WO2022057556A1 - 一种基于深度学习的端到端散斑投影三维测量方法 - Google Patents
一种基于深度学习的端到端散斑投影三维测量方法 Download PDFInfo
- Publication number
- WO2022057556A1 WO2022057556A1 PCT/CN2021/113187 CN2021113187W WO2022057556A1 WO 2022057556 A1 WO2022057556 A1 WO 2022057556A1 CN 2021113187 W CN2021113187 W CN 2021113187W WO 2022057556 A1 WO2022057556 A1 WO 2022057556A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- tensor
- size
- disparity
- feature
- map
- Prior art date
Links
- 238000013135 deep learning Methods 0.000 title claims abstract description 17
- 238000000691 measurement method Methods 0.000 title claims abstract description 15
- 238000000034 method Methods 0.000 claims abstract description 12
- 230000002776 aggregation Effects 0.000 claims abstract description 9
- 238000004220 aggregation Methods 0.000 claims abstract description 9
- 238000001514 detection method Methods 0.000 claims abstract description 9
- 238000000605 extraction Methods 0.000 claims abstract description 9
- 238000001914 filtration Methods 0.000 claims abstract description 8
- 238000012545 processing Methods 0.000 claims abstract description 4
- 238000012937 correction Methods 0.000 claims abstract description 3
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 9
- 238000011176 pooling Methods 0.000 claims description 3
- 238000005259 measurement Methods 0.000 description 10
- 238000012876 topography Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000001314 profilometry Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/64—Three-dimensional objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
- G06T7/55—Depth or shape recovery from multiple images
- G06T7/593—Depth or shape recovery from multiple images from stereo images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
- G06T7/55—Depth or shape recovery from multiple images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4007—Scaling of whole images or parts thereof, e.g. expanding or contracting based on interpolation, e.g. bilinear interpolation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20228—Disparity calculation for image-based rendering
Definitions
- the invention belongs to the technical field of optical measurement, in particular to an end-to-end speckle projection three-dimensional measurement method based on deep learning.
- the purpose of the present invention is to propose an end-to-end speckle projection three-dimensional measurement method based on deep learning.
- the technical solution for realizing the purpose of the present invention is: an end-to-end speckle projection three-dimensional measurement method based on deep learning, comprising the following steps:
- Step 1 The projector projects and the binocular camera synchronously collects the speckle pattern, and performs stereo correction on the speckle pattern;
- Step 2 processing the speckle pattern based on the feature extraction sub-network of shared weights to obtain a low-resolution 3-dimensional feature tensor of a set size;
- Step 3 Input the feature tensor into the salient object detection sub-network to detect the foreground information in the speckle map, and generate a full-resolution effective mask map;
- Step 4 Generate a 4-dimensional matching cost body according to the feature tensors of the two perspectives and the candidate disparity range, input the 4-dimensional matching cost body into the three-dimensional convolution layer filtering to achieve cost aggregation, and obtain the initial disparity map through disparity regression;
- Step 5 Obtain the final disparity map according to the effective mask map and the initial disparity map.
- step 2 the specific process of processing the speckle pattern based on the shared weight feature extraction sub-network to obtain a low-resolution 3-dimensional feature tensor with a set size is as follows: the size is H ⁇ W, and the speckle pattern passes through 3 Convolutional layers with the same number of output channels get a tensor of size 32 ⁇ H ⁇ W;
- a tensor of size 128 ⁇ H/2 ⁇ W/2 is then passed through average pooling and convolution layers of size (5,5), (10,10), (20,20), (40,40), respectively.
- the layers perform downsampling at different scales, and use bilinear interpolation to obtain the original resolution tensor;
- the original resolution tensor is spliced with a tensor of size 64 ⁇ H/2 ⁇ W/2 and a tensor of size 128 ⁇ H/2 ⁇ W/2 on the feature channel, and the size is 320 ⁇ A tensor of H/2 ⁇ W/2;
- the feature tensor is input into the salient object detection sub-network to detect foreground information in the speckle map, and the specific process of generating a full-resolution effective mask map is: the size is 32 ⁇ H/2 ⁇
- the tensor of W/2 passes through 3 residual blocks to obtain a tensor of size 64 ⁇ H/2 ⁇ W/2; passes through a deconvolution layer to obtain a tensor of size 32 ⁇ H ⁇ W; passes through 3 residual blocks
- the difference block obtains a tensor with a size of 32 ⁇ H ⁇ W; a tensor with a size of 1 ⁇ H ⁇ W is obtained through a convolutional layer without activation operations; the final full-resolution effective mask image is obtained through a sigmoid layer .
