CN113740216B - Air-ground integrated detection method for wide-gradation mixed aggregate - Google Patents
Air-ground integrated detection method for wide-gradation mixed aggregateDownload PDFInfo
- Publication number
- CN113740216B CN113740216BCN202111058008.XACN202111058008ACN113740216BCN 113740216 BCN113740216 BCN 113740216BCN 202111058008 ACN202111058008 ACN 202111058008ACN 113740216 BCN113740216 BCN 113740216B
- Authority
- CN
- China
- Prior art keywords
- aggregate
- wide
- detection
- conveyor belt
- predicted
- 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
- 238000001514detection methodMethods0.000titleclaimsabstractdescription47
- 239000002245particleSubstances0.000claimsabstractdescription40
- 238000005516engineering processMethods0.000claimsabstractdescription35
- 230000011218segmentationEffects0.000claimsabstractdescription32
- 238000000034methodMethods0.000claimsabstractdescription20
- 238000005070samplingMethods0.000claimsabstractdescription6
- 238000010276constructionMethods0.000claimsdescription14
- 238000012216screeningMethods0.000claimsdescription14
- 239000000463materialSubstances0.000claimsdescription13
- 238000004364calculation methodMethods0.000claimsdescription9
- 239000002344surface layerSubstances0.000claimsdescription7
- 238000013145classification modelMethods0.000claimsdescription6
- 238000010801machine learningMethods0.000claimsdescription6
- 230000010354integrationEffects0.000claimsdescription5
- 239000002184metalSubstances0.000claims1
- 238000007873sievingMethods0.000abstractdescription9
- 238000005259measurementMethods0.000abstractdescription4
- 238000013135deep learningMethods0.000abstractdescription2
- 238000013527convolutional neural networkMethods0.000description13
- 238000007689inspectionMethods0.000description8
- 230000006870functionEffects0.000description6
- 238000012360testing methodMethods0.000description5
- 239000004576sandSubstances0.000description4
- 238000004458analytical methodMethods0.000description3
- 210000000988bone and boneAnatomy0.000description3
- 239000000203mixtureSubstances0.000description3
- 238000012545processingMethods0.000description3
- 239000004575stoneSubstances0.000description3
- XLYOFNOQVPJJNP-UHFFFAOYSA-NwaterSubstancesOXLYOFNOQVPJJNP-UHFFFAOYSA-N0.000description3
- 239000010426asphaltSubstances0.000description2
- 239000004567concreteSubstances0.000description2
- 238000011156evaluationMethods0.000description2
- 239000010410layerSubstances0.000description2
- 238000011160researchMethods0.000description2
- 239000013049sedimentSubstances0.000description2
- 238000013473artificial intelligenceMethods0.000description1
- 230000009286beneficial effectEffects0.000description1
- 239000004566building materialSubstances0.000description1
- 239000011362coarse particleSubstances0.000description1
- 239000003086colorantSubstances0.000description1
- 238000005056compactionMethods0.000description1
- 230000007812deficiencyEffects0.000description1
- 238000011161developmentMethods0.000description1
- 230000018109developmental processEffects0.000description1
- 238000002474experimental methodMethods0.000description1
- 239000000945fillerSubstances0.000description1
- 238000005457optimizationMethods0.000description1
- 238000003672processing methodMethods0.000description1
- 239000002689soilSubstances0.000description1
- 238000012549trainingMethods0.000description1
- 238000010200validation analysisMethods0.000description1
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
Translated fromChinese
Description
Translated fromChinese技术领域technical field
本发明涉及骨料检测技术领域，具体涉及一种宽级配混合骨料空地一体级配检测方法。The invention relates to the technical field of aggregate detection, in particular to a wide-graded mixed aggregate open-ground integrated gradation detection method.
背景技术Background technique
建设工程和水利工程作为我国国民经济建设及社会发展的重要基础设施，全面提升其安全性刻不容缓。各种宽级配混合骨料是水利、公路、铁路及建筑工程的主要建筑材料。合理的骨料配合比，是影响工程结构耐久性和保障工程运行安全性的重要因素。为了保证填方工程的压实质量，提高抗变形能力和抗渗性能，采用级配合理的填料尤为重要。其中，骨料在混凝土和胶结砂砾石颗粒料中的体积比占50％～70％以上，骨料级配是衡量宽级配混合材料力学性能的重要因素。如何快速、准确检测并判断填筑土石料的级配的合理性一直是工程界关注的焦点问题。目前，对于骨料级配的检测方法主要是依靠传统的人工或机械筛分法，经统计计算后得到宽级配混合骨料的级配曲线。传统筛分法虽然技术成熟，但每次计算的样品数量有限、耗费人工、效率低影响施工进度，已无法满足当前施工技术数字化、自动化与智能化的发展需求。因此，有必要开发一种非接触式智能化快速检测骨料粒径的方法，从而实现快速动态获取宽级配混合骨料级配。Construction projects and water conservancy projects are important infrastructures for my country's national economic construction and social development, and it is imperative to comprehensively improve their safety. Various wide-graded mixed aggregates are the main building materials for water conservancy, highway, railway and construction projects. Reasonable aggregate ratio is an important factor affecting the durability of engineering structures and ensuring the safety of engineering operations. In order to ensure the compaction quality of the filling project and improve the anti-deformation ability and impermeability performance, it is particularly important to use a reasonable gradation of fillers. Among them, the volume ratio of aggregate in concrete and cemented sand and gravel particles accounts for more than 50% to 70%, and aggregate gradation is an important factor to measure the mechanical properties of wide-graded mixed materials. How to quickly and accurately detect and judge the rationality of the grading of earth and stone for filling has always been the focus of the engineering community. At present, the detection method for aggregate gradation mainly relies on the traditional manual or mechanical sieving method, and the gradation curve of the wide-graded mixed aggregate is obtained after statistical calculation. Although the traditional sieving method is mature in technology, the number of samples calculated each time is limited, labor-intensive, and low efficiency affect the construction progress. Therefore, it is necessary to develop a non-contact intelligent and rapid detection method of aggregate particle size, so as to realize the rapid and dynamic acquisition of wide gradation mixed aggregate gradation.
