CN106780588A - A kind of image depth estimation method based on sparse laser observations - Google Patents
A kind of image depth estimation method based on sparse laser observations Download PDFInfo
- Publication number
- CN106780588A CN106780588A CN201611126056.7A CN201611126056A CN106780588A CN 106780588 A CN106780588 A CN 106780588A CN 201611126056 A CN201611126056 A CN 201611126056A CN 106780588 A CN106780588 A CN 106780588A
- Authority
- CN
- China
- Prior art keywords
- depth
- laser
- sparse
- estimation
- image
- 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
Classifications
-
- 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
- G06T2207/10004—Still image; Photographic image
-
- 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]
-
- 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
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
Landscapes
- Length Measuring Devices By Optical Means (AREA)
Abstract
This method discloses a kind of image depth estimation method based on sparse laser observations, and the method proposes to realize the dense reconstruct of depth based on monocular image using the sparse observation of single line or many line lasers.Deep neural network is trained by way of constructing reference depth figure and residual error depth map, sparse part Observational depth information is made full use of.Method of the present invention compared to estimation of Depth is only carried out with monocular image, the method embodies clear superiority.
Description
Technical field
It is thick the present invention relates to scene depth estimation field, more particularly to a kind of scene based on monocular image and sparse laser
Close depth estimation method.
Background technology
The mankind are based on rich experience and constantly learn, and also have the energy for estimating that objects in images is far and near from monocular image
Power, i.e., estimation of Depth ability to a certain extent.In recent years, machine learning method is also imitating this estimation of Depth ability of the mankind
On make remarkable progress, wherein being had outstanding performance with the depth learning technology of data-driven especially.Such a technique avoids manual spy
Design process is levied, based on original monocular RGB image learning characteristic, and the prediction for image correspondence depth is exported.
Eigen et al. proposes monocular depth based on deep learning and estimates first, they construct one it is two stage
Estimation of Depth network, first stage generation rough estimate simultaneously carries out fine adjustment in second stage.Then, they are by the work
Expand to and estimate scene depth, depth normal vector and Scene Semantics simultaneously, and demonstrate estimating depth normal vector simultaneously and
Semantic information contributes to scene depth to estimate performance boost.Liu et al. has inquired into the depth with condition random field with reference to deep learning
Degree estimation, super-pixel segmentation is carried out to image, and all super-pixel structural environment random fields are optimized.Li and Wang distinguishes
Extended thereon, successively optimized from super-pixel aspect to pixel aspect by the condition random field being layered.
Although these method validations from the possibility of monocular image estimating depth, actually monocular image be in itself yardstick letter
Breath missing.Eigen et al. is also mentioned, and the estimation of Depth based on monocular image there may be a deviation for the overall situation.
The content of the invention
Image dense depth is estimated it is an object of the invention to combine sparse single line laser information, to reduce scene depth
Estimate global deviation, obtain confidence level scene depth higher and estimate.
To achieve the above object, the present invention is based on deep learning method, is input with monocular image and sparse single line laser,
Autonomous learning feature simultaneously obtains dense depth estimation, and training process is comprised the following steps that：
A kind of depth image estimating method based on sparse laser observations, it is characterised in that it comprises the following steps：
Step one, is that, by sparse single line laser information denseization, the sparse laser includes single line laser and many line lasers,
Wherein with the single line laser construction reference depth figure and residual error depth map in sparse laser, in three dimensions in single line laser
Each laser spots stretched with direction perpendicular to the ground, obtain a reference depth face perpendicular to the ground；According to monocular
The calibration information of camera and single line laser, the reference depth plane projection that will be obtained in three dimensions to monocular camera obtains image
Image plane on, obtain a reference depth figure corresponding with described image, the absolute depth that will be obtained by depth transducer
Figure makes the difference with reference depth figure, obtains residual error depth map；
Step 2, the monocular image that monocular camera is obtained and the reference depth figure that obtains as described in step one are used as instruction
Practice data, training volume machine neural network estimates corresponding residual error depth map；
Step 3, the residual error depth map that will roll up machine neural network estimation is added with reference depth figure, and that is estimated is absolute
Depth map, referred to as absolute depth estimate figure, and the volume machine neural network of further constitution optimization on this basis, reduce this absolute
Difference between the absolute depth figure that estimation of Depth figure and depth transducer are obtained；The volume machine neural network and step 2 of the optimization
It is described to carry out end-to-end optimization for estimating that the volume machine neural network of residual error depth be superimposed, that is, it is input into monocular figure
Picture and reference depth figure, output obtain the absolute depth estimation figure by optimizing.
