CN109657702B - 3D depth semantic perception method and device - Google Patents
3D depth semantic perception method and device Download PDFInfo
- Publication number
- CN109657702B CN109657702B CN201811403410.5A CN201811403410A CN109657702B CN 109657702 B CN109657702 B CN 109657702B CN 201811403410 A CN201811403410 A CN 201811403410A CN 109657702 B CN109657702 B CN 109657702B
- Authority
- CN
- China
- Prior art keywords
- image
- stored
- stored image
- points
- target
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a 3D depth semantic perception algorithm and a device, wherein the 3D depth semantic perception algorithm comprises the following steps: acquiring a 3D image; matching the 3D image with a pre-stored image library, wherein semantic information is marked on an image point on each pre-stored image in the pre-stored image library; and sensing semantic information of the image points on the 3D image by utilizing the semantic information of the image points on the pre-stored image through an artificial intelligence deep learning algorithm. The 3D depth semantic perception algorithm and the device can acquire more standard digital point clouds, so that the acquired 3D images are easier to manage and control, and resources consumed by operation can be reduced.
Description
Technical Field
The invention relates to a 3D depth semantic perception method and device.
Background
The 3D camera, which is manufactured by using a 3D lens, generally has two or more image pickup lenses, and has a pitch close to the pitch of human eyes, and can capture different images of the same scene seen by similar human eyes. The holographic 3D has a disc 5 above the lens.
The first 3D camera to date the 3D revolution has all been around the hollywood heavy-pound large and major sporting events. With the advent of 3D cameras, this technology is one step closer to home users. After the camera is introduced, people can capture each memorable moment of life, such as the first step taken by children, university graduate celebration and the like, by using a 3D lens in the future.
A 3D camera typically has more than two lenses. The 3D camera functions like a human brain, and can fuse two lens images together to form a 3D image. These images can be played on a 3D television, and can be viewed by viewers wearing so-called actively shuttered glasses, or directly viewed by naked-eye 3D display devices. The 3D shutter glasses can rapidly alternately open and close the lenses of the left and right glasses at a rate of 60 times per second. This means that each eye sees a slightly different picture of the same scene, so the brain can thus think that it is enjoying a single picture in 3D.
The existing 3D camera has the defects that the image acquired by the 3D camera is not easy to process and control, and the 3D image occupies a larger space.
Disclosure of Invention
The invention aims to overcome the defects that images acquired by a 3D camera are not easy to process and control and the occupied space of the 3D images is large in the prior art, and provides a 3D depth semantic perception method and a 3D depth semantic perception device which can acquire more standard digital point clouds and enable the acquired 3D images to be easier to manage and control.
The invention solves the technical problems through the following technical scheme:
A3D depth semantic perception method is characterized in that the 3D depth semantic perception method comprises the following steps:
acquiring a 3D image;
matching the 3D image with a pre-stored image library, wherein semantic information is marked on an image point on each pre-stored image in the pre-stored image library;
and sensing semantic information of the image points on the 3D image by utilizing the semantic information of the image points on the pre-stored image through an artificial intelligence deep learning algorithm.
The 3D image is a face image.
Machine learning is achieved through an algorithm, so that a machine can learn rules from a large amount of data input from the outside, and recognition and judgment are carried out. According to the method and the device, the images in the standard images (pre-stored image library) are learned, the rules of the images in the pre-stored image library are obtained, and therefore the 3D images can be marked, and a computer can automatically identify the semantics of each digital point (the meaning contained in the symbol is the semantics).
Preferably, the 3D depth semantic perception method includes:
and acquiring a target image to generate the pre-stored image library, wherein the target image is an accurate image acquired by an industrial 3D camera, and semantic information is marked on a target image point on the accurate image.
Preferably, each pre-stored image in the pre-stored image library is provided with a function expression representing the relationship between image points, and the 3D depth semantic perception method includes:
and setting a function formula between image points on the 3D image by utilizing the pre-stored function formula on the image through an artificial intelligence deep learning algorithm.