目前，国内外研究主要集中在矿石粒度、沥青混合料均匀性、粒径分布、集料颗粒特征等方面，有学者基于数字图像处理技术及分析理论对骨料粒径进行检测。DayakarPenumadu博士使用自动测试机快速确定未结合的骨料的粒度分布。美国的LeeJ.R.J等利用激光三角测量法，开发了从粗粒料颗粒表面采集分析颗粒的三维数据的系统。Sulaiman M S等人以河床砂石为试样，使用自动图像处理技术分析粒径分布特征。NorbertH.Maerz设计一种面向单源视觉多视角采集、评价颗粒形态的系统。Hyoungkwan Kim基于激光集料分析系统，进行了各个具体参数的实验与评价，并验证该系统的可靠性。周建华等人研究骨料采集装置，得到分散骨料时的各参数指数。蔡改贫等人利用CCD摄像组件采集颗粒图像，基于图像处理方法监测矿石粒度。孙东坡基于图像识别技术，确定推移质泥沙的输砂率。彭勇利用数字图像处理技术对定量描述沥青混合料均匀性的方法进行了研究，提出了描述混合料均匀性的指标参数。图像识别技术方兴未艾，已经在部分行业有所尝试，土石料与矿石、泥沙、混凝土骨料具有几何特征上的相似性，利用图像识别技术进行土石料粒径级配的检测具有技术上的可行性。但是单纯的图像识别技术，无法解决混合堆叠骨料的级配识别及检测问题，因此，亟需开发一种应用于大量宽级配混合骨料粒径级配检测的高效、便捷方法。At present, research at home and abroad mainly focuses on ore particle size, asphalt mixture uniformity, particle size distribution, aggregate particle characteristics, etc. Some scholars detect aggregate particle size based on digital image processing technology and analysis theory. Dr. Dayakar Penumadu used an automatic testing machine to quickly determine the particle size distribution of unbound aggregates. LeeJ.R.J in the United States developed a system for collecting and analyzing three-dimensional data of particles from the surface of coarse particles by using laser triangulation method. Sulaiman M S et al. used riverbed sand and gravel as samples to analyze the particle size distribution characteristics using automatic image processing technology. NorbertH.Maerz designed a system for single-source vision multi-view acquisition and evaluation of particle morphology. Based on the laser aggregate analysis system, Hyoungkwan Kim conducted experiments and evaluations of various specific parameters, and verified the reliability of the system. Zhou Jianhua et al. studied the aggregate collection device and obtained the parameter indices when dispersing the aggregate. Cai Gaipian et al. used CCD camera components to collect particle images, and monitor ore particle size based on image processing methods. Sun Dongpo determined the sand transport rate of bedding sediment based on image recognition technology. Peng Yong used digital image processing technology to study the method of quantitatively describing the uniformity of asphalt mixture, and put forward the index parameters to describe the uniformity of the mixture. Image recognition technology is in the ascendant, and it has been tried in some industries. Earth and stone materials have similarities in geometric characteristics with ore, sediment, and concrete aggregates. It is technically feasible to use image recognition technology to detect the particle size gradation of soil and stone materials. sex. However, simple image recognition technology cannot solve the problem of gradation identification and detection of mixed stacked aggregates. Therefore, it is urgent to develop an efficient and convenient method for particle size gradation detection of a large number of wide-graded mixed aggregates.
发明内容SUMMARY OF THE INVENTION
针对现有技术中的上述不足，本发明提供的一种宽级配混合骨料空地一体级配检测方法解决了传统筛分法存在每次计算的样品数量有限、耗费人工、效率低影响施工进度的问题。In view of the above deficiencies in the prior art, the invention provides a wide-graded mixed aggregate open space integrated grading detection method, which solves the problems of the traditional sieving method that the number of samples for each calculation is limited, labor-intensive, and low efficiency, which affects the construction progress. The problem.