On the basis of above-mentioned technical proposal, the present invention can also be using further technical scheme once：
The absolute depth estimation figure that the end-to-end output of deep neural network is obtained is passed through into condition with sparse laser depth map
Random field is merged, so as to confirm that its depth value of position for having single line laser to observe in absolute depth estimation figure is seen with laser
The depth value of survey is consistent.
In step 2, training volume machine neural network estimates that corresponding residual error depth map mode is as follows：By depth to be fitted
The value discretization of the residual error depth of each pixel on residual plot to several natural numbers numerically, to classify, realize to residual by form
The estimation of Depth of difference depth.
Due to using technical scheme, beneficial effects of the present invention to be：It is of the invention can sparse true of bound fraction
The observation of real depth, such as single line laser radar, and more accurately estimation of Depth is obtained, the present invention can reduce scene depth estimation
Global deviation, obtains confidence level scene depth higher and estimates.
Brief description of the drawings
Fig. 1 a are input monocular image；
Fig. 1 b are the depth image example of expectation estimation；
Fig. 2 a are sparse laser observations；
Fig. 2 b be reference depth figure with；
Fig. 2 c residual error depth illustrated examples；
Fig. 3 a are that depth image is true；
Fig. 3 b are estimation of Depth before optimization；
Fig. 3 c are estimation of Depth after optimization.
Specific embodiment
In order to be better understood from technical scheme, it is further described below in conjunction with accompanying drawing.Fig. 1 illustrates depth
The example of estimation, is input into the monocular image shown in Fig. 1 a, it is desirable to estimate the scene depth shown in Fig. 1 b.
Step one, based on single line laser construction reference depth figure and residual error depth map.Fig. 2 a illustrate known in Fig. 1
Single line laser information, it is seen that single line laser information is very sparse and limited.Be by sparse single line laser information denseization,
Each laser spots is stretched with direction perpendicular to the ground in three dimensions, obtains a reference depth perpendicular to the ground
Face.According to monocular camera and the calibration information of single line laser, the reference depth plane correspondence that will be obtained in three-dimensional is plotted to image
On, a dense reference depth figure corresponding with image is obtained, as shown in Figure 2 b.Real depth map is done with reference depth figure
Difference, obtains residual error depth map, as shown in Figure 2 c.
Step 2, is input, regression criterion depth map with monocular image and reference depth figure based on deep learning.Will be every
To in several integer values, to classify, form realizes estimation of Depth to the residual error depth value discretization of individual pixel.Construct full convolution form
Deep neural network, realize that depth value classification in each pixel is estimated.To obtain preferably fitting performance and bigger appearance
Amount, the 50 layers of Deep Residual Network proposed using He et al., and the network for obtaining is trained on ImageNet with it
It is trained as initial value.
Step 3, network-evaluated residual error depth map is added with reference depth figure, the real depth map estimated, and
Further constitution optimization network, the difference reduced between the real depth map of the estimation and actual real depth map on this basis
It is different.The optimization network can estimate that network is superimposed with residual error, carry out end-to-end optimization.Fig. 2 illustrates depth true value, excellent
The comparing of estimation of Depth after changing preceding estimation of Depth and optimizing.
Validity by carrying out methods of experiments on NYUD2 data sets of the invention.NYUD2 is an interior RGB-
D data sets, this method simulates generation single line laser data in RGB-D data.Main advantage of the invention with reference depth
Figure and the map generalization of residual error depth and estimation.Therefore, experiment is compared under identical neural network structure, only with RGB as defeated
Enter predetermined depth and estimate real depth (scheme one), real depth (scheme is estimated as input using RGB and reference depth figure
Two), and using RGB and reference depth figure estimate that residual error depth map further obtains the knot of real depth map (scheme three)
Really, and the depth estimation method based on monocular image leading with our times is compared, the specific comparative result such as institute of table 1
Show.
To assess estimation of Depth effect in all directions, table 1 uses 6 Measure Indexes.Make the estimation of Depth value of each pixel
ForReal depth value y, T are the intersection of all pixels point, and 6 Measure Indexes are as follows respectively：
1. absolute relative error (rel),
2. average Log errors (log10),
3. square mean error (rms),
4. three threshold value accuracy rate (δi), meet condition'sIt is shared
The ratio of all pixels point.
