empty space and the ground of wide-graded mixed aggregate is characterized by comprising the following steps:
s1, selecting a parent metal excavated from a river bed or a construction site as a detection aggregate, stacking all the detection aggregates in a primary storage pile, airing for a period of time, carrying out grading detection on the surface layer stacked aggregates by adopting an unmanned aerial vehicle LiDAR and unmanned aerial vehicle photogrammetry technology, and establishing a first grading curve;
step S1 includes the following substeps:
s11, selecting a parent material excavated from a river bed or a construction site as a detection aggregate, stacking all the detection aggregates in a primary storage pile for airing for a period of time, and acquiring surface aggregate images of the surface stacked aggregates by adopting an unmanned aerial vehicle LiDAR and an unmanned aerial vehicle photogrammetry technology;
s12, performing example segmentation and recognition detection on the surface Aggregate image by adopting a pre-trained Aggregate Net model based on machine learning to obtain a predicted frame and a mask of each surface Aggregate;
s13, assuming the shape of aggregate particles as an ellipsoid, and calculating the volume of the ellipsoid of the wide-graded mixed aggregate according to the predicted frame and the mask of each surface aggregate;
s14, converting the screening mass ratio of the wide-graded mixed aggregate into a volume ratio according to the same density of the same aggregate particles, obtaining the mass of each particle group of the wide-graded mixed aggregate in the total mass percentage, and constructing a first grading curve of the wide-graded mixed aggregate;
the Aggregate Net model includes: the method comprises the following steps of (1) providing a region suggestion network, a scene classification model and a semantic segmentation model;
the area suggestion network generates candidate areas, and each anchor point generates 9 candidate areas with 3 sizes and 3 length-width ratios;
the scene classification model is used for classifying and regressing the candidate regions, namely classifying the candidate regions into pebble and non-pebble, and the regression is to adjust the positions of the frame marks to be closer to the actual positions;
the semantic segmentation model is responsible for segmenting pebble boundaries in each candidate frame, a scale A obtained by photogrammetry of an unmanned aerial vehicle and a scale B obtained by an industrial camera erected above a conveying ground are respectively obtained through the flight height of the unmanned aerial vehicle and the height of two camera sensors erected above a conveying belt, the predicted major diameter a and the predicted minor diameter B of each aggregate are calculated, and then the predicted particle diameter d of each aggregate is calculated;
s2, randomly sampling the wide-graded mixed aggregate on the river bed or the primary storage pile excavated on the construction site in a large range on site by adopting a manual screening method, and establishing a second grading curve;
s3, performing curve fitting on the first grading curve and the second grading curve to establish a third grading curve;
s4, grabbing a certain amount of wide-gradation mixed aggregate from the primary storage pile to serve as aggregate to be detected, collecting images of the aggregate to be detected on a conveyor belt through a combined device of a photogrammetry technology and a laser radar technology, and establishing a fourth gradation curve;
s5, randomly sampling the wide-gradation mixed aggregate in the step S4 to obtain a manual screening detection sample, and performing field manual screening on the manual screening detection sample to establish a fifth gradation curve;
s6, performing curve fitting on the fourth grading curve and the fifth grading curve to establish a sixth grading curve;
and S7, fitting the third grading curve and the sixth grading curve by adopting a least square method to establish a seventh grading curve, namely the final grading curve based on the wide-grading mixed aggregate 'air-ground integration' detection method.