Preferably, each pre-stored image in the pre-stored image library is divided into a plurality of regions, and each region is provided with a function expression representing the relationship between image points in the same region, and the 3D depth semantic perception method includes:
dividing regions on the 3D image by using the region positions on the pre-stored image through an artificial intelligence deep learning algorithm;
and for a target area on the 3D image, setting a function between image points in the target area on the 3D image by utilizing the pre-stored function on the image through an artificial intelligence deep learning algorithm.
Preferably, the 3D depth semantic perception method includes:
for a target pre-stored image in a pre-stored image library, acquiring a function formula between adjacent image points in the target pre-stored image, wherein the function formula is a polynomial function;
and obtaining a plurality of parting lines passing through adjacent image points through artificial intelligence deep learning, calculating the sum of times of the highest-order item of a polynomial function between all the adjacent image points passing through the parting lines, and dividing the region of the target pre-stored image by the parting lines with the sum of times lower than a preset value.
The invention also provides a 3D depth semantic perception device which is characterized in that the 3D depth semantic perception device comprises an acquisition module, a matching module and a processing module,
the acquisition module is used for acquiring a 3D image;
the matching module is used for matching the 3D image with a pre-stored image library, and semantic information is marked on an image point on each pre-stored image in the pre-stored image library;
the processing module is used for sensing semantic information of the image points on the 3D image by utilizing the semantic information of the image points on the pre-stored image through an artificial intelligence deep learning algorithm.
Preferably, the 3D depth semantic perception device comprises a generation module,
the generating module is used for acquiring a target image to generate the pre-stored image library, wherein the target image is an accurate image acquired by an industrial 3D camera, and semantic information is marked on a target image point on the accurate image.
Preferably, each pre-stored image in the pre-stored image library is provided with a function expression representing the relationship between the image points,
the processing module is further used for setting a function between image points on the 3D image by utilizing the pre-stored function on the image through an artificial intelligence deep learning algorithm.
Preferably, each pre-stored image in the pre-stored image library is divided into several areas, each area is provided with a function formula representing the relationship between the image points in the same area,
the processing module is also used for dividing the 3D image into regions by utilizing the region positions on the pre-stored image through an artificial intelligence deep learning algorithm;
for a target area on the 3D image, the processing module is further used for setting a function between image points in the target area on the 3D image by utilizing the function on the pre-stored image through an artificial intelligence deep learning algorithm.
Preferably, the 3D depth semantic perception device includes an analysis module and a calculation module,
for a target pre-stored image in a pre-stored image library, the analysis module is used for acquiring a function formula between adjacent image points in the target pre-stored image, and the function formula is a polynomial function;
the calculation module is used for acquiring a plurality of dividing lines passing through adjacent image points through artificial intelligence deep learning, calculating the sum of times of the highest-order item of a polynomial function between all the adjacent image points passing through the dividing lines, and dividing the region of the target pre-stored image by the dividing lines with the sum of times lower than a preset value.
On the basis of the common knowledge in the field, the above preferred conditions can be combined randomly to obtain the preferred embodiments of the invention.
The positive progress effects of the invention are as follows:
the 3D depth semantic perception method and the device can acquire more standard digital point clouds, so that the acquired 3D images are easier to manage and control, and resources consumed by operation can be reduced.
Drawings
Fig. 1 is a flowchart of a 3D depth semantic perception method according to embodiment 1 of the present invention.
Fig. 2 is a flowchart of a 3D depth semantic perception method according to embodiment 2 of the present invention.
Detailed Description
The invention is further illustrated by the following examples, which are not intended to limit the scope of the invention.
Example 1
The embodiment provides a 3D depth semantic sensing device, which includes an obtaining module, a matching module, a generating module and a processing module,
the acquisition module is used for acquiring a 3D image;
the matching module is used for matching the 3D image with a pre-stored image library, and semantic information is marked on an image point of each pre-stored image in the pre-stored image library;
the processing module is used for sensing semantic information of the image points on the 3D image by utilizing the semantic information of the image points on the pre-stored image through an artificial intelligence deep learning algorithm.
The matching of the pre-stored image library is obtained through a generating module, the generating module is used for obtaining a target image and generating the pre-stored image library, the target image is obtained through an industrial 3D camera, and semantic information is marked on a target image point on the accurate image.