为了达到上述发明目的，本发明采用的技术方案为：一种宽级配混合骨料空地一体级配检测方法，包括以下步骤：In order to achieve the above-mentioned purpose of the invention, the technical solution adopted in the present invention is: a wide-graded mixed aggregate open-ground integrated grading detection method, comprising the following steps:
S1、选取河床或施工现场开挖的母材作为检测骨料，将检测骨料全部堆放在一次储料堆进行一段时间的晾晒，采用无人机LiDAR和无人机摄影测量技术进行表层堆叠骨料的级配检测，建立第一级配曲线；S1. Select the base material excavated from the river bed or the construction site as the inspection aggregate, stack all the inspection aggregates in a storage pile for a period of time to dry, and use the UAV LiDAR and UAV photogrammetry technology to stack the surface layer of the aggregate. The gradation detection of the material is carried out, and the first gradation curve is established;
S2、对河床或施工现场开挖的一次储料堆上的宽级配混合骨料采用人工筛分法进行现场大范围随机采样，建立第二级配曲线；S2. Use the manual screening method to randomly sample the wide-scale mixed aggregate on the primary storage pile excavated on the river bed or the construction site, and establish the second gradation curve;
S3、将第一级配曲线与第二级配曲线进行曲线拟合建立第三级配曲线；S3, performing curve fitting on the first gradation curve and the second gradation curve to establish a third gradation curve;
S4、从一次储料堆上抓取一定量的宽级配混合骨料作为待检测骨料，通过摄影测量技术和激光雷达技术的组合装置采集传送带上待检测骨料的图像，建立第四级配曲线；S4. Grab a certain amount of wide-graded mixed aggregate from the primary storage pile as the aggregate to be inspected, and collect the image of the aggregate to be inspected on the conveyor belt through the combined device of photogrammetry technology and lidar technology to establish the fourth level matching curve;
S5、对步骤S4中的宽级配混合骨料进行随机抽样，得到人工筛分检测样本，对人工筛分检测样本进行现场手动筛分，建立第五级配曲线；S5. Perform random sampling on the wide-graded mixed aggregate in step S4 to obtain an artificial screening test sample, perform manual screening on the artificial screening test sample on-site, and establish a fifth gradation curve;
S6、将第四级配曲线和第五级配曲线进行曲线拟合建立第六级配曲线；S6, perform curve fitting on the fourth gradation curve and the fifth gradation curve to establish the sixth gradation curve;
S7、采用最小二乘法曲线拟合第三级配曲线和第六级配曲线，建立第七级配曲线，即为基于宽级配混合骨料“空地一体”检测方法的最终级配曲线。S7. Use the least squares curve to fit the third gradation curve and the sixth gradation curve, and establish the seventh gradation curve, which is the final gradation curve based on the "air-ground integration" detection method of the wide-graded mixed aggregate.
进一步地，步骤S1包括以下分步骤：Further, step S1 includes the following sub-steps:
S11、选取河床或施工现场开挖的母材作为检测骨料，将检测骨料全部堆放在一次储料堆进行一段时间的晾晒，采用无人机LiDAR和无人机摄影测量技术采集表层堆叠骨料的表层骨料图像；S11. Select the base material excavated from the river bed or the construction site as the inspection aggregate, stack all the inspection aggregates in a storage pile for a period of time to dry, and use the UAV LiDAR and UAV photogrammetry technology to collect the surface stacked bone The surface aggregate image of the material;
S12、基于机器学习采用预先训练的Aggregate Net模型对表层骨料图像进行实例分割及识别检测，得到每颗表层骨料的预测边框和掩膜；S12, using a pre-trained Aggregate Net model based on machine learning to perform instance segmentation, identification and detection on the surface aggregate image, and obtain the predicted frame and mask of each surface aggregate;
S13、将骨料颗粒形状假设成椭球，根据每颗表层骨料的预测边框和掩膜，计算宽级配混合骨料的椭球体体积；S13, assuming the shape of the aggregate particles to be an ellipsoid, and calculating the ellipsoid volume of the wide-graded mixed aggregate according to the predicted frame and mask of each surface aggregate;
S14、根据同一骨料颗粒的密度相同，将宽级配混合骨料的筛分质量比转化为体积比，得到宽级配混合骨料的各个粒组质量占总质量百分比，构建宽级配混合骨料的第一级配曲线。S14. According to the same density of the same aggregate particles, the sieving mass ratio of the wide-graded mixed aggregate is converted into a volume ratio, and the mass of each particle group of the wide-graded mixed aggregate is obtained as a percentage of the total mass, and a wide-graded mixed aggregate is constructed. The first gradation curve of the aggregate.
进一步地，步骤S4包括以下分步骤：Further, step S4 includes the following sub-steps:
S41、从一次储料堆上抓取一定量的宽级配混合骨料作为待检测骨料，将待检测骨料放置在传送带上；S41. Grab a certain amount of wide-graded mixed aggregate from the primary storage pile as the aggregate to be inspected, and place the aggregate to be inspected on the conveyor belt;
S42、在传送带上架设摄影测量技术和激光雷达技术的组合装置，对传送带上的待检测骨料采集传送带上待检测骨料图像；S42. A combined device of photogrammetry technology and lidar technology is erected on the conveyor belt, and images of the aggregate to be detected on the conveyor belt are collected from the aggregate to be detected on the conveyor belt;
S43、基于机器学习使用预先训练的Aggregate Net模型对传送带上待检测骨料图像进行实例分割及识别检测，得到传送带上每颗表层骨料的预测边框和掩膜；S43, using the pre-trained Aggregate Net model based on machine learning to perform instance segmentation and identification detection on the image of the aggregate to be detected on the conveyor belt, and obtain the predicted frame and mask of each surface aggregate on the conveyor belt;
S44、将骨料颗粒形状假设成椭球，根据传送带上每颗表层骨料的预测边框和掩膜，计算传送带上宽级配混合骨料的椭球体面积，通过激光雷达技术获取骨料厚度，得到近似椭球体体积；S44. Assuming the shape of the aggregate particles to be an ellipsoid, according to the predicted frame and mask of each surface aggregate on the conveyor belt, calculate the ellipsoid area of the wide-graded mixed aggregate on the conveyor belt, and obtain the thickness of the aggregate through lidar technology. get the approximate ellipsoid volume;
S45、根据同一骨料颗粒的密度相同，将宽级配混合骨料的筛分质量比转化为体积比，得到传送带上宽级配混合骨料的各个粒组质量占总质量百分比，构建宽级配混合骨料的第四级配曲线。S45. According to the same density of the same aggregate particles, convert the sieving mass ratio of the wide-graded mixed aggregate into a volume ratio to obtain the mass percentage of each particle group of the wide-graded mixed aggregate on the conveyor belt to the total mass, and construct a wide-graded mixed aggregate. The fourth gradation curve of the mixed aggregate.