- the 4-dimensional matching cost body generated according to the candidate disparity range using the feature tensors of the two perspectives is specifically:
- Cost(1:32,D i -D min +1,1:H,1:WD i ) Feature left (1:32,1:H,1:WD i )
- Cost(33:64,D i -D min +1,1:H,1:WD i ) Feature right (1:32,1:H,D i :W)
- Feature left and Feature right are the feature tensors of the two viewing angles
- 32 ⁇ H/2 ⁇ W/2 is the size of the feature tensors of the two viewing angles
- [D min , D max ] is the disparity range
- D i is Candidate disparity.
- the process of obtaining the initial disparity map through disparity regression is as follows:
- the matching cost body passes through the Softmax layer, and the initial disparity map is obtained through disparity regression, as shown in the following formula:
- [D min , D max ] is the disparity range
- Softmax( ) represents the Softmax operation
- Disparity represents the initial disparity map obtained by parallax regression
- Cost is the 4-dimensional matching cost body after cost filtering
- step 5 obtains the final disparity map according to the effective mask map and the initial disparity map, as shown in the following formula:
- Disparity final (x,y) Disparity(x,y)*Mask(x,y)
- Disparity is the initial disparity map
- Mask is the effective mask map
- the present invention has the significant advantages that the present invention only needs to project a speckle pattern to realize single-frame and robust absolute three-dimensional topography measurement.
- FIG. 1 is a schematic flowchart of an end-to-end speckle projection 3D measurement method based on deep learning.
- FIG. 2 is a basic schematic diagram of the deep learning-based stereo matching algorithm of the present invention.
- Figure 3 is a schematic representation of the results obtained using the present invention.
- An end-to-end speckle projection 3D measurement method based on deep learning firstly uses projector projection and binocular camera to synchronously collect speckle patterns.
- the speckle pattern is stereo corrected and fed into the stereo matching network.
- a feature extraction sub-network based on shared weights processes the speckle pattern to obtain a series of low-resolution 3D feature tensors.
- the feature tensor is input into the salient object detection sub-network to detect foreground information in the speckle map to generate a full-resolution effective mask map.
- the feature tensors of the two perspectives are used to generate a 4-dimensional matching cost body according to the candidate disparity range.
- the cost aggregation is realized, and the initial disparity map is obtained through disparity regression.
- the final disparity map is obtained by combining the effective mask map and the initial disparity map, enabling single-frame, robust absolute 3D topography measurement.
- the invention only needs to project a speckle pattern to realize single-frame and robust absolute three-dimensional topography measurement. The specific steps are:
- Step 1 First, use the projector to project and the binocular camera to collect the speckle pattern synchronously.
- the speckle pattern is stereo corrected and fed into the stereo matching network.
- Step 2 The feature extraction sub-network based on shared weights processes the speckle pattern to obtain a series of low-resolution 3-dimensional feature tensors.
- the size of the speckle pattern is H ⁇ W.
- the tensor of size 32 ⁇ H/2 ⁇ W/2 is obtained through 3 residual blocks in succession, and the tensor of size 64 ⁇ H/2 ⁇ W/2 is obtained after 16 residual blocks , and then 6 residual blocks are used to obtain a tensor of size 128 ⁇ H/2 ⁇ W/2.
- a tensor of size 128 ⁇ H/2 ⁇ W/2 is then passed through average pooling and convolution layers of size (5,5), (10,10), (20,20), (40,40), respectively.
- the layers downsample the tensors at different scales, and then use bilinear interpolation to get the original resolution tensors.
- These original resolution tensors are spliced with tensors of size 64 ⁇ H/2 ⁇ W/2 and tensors of size 128 ⁇ H/2 ⁇ W/2 on the feature channel, and the size is 320 A tensor of ⁇ H/2 ⁇ W/2.
- a tensor of size 32 ⁇ H/2 ⁇ W/2 is obtained through two convolutional layers.
- Step 3 The feature tensor is input into the saliency object detection sub-network to detect foreground information in the speckle map to generate a full-resolution effective mask map.
