CN108471495B - Object multi-angle image acquisition system and method for machine learning and deep learning training - Google Patents
Object multi-angle image acquisition system and method for machine learning and deep learning training Download PDFInfo
- Publication number
- CN108471495B CN108471495B CN201810104230.0A CN201810104230A CN108471495B CN 108471495 B CN108471495 B CN 108471495B CN 201810104230 A CN201810104230 A CN 201810104230A CN 108471495 B CN108471495 B CN 108471495B
- Authority
- CN
- China
- Prior art keywords
- camera
- camera equipment
- grid
- button
- equipment
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
- H04N23/60—Control of cameras or camera modules
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
- H04N23/70—Circuitry for compensating brightness variation in the scene
- H04N23/74—Circuitry for compensating brightness variation in the scene by influencing the scene brightness using illuminating means
Abstract
The invention relates to a system and a method for acquiring multiple angle images of an object for machine learning and deep learning training. The system fixes a plurality of camera devices on a rotatable camera device array fixed grid and controls shooting of the plurality of camera devices by a program to improve efficiency of acquiring an object recognition training set for machine learning training. The number of cameras in the fixed grid of the camera array can be dynamically expanded as desired. The whole camera equipment array fixed grid is controlled by a computer, so that the cost of manual operation is saved, and meanwhile, light sources with different angles can be easily manufactured by matching the lighting lamp which can be fixed according to requirements with the camera equipment array fixed grid. The defects that a traditional manpower acquisition mode is time-consuming and labor-consuming are overcome, and high acquisition demand change inclusion and expansibility of acquisition equipment are realized.
Description
Technical Field
The invention relates to a system and a method for acquiring multiple angle images of an object for machine learning and deep learning training, which are applied to the technical field of object recognition, image processing and other research fields based on machine learning and deep learning. The invention can quickly and accurately acquire the image data of the same object at various angles at the same time, thereby saving the time of research personnel in related fields such as machine learning and the like. In addition, since a large amount of image data of the object in the same state is obtained at the same time, the reliability of the state of the image data of the object in machine learning and deep learning is ensured.
Technical Field
Artificial intelligence is now one of the most popular areas in computer science. In the field of artificial intelligence, machine learning and deep learning are very common algorithms, and in particular, deep learning is the fastest-developing method in recent ten years. A basic condition for object recognition and image processing using machine learning and deep learning methods is that a large amount of image data is required as a training data set. The traditional acquisition of a large amount of image data often requires a great deal of manpower and time. For example, in many face recognition public databases, the method for shooting the head image of the shot person in the face database samples such as ORL and Yale is a single camera, and the shot person continuously rotates to shoot, and the single picture is generally only about 10; and a large-scale face database such as LFW obtains face photos with known identities through a network, the number of the photos of a single person cannot be guaranteed, and a data set for training cannot be built for a specific crowd. The image data acquired under such conventional methods are not able to fully reflect the characteristics of the observed object.
Compared with the traditional method, the method has the advantage that the object image data acquisition efficiency is greatly improved. The image data acquisition device can efficiently acquire the image data of each angle of an object at a certain distance without rotating the object to be shot or a photographer by multiple times of shooting angles, thereby greatly reducing the consumption of manpower and time, improving the completeness of an image data set, and relieving the occurrence of low accuracy of a data model obtained by learning due to few object angles and incomplete data set in the acquisition process of a manual single camera.
Disclosure of Invention
The invention aims to provide a system and a method for acquiring multi-angle images of an object for machine learning and deep learning training, aiming at overcoming the defects in the prior art, and the system and the method can be used for acquiring image data of a target object from different angles at high speed. The device can save a great deal of time and labor input of researchers and developers, and effectively improve the image diversity of each target object, thereby improving the accuracy of the mathematical model obtained after machine learning and deep learning.
In order to achieve the purpose, the invention has the following conception: the computer is simultaneously connected with a plurality of camera devices through data transmission interfaces such as a network cable port, a USB port and the like; the camera equipment and the illuminating lamp are built on the fixed grid; the grid is fixed on a rotatable support. On the basis of the completion of the construction of hardware equipment, a control program of the camera equipment is designed and developed, so that the rapid image acquisition of a multi-angle object is completed.