进一步地，步骤S44中近似椭球体体积的计算公式为：Further, the calculation formula of the approximate ellipsoid volume in step S44 is:
其中，a为预测长径，b为预测短径尺寸，h为激光雷达获取的表层各骨料高度。Among them, a is the predicted long diameter, b is the predicted short diameter size, and h is the height of each aggregate in the surface layer obtained by the lidar.
进一步地，Aggregate Net模型包括：区域建议网络、场景分类模型和语义分割模型；Further, the Aggregate Net model includes: a region proposal network, a scene classification model and a semantic segmentation model;
所述区域建议网络生成候选区域，每一个锚点生成3种尺寸，3种长宽比，共9个候选区；The region proposal network generates candidate regions, and each anchor point generates 3 sizes, 3 aspect ratios, and a total of 9 candidate regions;
所述场景分类模型用于对候选区域进行分类和回归，分类即将候选区域分为卵石类和非卵石类，回归是将框标位置进行调整使其与实际更为接近；The scene classification model is used for classifying and regressing the candidate regions. The classification is to divide the candidate regions into pebbles and non-pebble classes, and the regression is to adjust the position of the frame mark to make it closer to the actual situation;
所述语义分割模型负责分割出每一个候选框内的卵石边界，通过无人机飞行高度与传送带上方架设的两个相机传感器高度，分别获取无人机摄影测量获取的比例尺A与传送地上方架设的工业相机获取的比例尺B，计算各个骨料的预测长径a和预测短径尺寸b，进而计算得到各骨料预测粒径d。The semantic segmentation model is responsible for segmenting the pebble boundary in each candidate frame, and through the flying height of the drone and the height of the two camera sensors erected above the conveyor belt, the scale A obtained by the photogrammetry of the UAV and the height of the two camera sensors erected above the conveyor belt are respectively obtained. The scale B obtained by the industrial camera is used to calculate the predicted long diameter a and predicted short diameter b of each aggregate, and then calculate the predicted particle size d of each aggregate.
进一步地，各骨料预测粒径d的计算公式为：Further, the calculation formula of the predicted particle size d of each aggregate is:
综上，本发明的有益效果为：To sum up, the beneficial effects of the present invention are:
(1)、本发明提出了一种宽级配混合骨料空地一体级配检测方法，基于大数据背景下的深度学习方法和激光雷达测量技术，提出了基于实例分割框架的骨料快速智能化检测分割算法Aggregate Net模型，该模型强大的并行目标检测和实例分割的能力，可以获取各骨料预测长径、预测短径以及骨料粒径，并且激光雷达测量技术获取骨料的厚度，从而精确计算出各骨料的等效体积，最后与人工采样筛分结果进行数据拟合，从而获得精度更高的级配曲线结果。(1) The present invention proposes a wide-graded mixed aggregate open-ground integrated grading detection method. Based on the deep learning method and lidar measurement technology under the background of big data, a fast and intelligent aggregate based on the instance segmentation framework is proposed. Detection and segmentation algorithm Aggregate Net model, which has powerful parallel target detection and instance segmentation capabilities, can obtain the predicted long diameter, predicted short diameter and aggregate particle size of each aggregate, and the LiDAR measurement technology can obtain the thickness of the aggregate, thereby Accurately calculate the equivalent volume of each aggregate, and finally perform data fitting with the manual sampling and sieving results, so as to obtain the gradation curve results with higher accuracy.
(2)、通过人工智能技术的骨料配比检测计算方法，实现骨料配比分析去除人工化、提高工作效率、实现动态非接触式检测，能够弥补我国水工及其他建筑物中传统骨料级配检测方法的不足，弥补当前筑坝骨料级配检测效率低、专业检测仪器不能大范围在现场使用、不能动态分析堆叠骨料形貌特征等短板。本发明研究对骨料级配检测理论和技术的提升与创新具有重要意义，进一步提高传统行业数字化、自动化、智能化的科技水平。(2) Through the calculation method of aggregate ratio detection and calculation of artificial intelligence technology, the artificial aggregate ratio analysis can be removed, the work efficiency can be improved, and the dynamic non-contact detection can be realized, which can make up for the traditional bone structure in hydraulic engineering and other buildings in my country. The shortcomings of the material gradation detection method make up for the shortcomings of the current dam-building aggregate gradation detection efficiency, the inability of professional detection instruments to be used on-site in a large range, and the inability to dynamically analyze the morphology and characteristics of stacked aggregates. The research of the invention is of great significance to the improvement and innovation of the aggregate gradation detection theory and technology, and further improves the technological level of digitization, automation and intelligence of traditional industries.