- a tensor with a size of 32 ⁇ H/2 ⁇ W/2 first passes through three residual blocks to obtain a tensor with a size of 64 ⁇ H/2 ⁇ W/2, and then passes through a deconvolution layer to obtain a size of A tensor of size 32 ⁇ H ⁇ W is obtained through 3 residual blocks to obtain a tensor of size 32 ⁇ H ⁇ W, and a tensor of size 1 ⁇ H ⁇ W is obtained through a convolutional layer without activation operations, and finally After a sigmoid layer, the final full-resolution effective mask image Mask is obtained, so as to realize the detection of foreground information in the speckle image.
- Step 4 Use the feature tensors of the two perspectives to generate a 4D matching cost body according to the candidate disparity range, achieve cost aggregation through multiple 3D convolution layer filtering, and obtain the initial disparity map through disparity regression.
- the size of the feature tensors of the two perspectives is 32 ⁇ H/2 ⁇ W/2
- the disparity range is [D min , D max ]
- the corresponding matching cost body Cost is calculated as follows:
- Cost(1:32,D i -D min +1,1:H,1:WD i ) Feature left (1:32,1:H,1:WD i )
- Cost(33:64,D i -D min +1,1:H,1:WD i ) Feature right (1:32,1:H,D i :W)
- the size of the 4D matching cost volume generated from the candidate disparity range is 64 ⁇ (D max - D min +1) ⁇ H/2 ⁇ W/2.
- the 4-dimensional matching cost body is filtered by multiple 3-D convolution layers to achieve cost aggregation, thereby obtaining a matching cost body of size 1 ⁇ (D max -D min +1) ⁇ H/2 ⁇ W/2.
- the initial disparity map is obtained through disparity regression, as shown in the following formula:
- Step 5 Combine the effective mask map Mask and the initial disparity map Disparity to obtain the final disparity map, so as to achieve single-frame, robust absolute 3D topography measurement.
- the final disparity map Disparity final is obtained by combining the effective mask map Mask and the initial disparity map Disparity, as shown in the following formula:
- Disparity final (x,y) Disparity(x,y)*Mask(x,y)
- the parallax data is converted into three-dimensional information, and finally a single-frame, robust absolute three-dimensional topography measurement is realized.
- the stereo matching network proposed by the present invention includes the following parts:
- the 4-dimensional matching cost after cost aggregation obtains the initial disparity map through disparity regression
- FIG. 2 is a basic schematic diagram of the deep learning-based stereo matching algorithm of the present invention. Using the steps 2 to 5, a single-frame, robust absolute 3D topography measurement is finally realized.
- Fig. 3 is a result diagram of the end-to-end speckle projection three-dimensional measurement method based on deep learning of the present invention. The result shown in FIG. 3 proves that the present invention only needs to project a speckle pattern to realize single-frame, robust absolute 3D topography measurement.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Image Analysis (AREA)
