CN109461178A - A kind of monocular image depth estimation method and device merging sparse known label - Google Patents
A kind of monocular image depth estimation method and device merging sparse known label Download PDFInfo
- Publication number
- CN109461178A CN109461178A CN201811050407.XA CN201811050407A CN109461178A CN 109461178 A CN109461178 A CN 109461178A CN 201811050407 A CN201811050407 A CN 201811050407A CN 109461178 A CN109461178 A CN 109461178A
- Authority
- CN
- China
- Prior art keywords
- depth
- sparse
- rgb image
- estimated
- known label
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 19
- 239000013598 vector Substances 0.000 claims description 21
- 230000004927 fusion Effects 0.000 claims description 16
- 238000012549 training Methods 0.000 claims description 10
- 238000010276 construction Methods 0.000 claims description 9
- 238000005457 optimization Methods 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 239000000284 extract Substances 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 241000208340 Araliaceae Species 0.000 claims 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 claims 1
- 235000003140 Panax quinquefolius Nutrition 0.000 claims 1
- 235000008434 ginseng Nutrition 0.000 claims 1
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000003475 lamination Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- 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
- G06T7/521—Depth or shape recovery from laser ranging, e.g. using interferometry; from the projection of structured light
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- 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
- G06T2207/20081—Training; Learning
-
- 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]
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Optics & Photonics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention proposes a kind of monocular image depth estimation methods and device for merging sparse known label, comprising: obtains RGB image to be estimated;Sparse known label is obtained using single line laser radar;The RGB image to be estimated is input in the estimation of Depth model pre-established, the first depth map of RGB image to be estimated is obtained;First depth map is merged with sparse known label by full articulamentum, obtains the ultimate depth figure of the RGB image to be estimated.Technical solution provided by the invention reduces the uncertainty that depth map is mapped to from monocular image, to effectively estimate relatively reliable scene depth by merging sparse known label.
Description
Technical field
The present invention relates to field of image processings, and in particular to a kind of monocular image estimation of Depth for merging sparse known label
Method and device.
Background technique
Monocular image estimates that scene depth is the important method for understanding geometry in scene, moreover, research it is many its
When his computer vision problem, incorporating depth information can be improved the performance of algorithm, such as semantic segmentation, Attitude estimation, target
Detection.There are the depth transducer (Kinect of such as Microsoft) of available RGB-D depth image, but this kind of sensor at present
Being limited in scope and (being less than 4m) for perceived depth, can generate a large amount of noise under strong light, so have its limitation in various application scenarios
Property.
Existing monocular image estimation of Depth there are many problems, for example, a two dimensional image correspond to it is infinite a variety of true
3D scene, this allows for for single image being mapped to depth map and exists uncertain, and this uncertainty determines computer
Vision mode only can not go out accurate depth value by voucher width Image estimation in principle.
Therefore the present invention provides a kind of monocular image depth estimation method for merging sparse known label and device to solve
The deficiencies in the prior art.
Summary of the invention
The present invention is intended to provide a kind of monocular image depth estimation method and device for merging sparse known label, solves mesh
Preceding picture depth estimates inaccurate problem.
According to an aspect of the present invention, a kind of monocular image depth estimation method for merging sparse known label is provided,
Include:
Obtain RGB image to be estimated;
Sparse known label is obtained using single line laser radar;
The RGB image to be estimated is input in the estimation of Depth model pre-established, RGB image to be estimated is obtained
First depth map;
First depth map is merged with sparse known label by full articulamentum, obtains the RGB image to be estimated
Ultimate depth figure.
It is further, described to obtain sparse known label using single line laser radar, comprising:
The point that single line laser radar is scanned projects to two-dimentional phase plane, obtains sparse known label.
Further, the estimation of Depth model includes:
Obtain multiple RGB images;
The feature of the RGB image is extracted using the depth residual error network of full convolution;
Using full articulamentum by the Feature Conversion be feature vector;
Network parameter according to the depth residual error network of the loss function training full convolution, after being optimized;
Estimation of Depth model is constructed according to the network parameter after optimization.