From the result of table 1, the performance of estimation of Depth can necessarily be carried directly as input after laser denseization
Rise, and estimation of Depth performance then can be further lifted by way of residual error is estimated plus follow-up optimization.Compared to remaining generation
The leading monocular image depth estimation algorithm in boundary, this method all has a clear superiority on indices.
The NYUD2 data sets estimation of Depth of table 1 is contrasted.
Above-described embodiment is the description of the invention, is not limitation of the invention, it is any to simple transformation of the present invention after
Scheme belong to protection scope of the present invention.
Claims (3)
1. a kind of depth image estimating method based on sparse laser observations, it is characterised in that it comprises the following steps：
Step one, is that, by sparse single line laser information denseization, the sparse laser includes single line laser and many line lasers, wherein
With single line laser construction reference depth figure and residual error depth map in sparse laser, in three dimensions to single line laser in it is every
Individual laser spots are stretched with direction perpendicular to the ground, obtain a reference depth face perpendicular to the ground；According to monocular camera
With the calibration information of single line laser, the reference depth face that will be obtained in three dimensions project to monocular camera obtain image picture put down
On face, a reference depth figure corresponding with described image, the absolute depth figure that will be obtained by depth transducer and ginseng are obtained
Examine depth map to make the difference, obtain residual error depth map；
Step 2, the monocular image that monocular camera is obtained and the reference depth figure that obtains as described in step one are used as training number
According to training volume machine neural network estimates corresponding residual error depth map；
Step 3, the residual error depth map that will roll up machine neural network estimation is added with reference depth figure, the absolute depth estimated
Figure, referred to as absolute depth estimate figure, and the volume machine neural network of further constitution optimization on this basis,；The volume machine of the optimization
Neutral net can be superimposed with described in step 2 for estimating the volume machine neural network of residual error depth, carry out end-to-end excellent
Change, that is, be input into monocular image and reference depth figure, output obtains the absolute depth estimation figure by optimizing.
2. as described in claim 1 a kind of image depth estimation method based on sparse laser observations, it is characterised in that will
The end-to-end absolute depth estimation figure for obtaining that exports of deep neural network is carried out with sparse laser depth map by condition random field
Fusion, so as to confirm the depth value of its depth value of the position and laser observations for thering is single line laser to observe in absolute depth estimation figure
It is consistent.
3. as described in claim 1 a kind of image depth estimation method based on sparse laser observations, it is characterised in that step
In rapid two, training volume machine neural network estimates that corresponding residual error depth map mode is as follows：By on depth residual plot to be fitted
The value discretization of the residual error depth of each pixel numerically, the depth to residual error depth is realized in form of classifying to several natural numbers
Degree is estimated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611126056.7A CN106780588A (en) | 2016-12-09 | 2016-12-09 | A kind of image depth estimation method based on sparse laser observations |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611126056.7A CN106780588A (en) | 2016-12-09 | 2016-12-09 | A kind of image depth estimation method based on sparse laser observations |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106780588A true CN106780588A (en) | 2017-05-31 |
Family
ID=58877585