Digital points (image points) in an image library are prestored, semantic information can be marked manually or the identities of the image points can be identified through artificial intelligence, and then the semantic information is added. The semantic information can record the identity of the image points, so that the initial 3D image is subjected to digital processing, and a machine acquires the meaning of each image point in the image.
According to the method and the device, the images in the standard images (pre-stored image library) are learned, the rules of the images in the pre-stored image library are obtained, and therefore the 3D images can be marked, and a computer can automatically identify the semantics of each digital point (the meaning contained in the symbol is the semantics).
Referring to fig. 1, with the 3D depth semantic sensing apparatus, the embodiment further provides a 3D depth semantic sensing method, including:
and 102, sensing semantic information of image points on the 3D image by utilizing the semantic information of the pre-stored image points on the image through an artificial intelligence deep learning algorithm.
The embodiment further provides a method for generating a pre-stored image library, including: and acquiring a target image to generate the pre-stored image library, wherein the target image is an accurate image acquired by an industrial 3D camera, and semantic information is marked on a target image point on the accurate image.
The 3D depth semantic perception method and the device can acquire more standard digital point clouds, so that the acquired 3D images are easier to manage and control, and resources consumed by operation can be reduced.
Example 2
This embodiment is substantially the same as embodiment 1 except that:
the 3D depth semantic perception device comprises an analysis module and a calculation module.
Each pre-stored image in the pre-stored image library is provided with a function expression representing the relationship between the image points,
the processing module is further used for setting a function between image points on the 3D image by utilizing the pre-stored function on the image through an artificial intelligence deep learning algorithm.
Specifically, each pre-stored image in the pre-stored image library is divided into several areas, each area is configured with a function expression representing the relationship between the image points in the same area,
the processing module is further used for dividing the 3D image into regions by utilizing the region positions on the pre-stored image through an artificial intelligence deep learning algorithm;
for a target area on the 3D image, the processing module is further used for setting a function between image points in the target area on the 3D image by utilizing the function on the pre-stored image through an artificial intelligence deep learning algorithm.
In order to further digitize the 3D image, the present embodiment adds a linkage relationship to the 3D image, so that when one image point is adjusted, linkage adjustment is made to other image points. Through the learning of the human face space shape, the association between image points can be obtained, so that the method is applied to the fields of image modification, reshaping and the like.
Because the relation between each image point is very complicated, if the image points which are involved in the movement of one image point are calculated from the whole, the calculation amount is very huge, so that the image points with obvious linkage relation are divided into the same area, the relation with the image points outside the area is cut off, and the calculation amount can be reduced.
Further, the present embodiment provides a way how to divide the regions, including:
for a target pre-stored image in a pre-stored image library, the analysis module is used for acquiring a function formula between adjacent image points in the target pre-stored image, and the function formula is a polynomial function;
the calculation module is used for acquiring a plurality of segmentation lines passing through adjacent image points through artificial intelligence deep learning, calculating the sum of times of the highest-order item of a polynomial function between all the adjacent image points through which the segmentation lines pass, and dividing the region of the target pre-stored image by the segmentation lines of which the sum of times is lower than a preset value.
Referring to fig. 2, correspondingly, the 3D depth semantic perception method according to this embodiment includes, after step 101 in embodiment 1:
200, for a target pre-stored image in a pre-stored image library, acquiring a function formula between adjacent image points in the target pre-stored image, wherein the function formula is a polynomial function;
Step 200 and step 201 realize that the region position on the pre-stored image is used for dividing the region on the 3D image through an artificial intelligence deep learning algorithm;
While specific embodiments of the invention have been described above, it will be appreciated by those skilled in the art that these are by way of example only, and that the scope of the invention is defined by the appended claims. Various changes or modifications to these embodiments may be made by those skilled in the art without departing from the principle and spirit of this invention, and these changes and modifications are within the scope of this invention.