附图说明Description of drawings
图1为一种宽级配混合骨料空地一体检测方法的流程图；Fig. 1 is a flow chart of an integrated detection method for wide-graded mixed aggregate open space;
图2为Aggregate Net模型算法流程图。Figure 2 is a flowchart of the Aggregate Net model algorithm.
具体实施方式Detailed ways
下面对本发明的具体实施方式进行描述，以便于本技术领域的技术人员理解本发明，但应该清楚，本发明不限于具体实施方式的范围，对本技术领域的普通技术人员来讲，只要各种变化在所附的权利要求限定和确定的本发明的精神和范围内，这些变化是显而易见的，一切利用本发明构思的发明创造均在保护之列。The specific embodiments of the present invention are described below to facilitate those skilled in the art to understand the present invention, but it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, as long as various changes Such changes are obvious within the spirit and scope of the present invention as defined and determined by the appended claims, and all inventions and creations utilizing the inventive concept are within the scope of protection.
如图1所示，一种宽级配混合骨料空地一体检测方法，包括以下步骤：As shown in Figure 1, an integrated detection method for wide-graded mixed aggregate open space includes the following steps:
S1、选取河床或施工现场开挖的母材作为检测骨料，将检测骨料全部堆放在一次储料堆进行一段时间的晾晒，采用无人机LiDAR和无人机摄影测量技术进行表层堆叠骨料的级配检测，建立第一级配曲线；S1. Select the base material excavated from the river bed or the construction site as the inspection aggregate, stack all the inspection aggregates in a storage pile for a period of time to dry, and use the UAV LiDAR and UAV photogrammetry technology to stack the surface layer of the aggregate. The gradation detection of the material is carried out, and the first gradation curve is established;
步骤S1包括以下分步骤：Step S1 includes the following sub-steps:
S11、选取河床或施工现场开挖的母材作为检测骨料，将检测骨料全部堆放在一次储料堆进行一段时间的晾晒，采用无人机LiDAR和无人机摄影测量技术采集表层堆叠骨料的表层骨料图像；S11. Select the base material excavated from the river bed or the construction site as the inspection aggregate, stack all the inspection aggregates in a storage pile for a period of time to dry, and use the UAV LiDAR and UAV photogrammetry technology to collect the surface stacked bone The surface aggregate image of the material;
S12、基于机器学习采用预先训练的Aggregate Net模型对表层骨料图像进行实例分割及识别检测，得到每颗表层骨料的预测边框和掩膜；S12, using a pre-trained Aggregate Net model based on machine learning to perform instance segmentation, identification and detection on the surface aggregate image, and obtain the predicted frame and mask of each surface aggregate;
S13、将骨料颗粒形状假设成椭球，根据每颗表层骨料的预测边框和掩膜，计算宽级配混合骨料的椭球体体积；S13, assuming the shape of the aggregate particles to be an ellipsoid, and calculating the ellipsoid volume of the wide-graded mixed aggregate according to the predicted frame and mask of each surface aggregate;
宽级配混合骨料的椭球体体积的计算公式为：The formula for calculating the ellipsoid volume of the wide graded mixed aggregate is:
其中，a为预测长径，b为预测短径尺寸，h为激光雷达获取的表层各骨料高度。Among them, a is the predicted long diameter, b is the predicted short diameter size, and h is the height of each aggregate in the surface layer obtained by the lidar.
S14、根据同一骨料颗粒的密度相同，将宽级配混合骨料的筛分质量比转化为体积比，得到宽级配混合骨料的各个粒组质量占总质量百分比，构建宽级配混合骨料的第一级配曲线。S14. According to the same density of the same aggregate particles, the sieving mass ratio of the wide-graded mixed aggregate is converted into a volume ratio, and the mass of each particle group of the wide-graded mixed aggregate is obtained as a percentage of the total mass, and a wide-graded mixed aggregate is constructed. The first gradation curve of the aggregate.
S2、对河床或施工现场开挖的一次储料堆上的宽级配混合骨料采用人工筛分法进行现场大范围随机采样，建立第二级配曲线；S2. Use the manual screening method to randomly sample the wide-scale mixed aggregate on the primary storage pile excavated on the river bed or the construction site, and establish the second gradation curve;
S3、将第一级配曲线与第二级配曲线进行曲线拟合建立第三级配曲线；S3, performing curve fitting on the first gradation curve and the second gradation curve to establish a third gradation curve;
S4、从一次储料堆上抓取一定量的宽级配混合骨料作为待检测骨料，通过摄影测量技术和激光雷达技术的组合装置采集传送带上待检测骨料的图像，建立第四级配曲线；S4. Grab a certain amount of wide-graded mixed aggregate from the primary storage pile as the aggregate to be inspected, and collect the image of the aggregate to be inspected on the conveyor belt through the combined device of photogrammetry technology and lidar technology to establish the fourth level matching curve;
步骤S4包括以下分步骤：Step S4 includes the following sub-steps:
S41、从一次储料堆上抓取一定量的宽级配混合骨料作为待检测骨料，将待检测骨料放置在传送带上；S41. Grab a certain amount of wide-graded mixed aggregate from the primary storage pile as the aggregate to be inspected, and place the aggregate to be inspected on the conveyor belt;
S42、在传送带上架设摄影测量技术和激光雷达技术的组合装置，对传送带上的待检测骨料采集传送带上待检测骨料图像；S42. A combined device of photogrammetry technology and lidar technology is erected on the conveyor belt, and images of the aggregate to be detected on the conveyor belt are collected from the aggregate to be detected on the conveyor belt;
S43、基于机器学习使用预先训练的Aggregate Net模型对传送带上待检测骨料图像进行实例分割及识别检测，得到传送带上每颗表层骨料的预测边框和掩膜；S43, using the pre-trained Aggregate Net model based on machine learning to perform instance segmentation and identification detection on the image of the aggregate to be detected on the conveyor belt, and obtain the predicted frame and mask of each surface aggregate on the conveyor belt;
S44、将骨料颗粒形状假设成椭球，根据传送带上每颗表层骨料的预测边框和掩膜，计算传送带上宽级配混合骨料的椭球体面积，通过激光雷达技术获取骨料厚度，得到近似椭球体体积；S44. Assuming the shape of the aggregate particles to be an ellipsoid, according to the predicted frame and mask of each surface aggregate on the conveyor belt, calculate the ellipsoid area of the wide-graded mixed aggregate on the conveyor belt, and obtain the thickness of the aggregate through lidar technology. get the approximate ellipsoid volume;
步骤S44中近似椭球体体积的计算公式为：The calculation formula of the approximate ellipsoid volume in step S44 is:
其中，a为预测长径，b为预测短径尺寸，h为激光雷达获取的表层各骨料高度。Among them, a is the predicted long diameter, b is the predicted short diameter size, and h is the height of each aggregate in the surface layer obtained by the lidar.