Further, the depth residual error network according to the loss function training full convolution, the net after being optimized
Network parameter, comprising:
The loss function is shown below:
Wherein,It is respectively the real depth value and predetermined depth value of pixel i with yi;Xi is the real depth value of pixel i
With the difference of predetermined depth value;C is threshold value.
Further, described to be merged first depth map with sparse known label by full articulamentum, it obtains described
The ultimate depth figure of RGB image to be estimated, comprising:
According to the feature of first depth map and sparse known label feature, using full articulamentum obtain merging it is sparse
Know the feature vector of label;
The feature vector of the sparse known label of fusion is converted to the ultimate depth figure of RGB image to be estimated.
According to a further aspect of the invention, it discloses a kind of monocular image estimation of Depth dresses for merging sparse known label
It sets, comprising:
First obtains module, for obtaining RGB image to be estimated;
Second obtains module, for obtaining sparse known label using single line laser radar;
Processing module, for the RGB image to be estimated to be input in the estimation of Depth model pre-established, obtain to
Estimate the first depth map of RGB image;
Fusion Module obtains described for being merged first depth map with sparse known label by full articulamentum
The ultimate depth figure of RGB image to be estimated.
Further, described second module is obtained, is used for,
The point that single line laser radar is scanned projects to two-dimentional phase plane, obtains sparse known label.
Further, the processing module includes model construction submodule, and the model construction submodule includes:
Acquiring unit, for obtaining multiple RGB images;
Extraction unit extracts the feature of the RGB image for the depth residual error network using full convolution;
Converting unit is used to using full articulamentum be feature vector by the Feature Conversion;
Optimize unit, the net for the depth residual error network according to the loss function training full convolution, after being optimized
Network parameter;
Construction unit, for constructing estimation of Depth model according to the network parameter after optimization.
Further, the Fusion Module, is used for,
According to the feature of first depth map and sparse known label feature, using full articulamentum obtain merging it is sparse
Know the feature vector of label;
The feature vector of the sparse known label of fusion is converted to the ultimate depth figure of RGB image to be estimated.
The beneficial effect of the technical program is compared with the immediate prior art:
RGB image to be estimated is input in the estimation of Depth model pre-established by technical solution provided by the invention, is obtained
To the first depth map of RGB image to be estimated;Sparse known label is obtained by the first depth map and using single line laser radar again
Fusion, obtains the ultimate depth figure of the RGB image to be estimated.Pass through the side for merging the output of model with sparse known label
Formula reduces the uncertainty that depth map is mapped to from monocular image, to effectively estimate relatively reliable scene depth.
Detailed description of the invention
Fig. 1 is flow chart of the method for the present invention;
Fig. 2 is estimation of Depth model building method flow chart in the embodiment of the present application;
Fig. 3 is that the structure of Encoder-Decoder in the embodiment of the present application builds schematic diagram;
Fig. 4 is the partial structure diagram of Decoder in the embodiment of the present application.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiments of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill people
Member's every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
As shown in Figure 1, the present invention provides a kind of monocular image depth estimation method for merging sparse known label, process
It is as follows:
S101, RGB image to be estimated is obtained;
S102, sparse known label is obtained using single line laser radar;
S103, the RGB image to be estimated is input in the estimation of Depth model pre-established, obtains RGB to be estimated
First depth map of image;
S104, first depth map is merged by full articulamentum with sparse known label, obtains the RGB to be estimated
The ultimate depth figure of image.
In the embodiment of the present application, RGB image to be estimated is input in the estimation of Depth model pre-established first,
Obtain the first depth map of RGB image to be estimated;Then sparse known label is obtained using single line laser radar;By the first depth
Figure is merged with sparse known label, obtains the depth map with Pixel-level deep tag.By the side for merging sparse known label
Formula reduces the uncertainty that depth map is mapped to from monocular image, to effectively estimate relatively reliable scene depth.
Since single line two-dimensional laser radar has the characteristics that structure is simple, low-power consumption, low cost.Such as the production of Sick company
Product LMS111, the price of the sensor are about 5 the percent of the HDL-64 of velodyne company.Because of these characteristics, the sensing
Device is widely installed on some robots or pilotless automobile, for carrying out range measurement.Swashed using single line two dimension
These features of optical radar, it is described to obtain sparse known label using single line laser radar in some embodiments of the present application,
Include:
The point that single line laser radar is scanned projects to two-dimentional phase plane, obtains sparse known label.