Claims (4)
1. A3D depth semantic perception method is characterized in that the 3D depth semantic perception method comprises the following steps:
acquiring a 3D image;
matching the 3D image with a pre-stored image library, wherein semantic information is marked on an image point on each pre-stored image in the pre-stored image library;
sensing semantic information of image points on the 3D image by utilizing the semantic information of the image points on the pre-stored image through an artificial intelligence deep learning algorithm;
each pre-stored image in the pre-stored image library is provided with a function expression for representing the relation between image points, and the 3D depth semantic perception method further comprises the following steps:
setting a function formula between image points on the 3D image by utilizing the pre-stored function formula on the image through an artificial intelligence deep learning algorithm;
each pre-stored image in the pre-stored image library is divided into a plurality of areas, a function expression for representing the relation between image points in the same area is arranged in each area, and the 3D depth semantic perception method further comprises the following steps:
dividing regions on the 3D image by using the region positions on the pre-stored image through an artificial intelligence deep learning algorithm;
setting a function formula between image points in a target area on the 3D image by utilizing the pre-stored function formula on the image through an artificial intelligence deep learning algorithm;
the 3D depth semantic perception method comprises the following steps:
for a target pre-stored image in a pre-stored image library, acquiring a function formula between adjacent image points in the target pre-stored image, wherein the function formula is a polynomial function;
and obtaining a plurality of parting lines passing through adjacent image points through artificial intelligence deep learning, calculating the sum of times of the highest-order item of a polynomial function between all the adjacent image points passing through the parting lines, and dividing the region of the target pre-stored image by the parting lines with the sum of times lower than a preset value.
2. The 3D depth semantic perception method according to claim 1, wherein the 3D depth semantic perception method includes:
and acquiring a target image to generate the pre-stored image library, wherein the target image is an accurate image acquired by an industrial 3D camera, and semantic information is marked on a target image point on the accurate image.
3. A3D depth semantic perception device is characterized in that the 3D depth semantic perception device comprises an acquisition module, a matching module and a processing module,
the acquisition module is used for acquiring a 3D image;
the matching module is used for matching the 3D image with a pre-stored image library, and semantic information is marked on an image point on each pre-stored image in the pre-stored image library;
the processing module is used for sensing semantic information of image points on the 3D image by utilizing the semantic information of the pre-stored image points on the image through an artificial intelligence deep learning algorithm;
wherein each pre-stored image in the pre-stored image library is provided with a function expression representing the relationship between the image points,
the processing module is also used for setting a function formula between image points on the 3D image by utilizing the pre-stored function formula on the image through an artificial intelligence deep learning algorithm;
wherein each pre-stored image in the pre-stored image library is divided into a plurality of regions, each region is provided with a function expression representing the relationship between image points in the same region,
the processing module is further used for dividing the 3D image into regions by utilizing the region positions on the pre-stored image through an artificial intelligence deep learning algorithm;
for a target area on the 3D image, the processing module is further used for setting a function between image points in the target area on the 3D image by using the pre-stored function on the image through an artificial intelligence deep learning algorithm;
wherein, the 3D depth semantic perception device comprises an analysis module and a calculation module,
for a target pre-stored image in a pre-stored image library, the analysis module is used for acquiring a function formula between adjacent image points in the target pre-stored image, and the function formula is a polynomial function;
the calculation module is used for acquiring a plurality of dividing lines passing through adjacent image points through artificial intelligence deep learning, calculating the sum of times of the highest-order item of a polynomial function between all the adjacent image points passing through the dividing lines, and dividing the region of the target pre-stored image by the dividing lines with the sum of times lower than a preset value.