S45、根据同一骨料颗粒的密度相同，将宽级配混合骨料的筛分质量比转化为体积比，得到传送带上宽级配混合骨料的各个粒组质量占总质量百分比，构建宽级配混合骨料的第四级配曲线。S45. According to the same density of the same aggregate particles, convert the sieving mass ratio of the wide-graded mixed aggregate into a volume ratio to obtain the mass percentage of each particle group of the wide-graded mixed aggregate on the conveyor belt to the total mass, and construct a wide-graded mixed aggregate. The fourth gradation curve of the mixed aggregate.
S5、对步骤S4中的宽级配混合骨料进行随机抽样，得到人工筛分检测样本，对人工筛分检测样本进行现场手动筛分，建立第五级配曲线；S5. Perform random sampling on the wide-graded mixed aggregate in step S4 to obtain an artificial screening test sample, perform manual screening on the artificial screening test sample on-site, and establish a fifth gradation curve;
S6、将第四级配曲线和第五级配曲线进行曲线拟合建立第六级配曲线；S6, perform curve fitting on the fourth gradation curve and the fifth gradation curve to establish the sixth gradation curve;
S7、采用最小二乘法曲线拟合第三级配曲线和第六级配曲线，建立第七级配曲线，即为基于宽级配混合骨料“空地一体”级配检测方法的最终级配曲线。S7. Use the least squares curve to fit the third gradation curve and the sixth gradation curve, and establish the seventh gradation curve, which is the final gradation curve based on the "air-ground integration" gradation detection method for wide-graded mixed aggregates .
Aggregate Net模型包括：区域建议网络、场景分类模型和语义分割模型；Aggregate Net models include: region proposal network, scene classification model and semantic segmentation model;
所述区域建议网络生成候选区域，每一个锚点生成3种尺寸，3种长宽比，共9个候选区；The region proposal network generates candidate regions, and each anchor point generates 3 sizes, 3 aspect ratios, and a total of 9 candidate regions;
所述场景分类模型用于对候选区域进行分类和回归，分类即将候选区域分为卵石类和非卵石类，回归是将框标位置进行调整使其与实际更为接近；The scene classification model is used for classifying and regressing the candidate regions. The classification is to divide the candidate regions into pebbles and non-pebble classes, and the regression is to adjust the position of the frame mark to make it closer to the actual situation;
所述语义分割模型负责分割出每一个候选框内的卵石边界，通过无人机飞行高度与传送带上方架设的两个相机传感器高度，分别获取无人机摄影测量获取的比例尺A与传送地上方架设的工业相机获取的比例尺B，计算各个骨料的预测长径a和预测短径尺寸b，进而计算得到各骨料预测粒径d。The semantic segmentation model is responsible for segmenting the pebble boundary in each candidate frame, and through the flying height of the drone and the height of the two camera sensors erected above the conveyor belt, the scale A obtained by the photogrammetry of the UAV and the height of the two camera sensors erected above the conveyor belt are respectively obtained. The scale B obtained by the industrial camera is used to calculate the predicted long diameter a and predicted short diameter b of each aggregate, and then calculate the predicted particle size d of each aggregate.