It is, the point that single line laser radar is scanned projects to two-dimentional phase plane, at this point, these points are respectively positioned on one
On level of approximation straight line, therefore the pixel on straight line is provided with depth value, the pixel outside straight line is with null filling.Due to whole
Only having straight line in a phase plane has reliable depth value, therefore this prior information obtained by single line laser radar exists
It is referred to as sparse known label in the present invention.
In some embodiments of the present application, as shown in Fig. 2, estimation of Depth model includes:
S201, multiple RGB images are obtained;
S202, the feature that the RGB image is extracted using the depth residual error network of full convolution;
S203, using full articulamentum by the Feature Conversion be feature vector;
S204, the network parameter according to the depth residual error network of the loss function training full convolution, after being optimized;
S205, estimation of Depth model is constructed according to the network parameter after optimization.
Model in the present embodiment is built according to the structure of Encoder-Decoder, as shown in Figure 3.Encoder is adopted part
With the depth residual error network of 152 layers of full convolution, the high dimensional feature of low resolution is successively extracted from input picture.
Decoder is mainly made of warp lamination, successively by the output of Encoder output up-sampling to 5 scales, final scale output
Size be input picture size.
It is as follows in the optimization operation that the part Decoder carries out:
Firstly, by the characteristic pattern, extracted with the Encoder of this feature figure correspondingly-sized of upper warp lamination output
Characteristic pattern and the output of a upper scale these three high dimension vectors are spliced into a high dimension vector;
Then deconvolution operation is carried out to it, can not only merged the information of three's high dimension vector, but also can amplify
The size of this layer output;
Finally, being exported by the prediction that a convolutional layer obtains the scale, 5 in Decoder upper sampling process of model
There is prediction to export on scale, so supervised training loss layer, the portion of Decoder are added in the prediction output of each scale
Separation structure is as shown in Figure 4.
For model by berHu function as supervised training loss function, concrete form is as follows:
Wherein,It is respectively the real depth value and predetermined depth value of pixel i with yi;Xi is the real depth value of pixel i
With the difference of predetermined depth value;C is threshold value, concrete form are as follows:
In some embodiments of the present application, it is described by full articulamentum by first depth map and sparse known label
Fusion, obtains the ultimate depth figure of the RGB image to be estimated, comprising:
According to the feature of first depth map and sparse known label feature, using full articulamentum obtain merging it is sparse
Know the feature vector of label;
The feature vector of the sparse known label of fusion is converted to the ultimate depth figure of RGB image to be estimated.
In the present embodiment, sparse known label is originally considered as the characteristic pattern in a channel, without the pixel of known depth value
By null filling.Since convolution operation has with transformation property characteristic pattern, convolution kernel can be by the zero value pixels in sparse known label
It is equal with the pixel of known depth value and treats.
In order to achieve the purpose that the sparse known label of fusion, Decoder final output and sparse known label are spliced into
Then one tensor returns to obtain final depth map according to this tensor using full articulamentum.Because full articulamentum is not
Weight is shared, the weight on each side be all it is different, zero and nonzero value on neuron node can be distinguished in this way,
To acquire the absolute reference information that model needs from that a line depth value of offer.
Based on identical inventive concept, the present invention also provides a kind of monocular image depth for merging sparse known label to estimate
Counter device, comprising:
First obtains module, for obtaining RGB image to be estimated;
Second obtains module, for obtaining sparse known label using single line laser radar;
Processing module, for the RGB image to be estimated to be input in the estimation of Depth model pre-established, obtain to
Estimate the first depth map of RGB image;
Fusion Module obtains described for being merged first depth map with sparse known label by full articulamentum
The ultimate depth figure of RGB image to be estimated.
Optionally, described second module is obtained, is used for,
The point that single line laser radar is scanned projects to two-dimentional phase plane, obtains sparse known label.