2. The method for integrally detecting the wide-graded mixed aggregate empty space according to claim 1, wherein the step S4 comprises the following substeps:
s41, grabbing a certain amount of wide-graded mixed aggregate from the primary storage pile to serve as aggregate to be detected, and placing the aggregate to be detected on a conveyor belt;
s42, erecting a combined device of photogrammetry technology and laser radar technology on the conveyor belt, and collecting the aggregate image to be detected on the conveyor belt for the aggregate to be detected on the conveyor belt;
s43, carrying out example segmentation and identification detection on the Aggregate image to be detected on the conveyor belt by using a pre-trained Aggregate Net model based on machine learning to obtain a prediction frame and a mask of each surface Aggregate on the conveyor belt;
s44, assuming the shape of aggregate particles as an ellipsoid, calculating the area of an ellipsoid of the wide-graded mixed aggregate on the conveyor belt according to the predicted frame and the mask of each surface aggregate on the conveyor belt, and obtaining the thickness of the aggregate by a laser radar technology to obtain the volume of the ellipsoid approximate to the volume of the aggregate;
s45, converting the screening mass ratio of the wide-graded mixed aggregate into a volume ratio according to the same density of the same aggregate particles, obtaining the mass of each particle group of the wide-graded mixed aggregate on the conveyor belt accounting for the total mass percentage, and constructing a fourth grading curve of the wide-graded mixed aggregate.
3. The method for detecting empty and ground integration of wide-graded mixed aggregate according to claim 2, wherein the approximate ellipsoid volume calculation formula in the step S44 is as follows:
wherein, a is the predicted major diameter, b is the predicted minor diameter size, and h is the height of each aggregate on the surface layer obtained by the laser radar.
4. The method for detecting the empty and land integration of the wide-graded mixed aggregates according to claim 1, wherein the calculation formula of the predicted particle size d of each aggregate is as follows:
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111058008.XA CN113740216B (en) | 2021-09-09 | 2021-09-09 | Air-ground integrated detection method for wide-gradation mixed aggregate |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111058008.XA CN113740216B (en) | 2021-09-09 | 2021-09-09 | Air-ground integrated detection method for wide-gradation mixed aggregate |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113740216A CN113740216A (en) | 2021-12-03 |
| CN113740216B true CN113740216B (en) | 2022-05-24 |
Family
ID=78737661
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111058008.XA Active CN113740216B (en) | 2021-09-09 | 2021-09-09 | Air-ground integrated detection method for wide-gradation mixed aggregate |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113740216B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101776566A (en) * | 2010-03-18 | 2010-07-14 | 长安大学 | Digital image-based aggregate grading real-time detection method |
| CN105699258A (en) * | 2016-01-28 | 2016-06-22 | 华侨大学 | Online detection device and method of fine aggregates |
| CN106228536A (en) * | 2016-07-12 | 2016-12-14 | 四川大学 | The earth and rockfill dam dam material grain composition method of inspection based on Digital Image Processing |
| CN109856377A (en) * | 2019-01-30 | 2019-06-07 | 中北大学 | The method that roadbase mixed coarse aggregate gradation is determined by crushing test |
| CN110349637A (en) * | 2019-06-13 | 2019-10-18 | 东南大学 | Aggregate ambient interfaces transition region volume fraction prediction technique, device and terminal device |
| CN110458119A (en) * | 2019-08-15 | 2019-11-15 | 中国水利水电科学研究院 | A kind of aggregate gradation method for quickly identifying of non-contact measurement |
| CN111751253A (en) * | 2020-07-06 | 2020-10-09 | 重庆理工大学 | Forming method and quality detection method of concrete aggregate detection model |
| CN112017164A (en) * | 2020-08-18 | 2020-12-01 | 中国水利水电科学研究院 | Soil and stone material grading detection method based on depth threshold convolution model |
| CN112528913A (en) * | 2020-12-18 | 2021-03-19 | 中山艾尚智同信息科技有限公司 | Grit particulate matter particle size detection analytic system based on image |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110004333A1 (en) * | 2009-07-01 | 2011-01-06 | Icrete International, Inc. | Superior concrete mix design with workability optimized gradation and fixed paste volume |
-
2021