- Length Measuring Devices By Optical Means (AREA)
- Image Processing (AREA)
Abstract
Description
Claims (6)
- 一种基于深度学习的端到端散斑投影三维测量方法,其特征在于,包括如下步骤:步骤一:投影仪投影、双目相机同步采集散斑图案,对散斑图案进行立体校正;步骤二:基于共享权重的特征提取子网络对散斑图案进行处理得到设定尺寸的低分辨率的3维特征张量;步骤三:将特征张量输入到显着性物体检测子网络中检测散斑图中的前景信息,生成全分辨率的有效掩膜图;步骤四:根据两个视角的特征张量以及候选视差范围生成4维匹配成本体,将4维匹配成本体输入三维卷积层滤波实现成本聚合,通过视差回归得到初始视差图;步骤五:根据有效掩膜图和初始视差图得到最终视差图。
- 根据权利要求1所述的基于深度学习的端到端散斑投影三维测量方法,其特征在于,步骤二基于共享权重的特征提取子网络对散斑图案进行处理得到设定尺寸的低分辨率的3维特征张量的具体过程为:尺寸大小为H×W,散斑图案经过3个具有相同输出通道数的卷积层得到尺寸为32×H×W的张量;经过1个具有步长为2的卷积层得到尺寸为32×H/2×W/2的张量;连续经过3个残差块得到尺寸为32×H/2×W/2的张量;经过16个残差块得到尺寸为64×H/2×W/2的张量;经过6个残差块得到尺寸为128×H/2×W/2的张量;尺寸为128×H/2×W/2的张量然后分别经过大小为(5,5)、(10,10)、(20,20)、(40,40)的平均池化层与卷积层进行不同尺度的降采样,并使用双线性插值得到原分辨率的张量;并将原分辨率的张量与尺寸为64×H/2×W/2的张量、尺寸为128×H/2×W/2的张量在特征通道上进行拼接,得到尺寸为320×H/2×W/2的张量;经过两个卷积层得到尺寸为32×H/2×W/2的张量。
- 根据权利要求1所述的基于深度学习的端到端散斑投影三维测量方法,其特征在于,步骤三将特征张量输入到显着性物体检测子网络中检测散斑图中的前景信息,生成全分辨率的有效掩膜图的具体过程为:尺寸为32×H/2×W/2的 张量经过3个残差块得到尺寸为64×H/2×W/2的张量;经过一个反卷积层得到尺寸为32×H×W的张量;经过3个残差块得到尺寸为32×H×W的张量;通过一个不含激活操作的卷积层得到尺寸为1×H×W的张量;经过一个Sigmoid层得到最终全分辨率的有效掩膜图。
- 根据权利要求1所述的基于深度学习的端到端散斑投影三维测量方法,其特征在于,步骤四中使用两个视角的特征张量根据候选视差范围生成的4维匹配成本体具体为:Cost(1:32,D i-D min+1,1:H,1:W-D i)=Feature left(1:32,1:H,1:W-D i)Cost(33:64,D i-D min+1,1:H,1:W-D i)=Feature right(1:32,1:H,D i:W)其中,Feature left和Feature right为两个视角的特征张量,32×H/2×W/2为两个视角的特征张量的尺寸,[D min,D max]为视差范围,D i为候选视差。
- 根据权利要求1所述的基于深度学习的端到端散斑投影三维测量方法,其特征在于,步骤五根据有效掩膜图和初始视差图得到的最终视差图,如下式所示:Disparity final(x,y)=Disparity(x,y)*Mask(x,y)式中,Disparity为初始视差图,Mask为有效掩膜图。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US18/025,815 US20240020866A1 (en) | 2020-09-18 | 2021-08-18 | A three-dimensional measurement method based on end-to-end deep learning for speckle projection |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010982857.3A CN112419386B (zh) | 2020-09-18 | 2020-09-18 | 一种基于深度学习的端到端散斑投影三维测量方法 |
CN202010982857.3 | 2020-09-18 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2022057556A1 true WO2022057556A1 (zh) | 2022-03-24 |
Family
ID=74855304
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2021/113187 WO2022057556A1 (zh) | 2020-09-18 | 2021-08-18 | 一种基于深度学习的端到端散斑投影三维测量方法 |
Country Status (3)
Country | Link |
---|---|
US (1) | US20240020866A1 (zh) |
CN (1) | CN112419386B (zh) |
WO (1) | WO2022057556A1 (zh) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116704000A (zh) * | 2023-08-04 | 2023-09-05 | 东莞市爱培科技术有限公司 | 一种无监督学习遮挡区域的立体匹配方法 |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112419386B (zh) * | 2020-09-18 | 2022-10-14 | 南京理工大学 | 一种基于深度学习的端到端散斑投影三维测量方法 |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110111271A (zh) * | 2019-04-24 | 2019-08-09 | 北京理工大学 | 一种基于侧抑制网络的单像素成像方法 |
US20200193580A1 (en) * | 2018-12-14 | 2020-06-18 | Spectral Md, Inc. | System and method for high precision multi-aperture spectral imaging |
CN111563564A (zh) * | 2020-07-20 | 2020-08-21 | 南京理工大学智能计算成像研究院有限公司 | 基于深度学习的散斑图像逐像素匹配方法 |