Optionally, the processing module includes model construction submodule, and the model construction submodule includes:
Acquiring unit, for obtaining multiple RGB images;
Extraction unit extracts the feature of the RGB image for the depth residual error network using full convolution;
Converting unit is used to using full articulamentum be feature vector by the Feature Conversion;
Optimize unit, the net for the depth residual error network according to the loss function training full convolution, after being optimized
Network parameter;
Construction unit, for constructing estimation of Depth model according to the network parameter after optimization.
Optionally, the Fusion Module, is used for,
According to the feature of first depth map and sparse known label feature, using full articulamentum obtain merging it is sparse
Know the feature vector of label;
The feature vector of the sparse known label of fusion is converted to the ultimate depth figure of RGB image to be estimated.
Particular embodiments described above has carried out further in detail the purpose of the present invention, technical scheme and beneficial effects
Describe in detail bright, it should be understood that the above is only a specific embodiment of the present invention, is not intended to restrict the invention, it is all
Within the spirit and principles in the present invention, any modification, equivalent replacement, improvement and so on should be included in protection of the invention
Within the scope of.
Claims (9)
1. a kind of monocular image depth estimation method for merging sparse known label characterized by comprising
Obtain RGB image to be estimated;
Sparse known label is obtained using single line laser radar;
The RGB image to be estimated is input in the estimation of Depth model pre-established, the first of RGB image to be estimated is obtained
Depth map;
First depth map is merged with sparse known label by full articulamentum, obtains the RGB image to be estimated most
Whole depth map.
2. the method according to claim 1, wherein described obtain sparse known mark using single line laser radar
Label, comprising:
The point that single line laser radar is scanned projects to two-dimentional phase plane, obtains sparse known label.
3. the method according to claim 1, wherein the estimation of Depth model includes:
Obtain multiple RGB images;
The feature of the RGB image is extracted using the depth residual error network of full convolution;
Using full articulamentum by the Feature Conversion be feature vector;
Network parameter according to the depth residual error network of the loss function training full convolution, after being optimized;
Estimation of Depth model is constructed according to the network parameter after optimization.
4. according to the method described in claim 3, it is characterized in that, the depth according to the loss function training full convolution
Residual error network, the network parameter after being optimized, comprising:
The loss function is shown below:
Wherein,With yiRespectively the real depth value of pixel i and predetermined depth value;xiFor the real depth value and prediction of pixel i
The difference of depth value;C is threshold value.
5. the method according to claim 1, wherein it is described by full articulamentum by first depth map with it is dilute
Known label fusion is dredged, the ultimate depth figure of the RGB image to be estimated is obtained, comprising:
According to the feature of first depth map and sparse known label feature, obtain merging sparse known mark using full articulamentum
The feature vector of label;
The feature vector of the sparse known label of fusion is converted to the ultimate depth figure of RGB image to be estimated.
6. a kind of monocular image estimation of Depth device for merging sparse known label characterized by comprising
First obtains module, for obtaining RGB image to be estimated;
Second obtains module, for obtaining sparse known label using single line laser radar;
Processing module obtains to be estimated for the RGB image to be estimated to be input in the estimation of Depth model pre-established
First depth map of RGB image;
Fusion Module obtains described wait estimate for being merged first depth map with sparse known label by full articulamentum
Count the ultimate depth figure of RGB image.
7. device according to claim 6, which is characterized in that described second obtains module, is used for,
The point that single line laser radar is scanned projects to two-dimentional phase plane, obtains sparse known label.
8. device according to claim 6, which is characterized in that the processing module includes model construction submodule, described
Model construction submodule includes:
Acquiring unit, for obtaining multiple RGB images;
Extraction unit extracts the feature of the RGB image for the depth residual error network using full convolution;
Converting unit is used to using full articulamentum be feature vector by the Feature Conversion;
Optimize unit, the network ginseng for the depth residual error network according to the loss function training full convolution, after being optimized
Number;
Construction unit, for constructing estimation of Depth model according to the network parameter after optimization.