各骨料预测粒径d的计算公式为：The formula for calculating the predicted particle size d of each aggregate is:
本发明在Mask R-CNN网络结构的基础上建立Aggregate Net网络模型，Mask R-CNN是一个实例分割算法，可以用来做“目标检测”、“目标实例分割”、“目标关键点检测”等，由目标检测模型生成预测框，语义分割模型分割得到具体的实例边界。Mask R-CNN网络结构是基于Faster-RCNN框架，在基础特征网络之后又加入了全连接的分割子网，由原来的“分类+回归”两个任务变为了“分类+回归+分割”三个任务。Mask R-CNN网络结构是一个两阶段的框架，第一个阶段扫描图像并生成提议(Proposals，即有可能包含一个目标的区域)，第二阶段分类提议并生成边界框和掩码。Mask R-CNN通过向Faster R-CNN添加一个分支来进行像素级分割，该分支是基于卷积神经网络特征映射的全卷积网络，继而分支输出一个二进制掩码来表示给定像素是否为目标对象的一部分。一旦生成这些掩码，MaskR-CNN将RoIAlign与来自Faster R-CNN的分类和边界框相结合，以便进行精确的分割。用平均二值交叉熵损失对每个目标独立地预测一个二值掩模，避免引入类别间竞争，大大提高了分割性能。The invention establishes the Aggregate Net network model on the basis of the Mask R-CNN network structure. Mask R-CNN is an instance segmentation algorithm, which can be used for "target detection", "target instance segmentation", "target key point detection", etc. , the prediction frame is generated by the target detection model, and the specific instance boundary is obtained by the semantic segmentation model. The Mask R-CNN network structure is based on the Faster-RCNN framework. After the basic feature network, a fully connected segmentation subnet is added. The original two tasks of "classification + regression" have become "classification + regression + segmentation" three tasks Task. The Mask R-CNN network structure is a two-stage framework, the first stage scans the image and generates proposals (Proposals, that is, regions that are likely to contain an object), and the second stage classifies the proposals and generates bounding boxes and masks. Mask R-CNN performs pixel-level segmentation by adding a branch to Faster R-CNN, which is a fully convolutional network based on convolutional neural network feature maps, which in turn outputs a binary mask to indicate whether a given pixel is a target part of the object. Once these masks are generated, MaskR-CNN combines RoIAlign with classification and bounding boxes from Faster R-CNN for accurate segmentation. A binary mask is independently predicted for each target with an average binary cross-entropy loss, avoiding the introduction of inter-class competition and greatly improving the segmentation performance.
Mask R-CNN实例分割网络定义了一个多任务的损失函数，包括3部分，公式为：The Mask R-CNN instance segmentation network defines a multi-task loss function, including 3 parts, the formula is:
L＝Lcls+Lbox+LmaxL=Lcls +Lbox +Lmax
其中，L为损失函数，Lcls为分类误差，Lbox为检测误差，Lmax是语义分割分支的损失。Lcls和Lbox利用全连接层处理，预测出每个ROI的所属类别和回归框坐标值。Lmax对每个ROI的目标进行分割，并赋予掩膜表示。输入mask分支的每个特征图，然后经过一系列卷积、转置卷积操作后输出k×m×m的特征图，其中k表示输出的维度也是总的类别数，每一个维度对应一个类别可以有效的避免类间竞争，m×m表示的是特征图的大小，Lmax为平均二值交叉熵函数，该函数会对每一个像素进行分类。每一个维度都利用sigmoid函数进行二分类，判断是否为此类别。Among them, L is the loss function, Lcls is the classification error, Lbox is the detection error, and Lmax is the loss of the semantic segmentation branch. Lcls and Lbox are processed by the fully connected layer to predict the category of each ROI and the coordinate value of the regression box.Lmax segments the object for each ROI and assigns a mask representation. Input each feature map of the mask branch, and then output a k×m×m feature map after a series of convolution and transposed convolution operations, where k represents the output dimension and the total number of categories, and each dimension corresponds to a category It can effectively avoid competition between classes, m×m represents the size of the feature map, andLmax is the average binary cross-entropy function, which classifies each pixel. Each dimension uses the sigmoid function for binary classification to determine whether it is this category.
对于获取骨料的RGB可见光图像识别，本发明使用改进的Mask R-CNN卷积神经网络模型——Aggregate Net模型，对宽级配混合卵石骨料进行分类和识别。Aggregate Net模型具有强大的并行目标检测和实例分割的能力，相对于其他算法来说，该算法能够将可见光图像更细节的特征检测出来，最终实现对现场堆叠含砂含水宽级配混合骨料的分割检测。本发明中的每个目标即为堆叠骨料中的每块卵石区域，最终对每块卵石预测一个二值掩模，根据掩膜计算其最小外接矩形为各骨料的长径、短径尺寸，本发明使用的AggregateNet网络结构算法流程图，如图2所示。For the RGB visible light image recognition of the obtained aggregate, the present invention uses the improved Mask R-CNN convolutional neural network model - the Aggregate Net model to classify and recognize the wide-graded mixed pebble aggregate. The Aggregate Net model has powerful parallel target detection and instance segmentation capabilities. Compared with other algorithms, this algorithm can detect more detailed features of visible light images, and finally realizes the on-site stacking of sand-containing and water-containing wide-graded mixed aggregates. Segmentation detection. Each target in the present invention is the area of each pebble in the stacked aggregate, and finally a binary mask is predicted for each pebble, and the minimum circumscribed rectangle calculated according to the mask is the length and short diameter of each aggregate. , the flow chart of the AggregateNet network structure algorithm used in the present invention is shown in FIG. 2 .