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611126056.7A Pending CN106780588A (en) | 2016-12-09 | 2016-12-09 | A kind of image depth estimation method based on sparse laser observations |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106780588A (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107992848A (en) * | 2017-12-19 | 2018-05-04 | 北京小米移动软件有限公司 | Obtain the method, apparatus and computer-readable recording medium of depth image |
| CN108416840A (en) * | 2018-03-14 | 2018-08-17 | 大连理工大学 | A kind of dense method for reconstructing of three-dimensional scenic based on monocular camera |
| CN108489496A (en) * | 2018-04-28 | 2018-09-04 | 北京空间飞行器总体设计部 | Noncooperative target Relative Navigation method for estimating based on Multi-source Information Fusion and system |
| CN108510535A (en) * | 2018-03-14 | 2018-09-07 | 大连理工大学 | A kind of high quality depth estimation method based on depth prediction and enhancing sub-network |
| CN109035319A (en) * | 2018-07-27 | 2018-12-18 | 深圳市商汤科技有限公司 | Monocular image depth estimation method and device, equipment, program and storage medium |
| CN109087349A (en) * | 2018-07-18 | 2018-12-25 | 亮风台（上海）信息科技有限公司 | A kind of monocular depth estimation method, device, terminal and storage medium |
| CN109146944A (en) * | 2018-10-30 | 2019-01-04 | 浙江科技学院 | A kind of space or depth perception estimation method based on the revoluble long-pending neural network of depth |
| CN109300151A (en) * | 2018-07-02 | 2019-02-01 | 浙江商汤科技开发有限公司 | Image processing method and device, electronic equipment |
| CN109325972A (en) * | 2018-07-25 | 2019-02-12 | 深圳市商汤科技有限公司 | Processing method, device, equipment and the medium of laser radar sparse depth figure |
| CN109461178A (en) * | 2018-09-10 | 2019-03-12 | 中国科学院自动化研究所 | A kind of monocular image depth estimation method and device merging sparse known label |
| CN110232361A (en) * | 2019-06-18 | 2019-09-13 | 中国科学院合肥物质科学研究院 | Human body behavior intension recognizing method and system based on the dense network of three-dimensional residual error |
| CN110428462A (en) * | 2019-07-17 | 2019-11-08 | 清华大学 | Polyphaser solid matching method and device |
| CN110992271A (en) * | 2020-03-04 | 2020-04-10 | 腾讯科技（深圳）有限公司 | Image processing method, path planning method, device, equipment and storage medium |
| CN111062981A (en) * | 2019-12-13 | 2020-04-24 | 腾讯科技（深圳）有限公司 | Image processing method, device and storage medium |
| CN111680554A (en) * | 2020-04-29 | 2020-09-18 | 北京三快在线科技有限公司 | Depth estimation method and device for automatic driving scene and autonomous vehicle |
| US10810754B2 (en) | 2018-04-24 | 2020-10-20 | Ford Global Technologies, Llc | Simultaneous localization and mapping constraints in generative adversarial networks for monocular depth estimation |
| 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 |
| CN113034562A (en) * | 2019-12-09 | 2021-06-25 | 百度在线网络技术（北京）有限公司 | Method and apparatus for optimizing depth information |
| CN113219475A (en) * | 2021-07-06 | 2021-08-06 | 北京理工大学 | Method and system for correcting monocular distance measurement by using single line laser radar |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103247075A (en) * | 2013-05-13 | 2013-08-14 | 北京工业大学 | Variational mechanism-based indoor scene three-dimensional reconstruction method |
| CN104346608A (en) * | 2013-07-26 | 2015-02-11 | 株式会社理光 | Sparse depth map densing method and device |
| CN106157307A (en) * | 2016-06-27 | 2016-11-23 | 浙江工商大学 | A kind of monocular image depth estimation method based on multiple dimensioned CNN and continuous CRF |
-
2016
- 2016-12-09 CN CN201611126056.7A patent/CN106780588A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103247075A (en) * | 2013-05-13 | 2013-08-14 | 北京工业大学 | Variational mechanism-based indoor scene three-dimensional reconstruction method |
| CN104346608A (en) * | 2013-07-26 | 2015-02-11 | 株式会社理光 | Sparse depth map densing method and device |
| CN106157307A (en) * | 2016-06-27 | 2016-11-23 | 浙江工商大学 | A kind of monocular image depth estimation method based on multiple dimensioned CNN and continuous CRF |
Non-Patent Citations (2)
| Title |
|---|
| FAYAO LIU 等: "Learning Depth from Single Monocular Images Using Deep Convolutional Neural Fields", 《IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE》 * |
| YIYI LIAO 等: "Parse Geometry from a Line: Monocular Depth Estimation with Partial Laser Observation", 《ARXIV:1611.02174V1 [CS.CV]》 * |