9. device according to claim 6, which is characterized in that the Fusion Module is used for,
According to the feature of first depth map and sparse known label feature, obtain merging sparse known mark using full articulamentum
The feature vector of label;
The feature vector of the sparse known label of fusion is converted to the ultimate depth figure of RGB image to be estimated.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811050407.XA CN109461178A (en) | 2018-09-10 | 2018-09-10 | A kind of monocular image depth estimation method and device merging sparse known label |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811050407.XA CN109461178A (en) | 2018-09-10 | 2018-09-10 | A kind of monocular image depth estimation method and device merging sparse known label |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109461178A true CN109461178A (en) | 2019-03-12 |
Family
ID=65606646
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811050407.XA Pending CN109461178A (en) | 2018-09-10 | 2018-09-10 | A kind of monocular image depth estimation method and device merging sparse known label |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109461178A (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110414418A (en) * | 2019-07-25 | 2019-11-05 | 电子科技大学 | A kind of Approach for road detection of image-lidar image data Multiscale Fusion |
CN111179331A (en) * | 2019-12-31 | 2020-05-19 | 智车优行科技(上海)有限公司 | Depth estimation method, depth estimation device, electronic equipment and computer-readable storage medium |
CN111340864A (en) * | 2020-02-26 | 2020-06-26 | 浙江大华技术股份有限公司 | Monocular estimation-based three-dimensional scene fusion method and device |
CN111476190A (en) * | 2020-04-14 | 2020-07-31 | 上海眼控科技股份有限公司 | Target detection method, apparatus and storage medium for unmanned driving |
CN111583663A (en) * | 2020-04-26 | 2020-08-25 | 宁波吉利汽车研究开发有限公司 | Monocular perception correction method and device based on sparse point cloud and storage medium |
CN112712017A (en) * | 2020-12-29 | 2021-04-27 | 上海智蕙林医疗科技有限公司 | Robot, monocular depth estimation method and system and storage medium |
CN113269118A (en) * | 2021-06-07 | 2021-08-17 | 重庆大学 | Monocular vision forward vehicle distance detection method based on depth estimation |
CN114627351A (en) * | 2022-02-18 | 2022-06-14 | 电子科技大学 | Fusion depth estimation method based on vision and millimeter wave radar |
CN114782782A (en) * | 2022-06-20 | 2022-07-22 | 武汉大学 | Uncertainty quantification method for learning performance of monocular depth estimation model |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106780588A (en) * | 2016-12-09 | 2017-05-31 | 浙江大学 | A kind of image depth estimation method based on sparse laser observations |
CN107204010A (en) * | 2017-04-28 | 2017-09-26 | 中国科学院计算技术研究所 | A kind of monocular image depth estimation method and system |
CN107578436A (en) * | 2017-08-02 | 2018-01-12 | 南京邮电大学 | A kind of monocular image depth estimation method based on full convolutional neural networks FCN |
-
2018
- 2018-09-10 CN CN201811050407.XA patent/CN109461178A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106780588A (en) * | 2016-12-09 | 2017-05-31 | 浙江大学 | A kind of image depth estimation method based on sparse laser observations |
CN107204010A (en) * | 2017-04-28 | 2017-09-26 | 中国科学院计算技术研究所 | A kind of monocular image depth estimation method and system |
CN107578436A (en) * | 2017-08-02 | 2018-01-12 | 南京邮电大学 | A kind of monocular image depth estimation method based on full convolutional neural networks FCN |
Non-Patent Citations (1)
Title |
---|
李耀宇: "《基于深度学习的单目图像深度估计》", 15 March 2018 * |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110414418B (en) * | 2019-07-25 | 2022-06-03 | 电子科技大学 | Road detection method for multi-scale fusion of image-laser radar image data |
CN110414418A (en) * | 2019-07-25 | 2019-11-05 | 电子科技大学 | A kind of Approach for road detection of image-lidar image data Multiscale Fusion |
CN111179331A (en) * | 2019-12-31 | 2020-05-19 | 智车优行科技(上海)有限公司 | Depth estimation method, depth estimation device, electronic equipment and computer-readable storage medium |
CN111179331B (en) * | 2019-12-31 | 2023-09-08 | 智车优行科技(上海)有限公司 | Depth estimation method, depth estimation device, electronic equipment and computer readable storage medium |