Acquisition device and technical scheme
The camera shooting part consists of a plurality of camera shooting devices, a camera shooting device array fixing grid, a fixing rotating shaft, a plurality of illuminating lamps and a tripod bracket; the main controller is composed of a computer and a display part. The grid can be very big according to different requirements of users, the grid can be fixed on the ground, portable small grids can be manufactured, and the plurality of camera shooting devices are fixed in the small grids at different positions in the grid. The fixed grid of camera equipment array can the incurvation, and its central point passes through the pivot with the A-frame to be connected for the fixed grid of camera equipment array can rotate around the central point, can adjust the angle that camera equipment shot like this. In addition, a plurality of illuminating lamps are fixed on the grid, and image data of objects under different illumination angles are collected.
The main controller is connected with the camera equipment through a transmission interface, controls the operation of the camera equipment, and stores the acquired image data into a storage device of the computer through the transmission interface. The main controller controls the plurality of camera devices through camera device control programs, and the programs include program interface design and back-end control. The main interface of the program comprises a display area of a plurality of shooting pictures of the shooting equipment, a storage path display bar, a storage folder name input box and 5 buttons, wherein the storage path display bar, the storage folder name input box and the 5 buttons are respectively as follows:
(1) a display area: arranging and displaying a dynamic picture connected with each camera device on a screen;
(2) starting: starting the camera equipment and displaying the dynamic picture of each camera equipment on the display area;
(3) saving the position: the storage position for setting the collected image;
(4) storing path display bar: a save location for displaying the captured image;
(5) storing folder name input box: a folder name for inputting a save image;
(6) setting: each camera device collects a single image or collects each frame of a dynamic image each time;
(7) collecting: the system is used for collecting dynamic pictures shot by all the camera devices and storing the image data in a hard disk;
(8) stopping: the system is used for stopping image acquisition of all the camera devices;
(9) and (3) exiting: the control software is closed and exited.
The operation steps of the object image acquisition device are as follows:
(1) starting a control program:
running a control program to control all connected camera devices and displaying dynamic images acquired by all the camera devices in a display area on a screen;
(2) setting a control mode:
setting a storage position of an acquired image, ensuring accurate shooting angle and setting an acquisition mode of single or multiple image acquisition;
(3) carrying out image acquisition of an object:
and turning on all lighting lights, pressing a 'collecting' button, controlling the camera equipment to collect images through the computer, and storing the collected images at a specified position.
(4) Adjusting lighting light for collection again:
turning on or turning off the lighting lights in sequence, and acquiring object image data under the condition of various different side lights;
(5) rotating the camera array stationary grid:
adjusting the position and the angle of each camera in the fixed grid of the camera array by rotating the grid according to the requirements of a user, and acquiring image data of a target object at more angles;
(6) repeating the step (4) and the step (5):
repeating the step (4) and the step (5) until the image data is completely acquired;
(7) stopping image acquisition:
pressing a 'stop' button to stop image acquisition, wherein the process can be carried out at any time in data acquisition;
(8) exiting and closing:
and pressing an exit button to stop program operation, closing each camera device and exiting the control system.
Compared with the prior equipment, the invention has the following obvious prominent substantive characteristics and remarkable technical progress:
the method breaks through the limitation that the traditional single camera equipment can only collect one target object image at the same time, can simultaneously collect images of a plurality of target objects, shortens the collection time and reduces the labor cost;
the device can increase or reduce the number of the camera devices, the camera devices can rotate, stretch back and forth and adjust positions up and down, good machine learning and deep learning training sets can be constructed rapidly, and great convenience is provided for researchers and application persons related to object image processing.
The invention arranges a plurality of illuminating lamps in the array fixed grid of the camera equipment as required, and can shoot images of objects under light sources of different angles by controlling the illuminating lamps in the shooting process, thereby helping to quickly construct a good machine learning and deep learning training set and providing great convenience for researchers and appliers related to object image processing.
Description of the drawings:
fig. 1 is a front view of a camera array fixed grid, tripod of the image acquisition system apparatus of the present invention.
Fig. 2 is a schematic structural diagram of an image acquisition system device incorporating an illumination lamp according to the present invention.