本发明建立自主搭建的Aggregate Net网络模型，对于每个ROI，Lcls负责预测目标骨料的类别，若预测感兴趣区域ROI为骨料的话，则在分割ROI过程的损失时只使用骨料类的相对熵误差作为误差计算，其他类别不参与此次损失函数中。Aggregate Net网络模型卷积计算得多个感兴趣区域，Lcls预测每个感兴趣区域类别为骨料，Lbox预测多个目标的回归框的位置坐标，Lmax采用平均二值交叉熵损失并使用Sigmoid函数对多个回归框的位置均独立地预测一个二值掩模，多个目标分割的目标轮廓信息放置在对应多个不同层深度，在分割图中用不同颜色的掩模来表示不同的目标。对于数量更多的宽级配混合骨料图像，Aggregate Net模型对每块骨料分割生成一个对应的二值掩模。The present invention establishes a self-built Aggregate Net network model. For each ROI, Lcls is responsible for predicting the category of the target aggregate. If the ROI in the region of interest is predicted to be aggregate, only the aggregate class is used when dividing the loss of the ROI process. The relative entropy error of is calculated as the error, and other categories do not participate in this loss function. Aggregate Net network model convolution calculates multiple regions of interest, Lcls predicts that each region of interest category is aggregate, Lbox predicts the position coordinates of the regression boxes of multiple targets, Lmax uses the average binary cross entropy loss and Use the Sigmoid function to independently predict a binary mask for the positions of multiple regression boxes. The target contour information of multiple target segmentations is placed in corresponding multiple different layer depths, and masks of different colors are used in the segmentation map to represent different The goal. For a larger number of wide-graded mixed aggregate images, the Aggregate Net model generates a corresponding binary mask for each aggregate segment.
本发明在Mask R-CNN网络结构的基础上建立Aggregate Net网络模型，利用ENVI软件进行图像数据的预处理，前期标注多角度堆叠骨料图像作为模型的训练集，使用ResNet50为特征提取器，采用留一法进行模型验证优化，最终确定较优的分割堆叠宽级配混合骨料图像的网络参数和模型。The invention establishes an Aggregate Net network model on the basis of the Mask R-CNN network structure, uses ENVI software to preprocess the image data, labels the multi-angle stacked aggregate images in the early stage as the training set of the model, uses ResNet50 as the feature extractor, and adopts The leave-one-out method is used for model validation and optimization, and finally the optimal network parameters and models for segmenting and stacking wide-graded mixed aggregate images are determined.
Claims (4)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111058008.XACN113740216B (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.XACN113740216B (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 |
| CN113740216Btrue CN113740216B (en) | 2022-05-24 |
Family
ID=78737661
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111058008.XAActiveCN113740216B (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 rapid identification method of concrete aggregate gradation by non-contact measurement |
| CN111751253A (en)* | 2020-07-06 | 2020-10-09 | 重庆理工大学 | A method for forming a concrete aggregate detection model and a quality detection method |
| CN112017164A (en)* | 2020-08-18 | 2020-12-01 | 中国水利水电科学研究院 | Soil and rock gradation 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
- 2021-09-09CNCN202111058008.XApatent/CN113740216B/enactiveActive
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 rapid identification method of concrete aggregate gradation by non-contact measurement |
| CN111751253A (en)* | 2020-07-06 | 2020-10-09 | 重庆理工大学 | A method for forming a concrete aggregate detection model and a quality detection method |
| CN112017164A (en)* | 2020-08-18 | 2020-12-01 | 中国水利水电科学研究院 | Soil and rock gradation 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 |
|---|---|---|
| CN110458119B (en) | A rapid identification method of concrete aggregate gradation by non-contact measurement | |
| Biçici et al. | An approach for the automated extraction of road surface distress from a UAV-derived point cloud | |
| Li et al. | Automated asphalt pavement damage rate detection based on optimized GA-CNN | |
| Koch et al. | Achievements and challenges in machine vision-based inspection of large concrete structures | |
| CN110263717B (en) | A land-use category determination method incorporating street view imagery | |
| 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 | |
| Liu et al. | Large-scale mapping of gully-affected areas: An approach integrating Google Earth images and terrain skeleton information | |
| CN114022474A (en) | Particle grading rapid detection method based on YOLO-V4 | |
| CN114723709A (en) | Tunnel disease detection method and device and electronic equipment | |
| Conley et al. | Using a deep learning model to quantify trash accumulation for cleaner urban stormwater | |
| CN113469097A (en) | SSD (solid State disk) network-based real-time detection method for water surface floating object multiple cameras | |
| Yu et al. | Road manhole cover delineation using mobile laser scanning point cloud data | |
| CN118038132A (en) | Method and equipment for analyzing real-time blending uniformity of gravel soil material | |
| CN118015494A (en) | Regional enhanced multi-phase three-dimensional monitoring method for erosion trench of mining area | |
| Shao et al. | Crack detection and measurement using PTZ camera–based image processing method on expressways | |
| CN113740216B (en) | Air-ground integrated detection method for wide-gradation mixed aggregate | |
| Dewali et al. | Combining OBIA, CNN, and UAV photogrammetry for automated avalanche deposit detection and characterization | |
| Zhang et al. | Research on pipeline defect detection based on optimized faster r-cnn algorithm | |
| Zhu et al. | A capsnets approach to pavement crack detection using mobile laser scannning point clouds | |
| CN109598310B (en) | A multi-factor sensitive facility identification method | |
| Bernardo et al. | Techniques of geomatics and soft computing for the monitoring of infrastructures and the management of big data | |
| Zhou et al. | A comparison of object-based with pixel-based land cover change detection in the Baltimore metropolitan area using multitemporal high resolution remote sensing data | |
| Korzeniowska et al. | Generating DEM from LiDAR data–comparison of available software tools | |
| Yastikli et al. | Building extraction using multi sensor systems |
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 |