Cited By (39)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107992848B (en) * | 2017-12-19 | 2020-09-25 | 北京小米移动软件有限公司 | Method and device for acquiring depth image and computer readable storage medium |
| CN107992848A (en) * | 2017-12-19 | 2018-05-04 | 北京小米移动软件有限公司 | Obtain the method, apparatus and computer-readable recording medium of depth image |
| CN108416840B (en) * | 2018-03-14 | 2020-02-18 | 大连理工大学 | Three-dimensional scene dense reconstruction method based on monocular camera |
| CN108416840A (en) * | 2018-03-14 | 2018-08-17 | 大连理工大学 | A kind of dense method for reconstructing of three-dimensional scenic based on monocular camera |
| CN108510535A (en) * | 2018-03-14 | 2018-09-07 | 大连理工大学 | A kind of high quality depth estimation method based on depth prediction and enhancing sub-network |
| CN108510535B (en) * | 2018-03-14 | 2020-04-24 | 大连理工大学 | High-quality depth estimation method based on depth prediction and enhancer network |
| US10810754B2 (en) | 2018-04-24 | 2020-10-20 | Ford Global Technologies, Llc | Simultaneous localization and mapping constraints in generative adversarial networks for monocular depth estimation |
| CN108489496A (en) * | 2018-04-28 | 2018-09-04 | 北京空间飞行器总体设计部 | Noncooperative target Relative Navigation method for estimating based on Multi-source Information Fusion and system |
| CN109300151A (en) * | 2018-07-02 | 2019-02-01 | 浙江商汤科技开发有限公司 | Image processing method and device, electronic equipment |
| CN109300151B (en) * | 2018-07-02 | 2021-02-12 | 浙江商汤科技开发有限公司 | Image processing method and device and electronic equipment |
| CN109087349A (en) * | 2018-07-18 | 2018-12-25 | 亮风台（上海）信息科技有限公司 | A kind of monocular depth estimation method, device, terminal and storage medium |
| CN109087349B (en) * | 2018-07-18 | 2021-01-26 | 亮风台（上海）信息科技有限公司 | Monocular depth estimation method, device, terminal and storage medium |
| CN109325972A (en) * | 2018-07-25 | 2019-02-12 | 深圳市商汤科技有限公司 | Processing method, device, equipment and the medium of laser radar sparse depth figure |
| CN109325972B (en) * | 2018-07-25 | 2020-10-27 | 深圳市商汤科技有限公司 | Laser radar sparse depth map processing method, device, equipment and medium |
| KR102292559B1 (en) * | 2018-07-27 | 2021-08-24 | 선전 센스타임 테크놀로지 컴퍼니 리미티드 | Monocular image depth estimation method and apparatus, apparatus, program and storage medium |
| WO2020019761A1 (en) * | 2018-07-27 | 2020-01-30 | 深圳市商汤科技有限公司 | Monocular image depth estimation method and apparatus, device, program and storage medium |
| TWI766175B (en) * | 2018-07-27 | 2022-06-01 | 大陸商深圳市商湯科技有限公司 | Method, device and apparatus for monocular image depth estimation, program and storage medium thereof |
| CN109035319B (en) * | 2018-07-27 | 2021-04-30 | 深圳市商汤科技有限公司 | Monocular image depth estimation method, monocular image depth estimation device, monocular image depth estimation apparatus, monocular image depth estimation program, and storage medium |
| KR20200044108A (en) * | 2018-07-27 | 2020-04-28 | 선전 센스타임 테크놀로지 컴퍼니 리미티드 | Method and apparatus for estimating monocular image depth, device, program and storage medium |
| US11443445B2 (en) | 2018-07-27 | 2022-09-13 | Shenzhen Sensetime Technology Co., Ltd. | Method and apparatus for depth estimation of monocular image, and storage medium |
| JP2021500689A (en) * | 2018-07-27 | 2021-01-07 | 深▲せん▼市商▲湯▼科技有限公司Ｓｈｅｎｚｈｅｎ Ｓｅｎｓｅｔｉｍｅ Ｔｅｃｈｎｏｌｏｇｙ Ｃｏ．， Ｌｔｄ． | Monocular image depth estimation method and equipment, equipment, programs and storage media |
| CN109035319A (en) * | 2018-07-27 | 2018-12-18 | 深圳市商汤科技有限公司 | Monocular image depth estimation method and device, equipment, program and storage medium |
| CN109461178A (en) * | 2018-09-10 | 2019-03-12 | 中国科学院自动化研究所 | A kind of monocular image depth estimation method and device merging sparse known label |
| CN109146944B (en) * | 2018-10-30 | 2020-06-26 | 浙江科技学院 | Visual depth estimation method based on depth separable convolutional neural network |
| CN109146944A (en) * | 2018-10-30 | 2019-01-04 | 浙江科技学院 | A kind of space or depth perception estimation method based on the revoluble long-pending neural network of depth |
| CN110232361A (en) * | 2019-06-18 | 2019-09-13 | 中国科学院合肥物质科学研究院 | Human body behavior intension recognizing method and system based on the dense network of three-dimensional residual error |