Fig. 3 is an interface schematic diagram of an image pickup apparatus control program in the present invention.
FIG. 4 is a workflow of human-computer interaction of the present invention.
The specific implementation mode is as follows:
the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings:
the first embodiment is as follows:
referring to fig. 1-3, the multi-angle object image capturing system for machine learning training of the present invention comprises a main controller (1) and a shooting frame (2), wherein the shooting frame (2) comprises a tripod (3) and a camera array fixing grid (4), and is characterized in that: the camera equipment array fixing grid (4) is connected with the tripod (3) through a fixing rotating shaft (6); a plurality of camera devices (5) are fixed on the camera device array fixing grid (4); the camera equipment (5) is fixed and connected with the camera equipment array fixed grid (4); the camera equipment (5) on the camera equipment array fixed grid (4) is connected to the main controller (1) through a connecting wire; the main controller (1) runs a camera equipment control program and controls the opening, closing and shooting of a plurality of camera equipment through a camera equipment control interface (7).
Example two:
referring to fig. 3, the present embodiment is substantially the same as the first embodiment, and is characterized in that the image pickup apparatus control program interface (7), referring to fig. 3, includes an image pickup apparatus shooting picture display area (8), a storage path display bar (9), a save position button (10), a start button (11), a stop button (12), an image capture button (13), a setting (14), and a storage folder name input box (15). The display area (8) of the shot picture of the camera equipment comprises a plurality of display frames (16) of the shot picture of the camera equipment, each display frame (16) of the shot picture of the corresponding camera equipment (5) displays the shot picture, a storage path display bar (9) is used for displaying a selected image storage directory, a storage position button (10) is used for browsing and selecting the image storage directory, a start button (11) is used for opening all connected camera equipment, a stop button (12) is used for closing all connected camera equipment, an image acquisition button (13) is used for controlling the camera equipment to acquire image data, an acquisition mode button (14) is used for setting whether each camera equipment acquires a single image or acquires each frame in a dynamic image each time, and a storage folder name input frame (15) is used for inputting the folder name of the acquired image storage.
Example three:
referring to fig. 2, the present embodiment is substantially the same as the first embodiment, and is characterized in that a plurality of illuminating lamps (17) are fixed on the imaging device array fixing grid (4) for manufacturing light sources at different angles in actual shooting.
Example four:
referring to fig. 4, the present embodiment is substantially the same as the first embodiment, and is characterized in that the operation flow of the multi-angle object image capturing system for machine learning training is as follows:
(1) starting a control program:
running a control program to control all connected camera devices and displaying dynamic images acquired by all the camera devices in a display area on a screen;
(2) setting a control system:
setting a storage position of an acquired image, ensuring accurate shooting angle and setting an acquisition mode of single or multiple image acquisition;
(3) carrying out image acquisition of an object:
and turning on all lighting lights, pressing a 'collecting' button, controlling the camera equipment to collect images through the computer, and storing the collected images at a specified position.
(4) Adjusting lighting light for collection again:
turning on or turning off the lighting lights in sequence, and acquiring object image data under the condition of various different side lights;
(5) rotating the camera array stationary grid:
adjusting the position and the angle of each camera in the fixed grid of the camera array by rotating the grid according to the requirements of a user, and acquiring image data of a target object at more angles;
(6) repeating the step (4) and the step (5):
repeating the step (4) and the step (5) until the image data is completely acquired;
(7) stopping image acquisition:
pressing a 'stop' button to stop image acquisition, wherein the process can be carried out at any time in data acquisition;
(8) exiting and closing:
and pressing an exit button to stop program operation, closing each camera device and exiting the control system.