CN111340864A (en) * | 2020-02-26 | 2020-06-26 | 浙江大华技术股份有限公司 | Monocular estimation-based three-dimensional scene fusion method and device |
CN111340864B (en) * | 2020-02-26 | 2023-12-12 | 浙江大华技术股份有限公司 | Three-dimensional scene fusion method and device based on monocular estimation |
CN111476190A (en) * | 2020-04-14 | 2020-07-31 | 上海眼控科技股份有限公司 | Target detection method, apparatus and storage medium for unmanned driving |
CN111583663B (en) * | 2020-04-26 | 2022-07-12 | 宁波吉利汽车研究开发有限公司 | Monocular perception correction method and device based on sparse point cloud and storage medium |
CN111583663A (en) * | 2020-04-26 | 2020-08-25 | 宁波吉利汽车研究开发有限公司 | Monocular perception correction method and device based on sparse point cloud and storage medium |
CN112712017A (en) * | 2020-12-29 | 2021-04-27 | 上海智蕙林医疗科技有限公司 | Robot, monocular depth estimation method and system and storage medium |
CN113269118A (en) * | 2021-06-07 | 2021-08-17 | 重庆大学 | Monocular vision forward vehicle distance detection method based on depth estimation |
CN114627351A (en) * | 2022-02-18 | 2022-06-14 | 电子科技大学 | Fusion depth estimation method based on vision and millimeter wave radar |
CN114782782A (en) * | 2022-06-20 | 2022-07-22 | 武汉大学 | Uncertainty quantification method for learning performance of monocular depth estimation model |
CN114782782B (en) * | 2022-06-20 | 2022-10-04 | 武汉大学 | Uncertainty quantification method for learning performance of monocular depth estimation model |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109461178A (en) | A kind of monocular image depth estimation method and device merging sparse known label | |
US11361456B2 (en) | Systems and methods for depth estimation via affinity learned with convolutional spatial propagation networks | |
KR102126724B1 (en) | Method and apparatus for restoring point cloud data | |
CN113902897B (en) | Training of target detection model, target detection method, device, equipment and medium | |
CN110910437B (en) | Depth prediction method for complex indoor scene | |
CN110689562A (en) | Trajectory loop detection optimization method based on generation of countermeasure network | |
CN111507222B (en) | Three-dimensional object detection frame based on multisource data knowledge migration | |
CN112435267B (en) | Disparity map calculation method for high-resolution urban satellite stereo image | |
CN111160293A (en) | Small target ship detection method and system based on characteristic pyramid network | |
WO2023155387A1 (en) | Multi-sensor target detection method and apparatus, electronic device and storage medium | |
CN110349186A (en) | Optical flow computation method is moved based on the matched big displacement of depth | |
CN116385660A (en) | Indoor single view scene semantic reconstruction method and system | |
Nguyen et al. | ROI-based LiDAR sampling algorithm in on-road environment for autonomous driving | |
CN114299242A (en) | Method, device and equipment for processing images in high-precision map and storage medium | |
CN114358133A (en) | Method for detecting looped frames based on semantic-assisted binocular vision SLAM | |
Wang et al. | Pedestrian detection based on YOLOv3 multimodal data fusion | |
Le Besnerais et al. | Dense height map estimation from oblique aerial image sequences | |
CN117808689A (en) | Depth complement method based on fusion of millimeter wave radar and camera | |
CN116912645A (en) | Three-dimensional target detection method and device integrating texture and geometric features | |
CN112561979A (en) | Self-supervision monocular depth estimation method based on deep learning | |
Hirata et al. | Real-time dense depth estimation using semantically-guided LIDAR data propagation and motion stereo | |
CN116778266A (en) | Multi-scale neighborhood diffusion remote sensing point cloud projection image processing method | |
CN116503686A (en) | Training method of image correction model, image correction method, device and medium | |
CN115731273A (en) | Pose graph optimization method and device, electronic equipment and storage medium | |
CN116129422A (en) | Monocular 3D target detection method, monocular 3D target detection device, electronic equipment and storage medium |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190312 |
|
WD01 | Invention patent application deemed withdrawn after publication |