| CN110232361B (en) * | 2019-06-18 | 2021-04-02 | 中国科学院合肥物质科学研究院 | Human behavior intention identification method and system based on three-dimensional residual dense network |
| CN110428462A (en) * | 2019-07-17 | 2019-11-08 | 清华大学 | Polyphaser solid matching method and device |
| CN110428462B (en) * | 2019-07-17 | 2022-04-08 | 清华大学 | Multi-camera stereo matching method and device |
| CN113034562A (en) * | 2019-12-09 | 2021-06-25 | 百度在线网络技术（北京）有限公司 | Method and apparatus for optimizing depth information |
| CN113034562B (en) * | 2019-12-09 | 2023-05-12 | 百度在线网络技术（北京）有限公司 | Method and apparatus for optimizing depth information |
| CN111062981A (en) * | 2019-12-13 | 2020-04-24 | 腾讯科技（深圳）有限公司 | Image processing method, device and storage medium |
| CN111062981B (en) * | 2019-12-13 | 2023-05-05 | 腾讯科技（深圳）有限公司 | Image processing method, device and storage medium |
| CN110992271A (en) * | 2020-03-04 | 2020-04-10 | 腾讯科技（深圳）有限公司 | Image processing method, path planning method, device, equipment and storage medium |
| CN110992271B (en) * | 2020-03-04 | 2020-07-07 | 腾讯科技（深圳）有限公司 | Image processing method, path planning method, device, equipment and storage medium |
| CN111680554A (en) * | 2020-04-29 | 2020-09-18 | 北京三快在线科技有限公司 | Depth estimation method and device for automatic driving scene and autonomous vehicle |
| 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 |
| CN113219475A (en) * | 2021-07-06 | 2021-08-06 | 北京理工大学 | Method and system for correcting monocular distance measurement by using single line laser radar |
| CN113219475B (en) * | 2021-07-06 | 2021-10-22 | 北京理工大学 | Method and system for correcting monocular distance measurement by using single line laser radar |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106780588A (en) | A kind of image depth estimation method based on sparse laser observations | |
| CN110555434B (en) | Method for detecting visual saliency of three-dimensional image through local contrast and global guidance | |
| CN103824050B (en) | A kind of face key independent positioning method returned based on cascade | |
| CN107204010A (en) | A kind of monocular image depth estimation method and system | |
| CN110419049A (en) | Room layout estimation method and technology | |
| US11182644B2 (en) | Method and apparatus for pose planar constraining on the basis of planar feature extraction | |
| CN109377530A (en) | A kind of binocular depth estimation method based on deep neural network | |
| CN108648161A (en) | The binocular vision obstacle detection system and method for asymmetric nuclear convolutional neural networks | |
| CN108537837A (en) | A kind of method and relevant apparatus of depth information determination | |
| CN108510535A (en) | A kind of high quality depth estimation method based on depth prediction and enhancing sub-network | |
| CN105956597A (en) | Binocular stereo matching method based on convolution neural network | |
| CN101394573B (en) | Panoramagram generation method and system based on characteristic matching | |
| CN106940704A (en) | A kind of localization method and device based on grating map | |
| CN111832655A (en) | Multi-scale three-dimensional target detection method based on characteristic pyramid network | |
| CN106127690A (en) | A kind of quick joining method of unmanned aerial vehicle remote sensing image | |
| CN101866497A (en) | Binocular stereo vision based intelligent three-dimensional human face rebuilding method and system | |
| RU2476825C2 (en) | Method of controlling moving object and apparatus for realising said method | |
| CN104517317A (en) | Three-dimensional reconstruction method of vehicle-borne infrared images | |
| CN111998862B (en) | BNN-based dense binocular SLAM method | |
| CN111402311A (en) | Knowledge distillation-based lightweight stereo parallax estimation method | |
| CN107944459A (en) | A kind of RGB D object identification methods | |
| CN111127522B (en) | Depth optical flow prediction method, device, equipment and medium based on monocular camera | |
| CN104156957A (en) | Stable and high-efficiency high-resolution stereo matching method | |
| CN113724379B (en) | Three-dimensional reconstruction method and device for fusing image and laser point cloud | |
| CN103955942A (en) | SVM-based depth map extraction method of 2D image |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170531 |