- 2021-09-09 CN CN202111058008.XA patent/CN113740216B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101776566A (en) * | 2010-03-18 | 2010-07-14 | 长安大学 | Digital image-based aggregate grading real-time detection method |
| CN105699258A (en) * | 2016-01-28 | 2016-06-22 | 华侨大学 | Online detection device and method of fine aggregates |
| CN106228536A (en) * | 2016-07-12 | 2016-12-14 | 四川大学 | The earth and rockfill dam dam material grain composition method of inspection based on Digital Image Processing |
| CN109856377A (en) * | 2019-01-30 | 2019-06-07 | 中北大学 | The method that roadbase mixed coarse aggregate gradation is determined by crushing test |
| CN110349637A (en) * | 2019-06-13 | 2019-10-18 | 东南大学 | Aggregate ambient interfaces transition region volume fraction prediction technique, device and terminal device |
| CN110458119A (en) * | 2019-08-15 | 2019-11-15 | 中国水利水电科学研究院 | A kind of aggregate gradation method for quickly identifying of non-contact measurement |
| CN111751253A (en) * | 2020-07-06 | 2020-10-09 | 重庆理工大学 | Forming method and quality detection method of concrete aggregate detection model |
| CN112017164A (en) * | 2020-08-18 | 2020-12-01 | 中国水利水电科学研究院 | Soil and stone material grading detection method based on depth threshold convolution model |
| CN112528913A (en) * | 2020-12-18 | 2021-03-19 | 中山艾尚智同信息科技有限公司 | Grit particulate matter particle size detection analytic system based on image |
Non-Patent Citations (1)
| Title |
|---|
| 采用数学图像法进行粗骨料粒度分布分析;范德增;《国外耐火材料》;19990430(第04期);第13-18页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113740216A (en) | 2021-12-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wu et al. | Road pothole extraction and safety evaluation by integration of point cloud and images derived from mobile mapping sensors | |
| Kumar et al. | Automated road markings extraction from mobile laser scanning data | |
| Biçici et al. | An approach for the automated extraction of road surface distress from a UAV-derived point cloud | |
| Brasington et al. | Methodological sensitivity of morphometric estimates of coarse fluvial sediment transport | |
| CN103389103B (en) | A kind of Characters of Geographical Environment map structuring based on data mining and air navigation aid | |
| Kim et al. | Deep learning-based underground object detection for urban road pavement | |
| CN110726677B (en) | Polluted site remote sensing detection and space hot area identification system and method | |
| CN108288059A (en) | A kind of building waste monitoring method based on high-definition remote sensing technology | |
| CN111028255A (en) | Farmland area pre-screening method and device based on prior information and deep learning | |
| CN109816707A (en) | A kind of field of opencast mining information extracting method based on high-resolution satellite image | |
| CN110363299B (en) | Spatial case reasoning method for outcrop rock stratum layering | |
| CN115512247A (en) | Regional building damage grade assessment method based on image multi-parameter extraction | |
| Yu et al. | Road manhole cover delineation using mobile laser scanning point cloud data | |
| CN115330720A (en) | Unmanned aerial vehicle image earth surface solid waste detection model and detection method | |
| CN113469097B (en) | Multi-camera real-time detection method for water surface floaters based on SSD network | |
| CN117151430B (en) | Small watershed soil and water conservation treatment priority remote sensing evaluation method | |
| CN113740216B (en) | Air-ground integrated detection method for wide-gradation mixed aggregate | |
| CN114067245A (en) | Method and system for identifying hidden danger of external environment of railway | |
| Korzeniowska et al. | Generating DEM from LiDAR data–comparison of available software tools | |
| Naorem et al. | Robustness of rule sets using VHR imagery to detect informal settlements-a case of Mumbai, India | |
| Bernardo et al. | Techniques of geomatics and soft computing for the monitoring of infrastructures and the management of big data | |
| Yastikli et al. | Automatic 3D building model generations with airborne LiDAR data | |
| Forghani et al. | Object-based classification of multi-sensor optical imagery to generate terrain surface roughness information for input to wind risk simulation | |
| Zhu et al. | A capsnets approach to pavement crack detection using mobile laser scannning point clouds | |
| Sun et al. | Check dam extraction from remote sensing images using deep learning and geospatial analysis: A case study in the Yanhe River Basin of the Loess Plateau, China |
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 |