Claims (3)
1. The utility model provides a multi-angle object image acquisition system that machine learning and deep learning training used, includes a main control unit (1) and one shoots frame (2), it includes a tripod (3) and a camera equipment array fixed grid (4) to shoot frame (2), its characterized in that: the camera equipment array fixed grid (4) is arranged vertical to the ground; the camera equipment array fixing grid (4) is connected with the tripod (3); a plurality of camera devices (5) are fixed on the camera device array fixed grid (4) and face to the collected object; the camera equipment (5) is fixed and connected with the camera equipment array fixed grid (4) through a fixed rotating shaft (6); the camera equipment (5) on the camera equipment array fixed grid (4) is connected to the main controller (1) through a connecting wire; the main controller (1) runs a camera equipment control program and can control the opening, closing and shooting of a plurality of camera equipment through a camera equipment control interface (7); the camera equipment array fixed grid (4) comprises a plurality of small square grids, the size of the grids is set according to different requirements of users, and the grids are fixed on the ground; or small portable grids are manufactured, and a plurality of camera devices (5) are fixed in small square grids at different positions in the grids; the camera equipment array fixing grid (4) can be bent inwards, and the central point of the camera equipment array fixing grid is connected with the triangular support (3) through a rotating shaft, so that the camera equipment array fixing grid (4) can rotate around the central point to adjust the shooting angle of the camera equipment (5); a plurality of illuminating lamps (17) are fixed on the camera equipment array fixing grid (4) and are used for manufacturing light sources at different angles in actual shooting so as to collect image data of objects at different illumination angles.
2. The multi-angle object image capturing system for machine learning and deep learning training as claimed in claim 1, wherein: the control program interface (7) of the camera equipment comprises a camera equipment shooting picture display area (8), a storage path display bar (9), a path selection button (10), a start button (11), a stop button (12), an image acquisition button (13), an acquisition mode setting button (14) and a storage folder name input box (15); the shooting picture display area (8) of the camera equipment comprises a plurality of shooting picture display frames (16) of the camera equipment, each shooting picture display frame (16) of the camera equipment displays a shooting picture corresponding to the camera equipment (5), a storage path display bar (9) is used for displaying a selected image storage directory, a path selection button (10) is used for browsing and selecting the image storage directory, a start button (11) is used for opening all connected camera equipment, a stop button (12) is used for closing all connected camera equipment, an image acquisition button (13) is used for controlling the camera equipment to shoot the current time, an acquisition mode button (14) is set for setting whether each camera equipment acquires a single image or acquires each frame in a dynamic image each time, and a storage folder name input frame (15) is used for inputting a folder name stored by the acquired image.
3. A multi-angle object image acquisition method for machine learning and deep learning training, which is operated by the multi-angle object image acquisition system for machine learning and deep learning training according to claim 1, and is characterized by comprising the following operation steps:
(1) starting a control program:
operating a control program of the camera shooting equipment, controlling all the connected camera shooting equipment, and displaying dynamic images acquired by all the camera shooting equipment in a display area on a screen;
(2) setting a control mode:
setting a storage position of an acquired image, ensuring accurate shooting angle and setting an acquisition mode of single or multiple image acquisition;
(3) carrying out image acquisition of an object:
turning on all lighting lights, pressing an acquisition button, controlling the camera equipment to acquire images through a computer, and storing the acquired images at a specified position;
(4) adjusting lighting light for collection again:
turning on or turning off the lighting lights in sequence, and acquiring object image data under the condition of various different side lights;
(5) rotating the camera array stationary grid:
adjusting the position and the angle of each camera in the fixed grid of the camera array by rotating the grid according to the requirements of a user, and acquiring image data of a target object at more angles;
(6) repeating the step (4) and the step (5):
repeating the step (4) and the step (5) until the image data is completely acquired;
(7) stopping image acquisition:
pressing a 'stop' button to stop image acquisition, wherein the process can be carried out at any time in data acquisition;
(8) exiting and closing:
and pressing an exit button to stop program operation, closing each camera device and exiting the control system.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810104230.0A CN108471495B (en) | 2018-02-02 | 2018-02-02 | Object multi-angle image acquisition system and method for machine learning and deep learning training |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810104230.0A CN108471495B (en) | 2018-02-02 | 2018-02-02 | Object multi-angle image acquisition system and method for machine learning and deep learning training |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108471495A CN108471495A (en) | 2018-08-31 |
CN108471495B true CN108471495B (en) | 2020-09-08 |
Family
ID=63266149
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810104230.0A Active CN108471495B (en) | 2018-02-02 | 2018-02-02 | Object multi-angle image acquisition system and method for machine learning and deep learning training |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108471495B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109410684B (en) * | 2018-12-21 | 2024-04-16 | 无锡职业技术学院 | Real standard system of making a video recording |
CN110031403B (en) * | 2019-04-04 | 2020-03-31 | 山东大学 | Full-automatic rock specimen image acquisition device and method |
GB202116558D0 (en) * | 2021-11-17 | 2021-12-29 | Ocado Innovation Ltd | Camera unit |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104616008A (en) * | 2015-02-15 | 2015-05-13 | 四川川大智胜软件股份有限公司 | Multi-view two-dimension facial image collecting platform and collecting method thereof |
CN106322062A (en) * | 2016-11-01 | 2017-01-11 | 北京视觉智能科技有限公司 | Multi-angle shooting support and system |
CN107135336A (en) * | 2016-02-29 | 2017-09-05 | 华为技术有限公司 | A kind of video camera array |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20130068193A (en) * | 2011-12-14 | 2013-06-26 | 한국전자통신연구원 | Multi images supplying system and multi images shooting device thereof |
WO2017215767A1 (en) * | 2016-06-17 | 2017-12-21 | Huawei Technologies Co., Ltd. | Exposure-related intensity transformation |
CN106791301A (en) * | 2016-12-13 | 2017-05-31 | 天津代双科技有限公司 | A kind of indoor unmanned plane multi-angled shooting control system of electronic information technical field |
-
2018
- 2018-02-02 CN CN201810104230.0A patent/CN108471495B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104616008A (en) * | 2015-02-15 | 2015-05-13 | 四川川大智胜软件股份有限公司 | Multi-view two-dimension facial image collecting platform and collecting method thereof |
CN107135336A (en) * | 2016-02-29 | 2017-09-05 | 华为技术有限公司 | A kind of video camera array |
CN106322062A (en) * | 2016-11-01 | 2017-01-11 | 北京视觉智能科技有限公司 | Multi-angle shooting support and system |
Also Published As
Publication number | Publication date |
---|---|
CN108471495A (en) | 2018-08-31 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108471495B (en) | Object multi-angle image acquisition system and method for machine learning and deep learning training | |
CN104427252B (en) | Method and its electronic equipment for composograph | |
Goh et al. | The CMU face in action (FIA) database | |
WO2022156683A1 (en) | Image processing method and apparatus, and photographic support, electronic device and readable storage medium | |
WO2020211427A1 (en) | Segmentation and recognition method, system, and storage medium based on scanning point cloud data | |
CN101828905B (en) | Chinese medicine color inspection system | |
WO2020258258A1 (en) | Target tracking method and system, readable storage medium and moving platform | |
WO2020119254A1 (en) | Method and device for filter recommendation, electronic equipment, and storage medium | |
CN106846462B (en) | insect recognition device and method based on three-dimensional simulation | |
CN111833423A (en) | Presentation method, presentation device, presentation equipment and computer-readable storage medium | |
CN113329172B (en) | Shooting method and device and electronic equipment | |
CN112348003A (en) | Airplane refueling scene recognition method and system based on deep convolutional neural network | |
CN106777071B (en) | Method and device for acquiring reference information by image recognition | |
CN111866374A (en) | Image shooting method and device, pan-tilt camera and storage medium | |
JP2004532441A5 (en) | ||
CN110969696A (en) | Method and system for three-dimensional modeling rapid space reconstruction | |
CN211240747U (en) | Soybean plant species test instrument | |
CN112906610A (en) | Method for living body detection, electronic circuit, electronic apparatus, and medium | |
CN116189076A (en) | Observation and identification system and method for bird observation station | |
CN207051928U (en) | A kind of plane pressed characters automatic identification equipment based on train bogie | |
CN107667522A (en) | Adjust the length of live image | |
CN105657265A (en) | Full-automatic shooting imaging system and method | |
CN102231800B (en) | Automatic aiming method for camera | |
JP2020144785A (en) | Image collection device, image collection system, image collection method, and image collection program | |
CN202771488U (en) | Image automatic mosaic device based on halcon algorithm |
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: 20211129 Address after: 200444 No. 668, SHANGDA Road, Baoshan District, Shanghai Patentee after: Shanghai Mingqi Network Technology Co.,Ltd. Address before: 200444 No. 99, upper road, Shanghai, Baoshan District Patentee before: Shanghai University |