4. The 3D depth semantic perception device according to claim 3, wherein the 3D depth semantic perception device includes a generation module,
the generating module is used for acquiring a target image to generate the pre-stored image library, wherein the target image is an accurate image acquired by an industrial 3D camera, and semantic information is marked on a target image point on the accurate image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811403410.5A CN109657702B (en) | 2018-11-23 | 2018-11-23 | 3D depth semantic perception method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811403410.5A CN109657702B (en) | 2018-11-23 | 2018-11-23 | 3D depth semantic perception method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109657702A CN109657702A (en) | 2019-04-19 |
CN109657702B true CN109657702B (en) | 2023-02-07 |
Family
ID=66112145
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811403410.5A Active CN109657702B (en) | 2018-11-23 | 2018-11-23 | 3D depth semantic perception method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109657702B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110188616B (en) * | 2019-05-05 | 2023-02-28 | 上海盎维信息技术有限公司 | Space modeling method and device based on 2D and 3D images |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105551084A (en) * | 2016-01-28 | 2016-05-04 | 北京航空航天大学 | Outdoor three-dimensional scene combined construction method based on image content parsing |
CN107577994A (en) * | 2017-08-17 | 2018-01-12 | 南京邮电大学 | A kind of pedestrian based on deep learning, the identification of vehicle auxiliary product and search method |
CN108848366A (en) * | 2018-07-05 | 2018-11-20 | 盎锐(上海)信息科技有限公司 | Information acquisition device and method based on 3D video camera |
-
2018
- 2018-11-23 CN CN201811403410.5A patent/CN109657702B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105551084A (en) * | 2016-01-28 | 2016-05-04 | 北京航空航天大学 | Outdoor three-dimensional scene combined construction method based on image content parsing |
CN107577994A (en) * | 2017-08-17 | 2018-01-12 | 南京邮电大学 | A kind of pedestrian based on deep learning, the identification of vehicle auxiliary product and search method |
CN108848366A (en) * | 2018-07-05 | 2018-11-20 | 盎锐(上海)信息科技有限公司 | Information acquisition device and method based on 3D video camera |
Non-Patent Citations (2)
Title |
---|
基于卷积神经网络的车道线语义分割算法;徐国晟等;《电子测量与仪器学报》;20180715(第07期);第94-99页 * |
基于深度学习的图像语义分割算法概述;华敏杰;《中国战略新兴产业》;20180128(第04期);第120页 * |
Also Published As
Publication number | Publication date |
---|---|
CN109657702A (en) | 2019-04-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP3236391B1 (en) | Object detection and recognition under out of focus conditions | |
JP5073670B2 (en) | Stereoscopic image display method and method and apparatus for generating three-dimensional image data from input of two-dimensional image data | |
US11508038B2 (en) | Image processing method, storage medium, image processing apparatus, learned model manufacturing method, and image processing system | |
CN108600729B (en) | Dynamic 3D model generation device and image generation method | |
CN101653011A (en) | System and method for combining text with three-dimensional content | |
CN109035147B (en) | Image processing method and device, electronic device, storage medium and computer equipment | |
Kim et al. | Binocular fusion net: deep learning visual comfort assessment for stereoscopic 3D | |
CN108347505B (en) | Mobile terminal with 3D imaging function and image generation method | |
CN103370943A (en) | Imaging device and imaging method | |
JP7479729B2 (en) | Three-dimensional representation method and device | |
CN108391116B (en) | Whole body scanning device and method based on 3D imaging technology | |
KR20190120106A (en) | Method for determining representative image of video, and electronic apparatus for processing the method | |
CN109657702B (en) | 3D depth semantic perception method and device | |
CN109636926B (en) | 3D global free deformation method and device | |
CN105263011B (en) | Multi-view image shows equipment and its multi-view image display methods | |
CN109657559B (en) | Point cloud depth perception coding engine device | |
CN108513122B (en) | Model adjusting method and model generating device based on 3D imaging technology | |
CN109379511B (en) | 3D data security encryption algorithm and device | |
CN111161399B (en) | Data processing method and assembly for generating three-dimensional model based on two-dimensional image | |
CN109448066B (en) | 3D data polar compression method and device | |
CN108848366B (en) | Information acquisition device and method based on 3D camera | |
CN109089105B (en) | Model generation device and method based on depth perception coding | |
CN109379509B (en) | Point cloud encryption compression engine | |
CN109218703B (en) | Data processing device and method based on 3D camera | |
CN110876050B (en) | Data processing device and method based on 3D camera |
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 | ||
TR01 | Transfer of patent right | ||
TR01 | Transfer of patent right |
Effective date of registration: 20230406 Address after: 518000 1101-g1, BIC science and technology building, No. 9, scientific research road, Maling community, Yuehai street, Nanshan District, Shenzhen, Guangdong Province Patentee after: Shenzhen Yujing Information Technology Co.,Ltd. Address before: 201703 No.206, building 1, no.3938 Huqingping Road, Qingpu District, Shanghai Patentee before: UNRE (SHANGHAI) INFORMATION TECHNOLOGY Co.,Ltd. |