CN112419386A (zh) * | 2020-09-18 | 2021-02-26 | 南京理工大学 | 一种基于深度学习的端到端散斑投影三维测量方法 |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101608908B (zh) * | 2009-07-20 | 2011-08-10 | 杭州先临三维科技股份有限公司 | 数字散斑投影和相位测量轮廓术相结合的三维数字成像方法 |
CN110487216B (zh) * | 2019-09-20 | 2021-05-25 | 西安知象光电科技有限公司 | 一种基于卷积神经网络的条纹投影三维扫描方法 |
CN111145342B (zh) * | 2019-12-27 | 2024-04-12 | 山东中科先进技术研究院有限公司 | 一种双目散斑结构光三维重建方法及系统 |
-
2020
- 2020-09-18 CN CN202010982857.3A patent/CN112419386B/zh active Active
-
2021
- 2021-08-18 US US18/025,815 patent/US20240020866A1/en active Pending
- 2021-08-18 WO PCT/CN2021/113187 patent/WO2022057556A1/zh active Application Filing
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20200193580A1 (en) * | 2018-12-14 | 2020-06-18 | Spectral Md, Inc. | System and method for high precision multi-aperture spectral imaging |
CN110111271A (zh) * | 2019-04-24 | 2019-08-09 | 北京理工大学 | 一种基于侧抑制网络的单像素成像方法 |
CN111563564A (zh) * | 2020-07-20 | 2020-08-21 | 南京理工大学智能计算成像研究院有限公司 | 基于深度学习的散斑图像逐像素匹配方法 |
CN112419386A (zh) * | 2020-09-18 | 2021-02-26 | 南京理工大学 | 一种基于深度学习的端到端散斑投影三维测量方法 |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116704000A (zh) * | 2023-08-04 | 2023-09-05 | 东莞市爱培科技术有限公司 | 一种无监督学习遮挡区域的立体匹配方法 |
CN116704000B (zh) * | 2023-08-04 | 2023-10-03 | 东莞市爱培科技术有限公司 | 一种无监督学习遮挡区域的立体匹配方法 |
Also Published As
Publication number | Publication date |
---|---|
CN112419386B (zh) | 2022-10-14 |
US20240020866A1 (en) | 2024-01-18 |
CN112419386A (zh) | 2021-02-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110288642B (zh) | 基于相机阵列的三维物体快速重建方法 | |
CN111563564B (zh) | 基于深度学习的散斑图像逐像素匹配方法 | |
CN109813251B (zh) | 用于三维测量的方法、装置以及系统 | |
CN111351450B (zh) | 基于深度学习的单帧条纹图像三维测量方法 | |
WO2020063013A1 (zh) | 一种基于深度学习的条纹投影时间相位展开方法 | |
WO2022057556A1 (zh) | 一种基于深度学习的端到端散斑投影三维测量方法 | |
CN109919911B (zh) | 基于多视角光度立体的移动三维重建方法 | |
KR101733228B1 (ko) | 구조광을 이용한 3차원 스캐닝 장치 | |
CN110956661B (zh) | 基于双向单应矩阵的可见光与红外相机动态位姿计算方法 | |
CN107990846B (zh) | 基于单帧结构光的主被动结合深度信息获取方法 | |
CN113129430B (zh) | 基于双目结构光的水下三维重建方法 | |
CN107564091A (zh) | 一种基于快速对应点搜索的三维重建方法及装置 | |
WO2024032233A1 (zh) | 基于双目视觉的立体摄影测量方法 | |
CN112001973B (zh) | 基于数字散斑相关的快速三维人头测量方法 | |
CN113393439A (zh) | 一种基于深度学习的锻件缺陷检测方法 | |
CN111696044B (zh) | 一种大场景动态视觉观测方法及装置 | |
CN111105451B (zh) | 一种克服遮挡效应的驾驶场景双目深度估计方法 | |
JP6285686B2 (ja) | 視差画像生成装置 | |
Furukawa et al. | Multiview projectors/cameras system for 3d reconstruction of dynamic scenes | |
JP2006337075A (ja) | 3次元情報復元装置 | |
JP7230722B2 (ja) | 画像処理装置及び画像処理方法 | |
CN116363226A (zh) | 一种实时的多相机多投影仪3d成像处理方法及装置 | |
CN114820563A (zh) | 一种基于多视点立体视觉的工业部件尺寸估计方法及系统 | |
CN111023999B (zh) | 一种基于空间编码结构光的稠密点云生成方法 | |
CN109389629B (zh) | 立体匹配自适应视差等级的确定方法 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 21868384 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 18025815 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 21868384 Country of ref document: EP Kind code of ref document: A1 |
|
32PN | Ep: public notification in the ep bulletin as address of the adressee cannot be established |
Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 05.10.2023) |