Machine learning device and learning method

Abstract

The invention provides a machine learning device and method, and the device comprises an objective table, the upper side of the objective table is provided with a lighting device, the upper side of the lighting device is provided with a camera, the positions of the lighting device and the camera can be adjusted in the vertical direction, the center of the lighting device is provided with a through hole, and the through hole is communicated with the camera. And the camera shoots through the through hole. The ratio value of the projection area of the shot object to the shooting distance can be obtained through learning, so that the shooting distance is adjusted according to the ratio value and the size of the shot object; the optimal shooting pattern is obtained by shooting and comparing the example pattern, so that the proper illumination intensity in the current illumination environment is determined, the equipment has the capability of automatically adjusting the shooting distance and the illumination intensity during shooting, and the technology aims at controlling the shooting distance and the shooting illumination intensity to obtain a picture with better quality, so that the shooting efficiency is improved. And the requirements of users are met.

Description

Technical Field

[0001] The present invention relates to the technical field of shooting lighting device control, and in particular to a machine learning device and a learning method. Background Art

[0002] As online e-commerce becomes increasingly popular, all kinds of products need to be photographed before being put online. Product photography is an important and technical task. Product photos taken by non-professionals without professional equipment are often unsatisfactory. Moreover, even if professional photography equipment is available, it is difficult to take good photos without certain photography skills. Therefore, product photography often requires paid professional photography. This often costs a lot of money and time for some start-up businesses.

[0003] Therefore, it is necessary to design an automatic photography device to help people photograph products. It can automatically adjust the shooting position and lighting brightness to obtain high-quality product images. However, the operation of such an automatic photography device depends on its own acquisition of shooting methods, which require extensive shooting learning to develop. Therefore, how the photography device learns to acquire shooting methods is a problem that needs to be considered. Summary of the Invention

[0004] In response to the problems pointed out in the background technology, the present invention proposes a machine learning device and a learning method to solve the above technical problems.

[0005] The technical solution of the present invention is achieved as follows: A machine learning device includes a stage, a lighting device is provided on the upper side of the stage, a camera is provided on the upper side of the lighting device, the lighting device and the camera can be adjusted in the upper and lower directions, a through hole is provided in the center of the lighting device, and the camera takes pictures through the through hole.

[0006] The present invention is further configured such that a plurality of light detectors are distributed on the stage, and the light detectors can emit electrical signals after being illuminated by light.

[0007] The present invention is further configured to further include a central processing unit, and the light detector, the lighting device, and the camera are all connected to the central processing unit.

[0008] The present invention is further configured such that the lighting device is in a circular shape, and a plurality of light sources are provided on the lighting device around the through hole.

[0009] The present invention is further configured to include a driving device 1 for driving the lighting device to move up and down and a driving device 2 for driving the camera to move up and down, and both the driving device 1 and the driving device 2 are controlled by a central processing unit.

[0010] The present invention is further configured such that an example pattern is provided on the stage.

[0011] The present invention is further configured such that a lens is provided in the through hole.

[0012] A machine learning method, using the above-mentioned machine learning device, comprises the following steps: Step (1) is to learn how to calculate the projected area of ​​the object. The object is placed on the stage, and the lighting device illuminates the object. The light detectors in the area on the stage that is not blocked by the object are activated by the light and send out signals. In this way, the total number of light detectors on the stage and the number of activated light detectors are known, and the number of unactivated light detectors is calculated. The projected area of ​​the object is calculated based on the number of unactivated light detectors. Step (2) is to learn the camera shooting distance, use objects with different projection areas to shoot, shoot the same object with the camera at different heights, select the best shot, store the projection area and shooting distance of the object, and obtain the ratio value A of the projection area of ​​the object to the shooting distance by shooting objects with different projection areas and storing the optimal shooting distance; The present invention is further configured to perform illumination intensity learning of the lighting device; photograph an example pattern on the stage, execute steps (1) and (2) in sequence, determine the camera shooting distance, adjust different illumination intensities, and photograph, compare the photographs taken at different illumination intensities with the example patterns stored in the central processing unit, obtain the best photographing pattern, and the illumination intensity corresponding to the best photographing pattern is selected as the photographing illumination intensity.

[0013] By adopting the above technical solution, the beneficial effects of the present invention are: The present invention provides a machine learning device and learning method, which can learn the ratio of the projected area of ​​the object to the shooting distance, and thus adjust the shooting distance according to the ratio and the size of the object; by shooting and comparing example patterns, the optimal shooting pattern is obtained, thereby determining the appropriate light intensity under the current lighting environment, so that the device has the ability to automatically adjust the shooting distance and the light intensity during shooting. The technology of this application aims to control the shooting distance and shooting light intensity to obtain high-quality photos to meet the needs of users. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0015] Figure 1 Schematic diagram of the structure of the machine learning device of the present invention; Figure 2 Schematic diagram of the structure of the stage of the present invention; Figure 3 This is a control principle block diagram of the present invention. DETAILED DESCRIPTION

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0017] Reference as follows Figure 1-3 The present invention will be described: A machine learning device includes a stage 1, which serves as a basic supporting component and is horizontally arranged to provide a stable surface for placing products to be photographed. A lighting device 2 is provided above the stage 1, and a camera 3 is located above the lighting device 2. Both the lighting device 2 and the camera 3 are adjustable in the vertical direction. A through hole is provided in the center of the lighting device 2, and a lens is located in the through hole. This design enables the camera 3 to photograph the products on the stage 1 through the through hole.

[0018] A frame 4 is provided on the side of the stage 1, and the frame 4 serves as a supporting structure for the lighting device 2 and the camera 3. The lighting device 2 and the camera 3 are both slidably connected to the frame 4, thereby realizing position adjustment of the two in the vertical direction.

[0019] It also includes a driving device 1 for driving the lighting device 2 to move up and down, and a driving device 2 for driving the camera 3 to move up and down. The lighting device 2 and the camera 3 can be controlled to move separately by the two driving devices, or a single driving device can be used to simultaneously control the movement of the lighting device 2 and the camera 3. The driving device can be a cylinder or an electric push rod, etc. When working, the lighting device 2 and the camera 3 move simultaneously, and the distance between them remains unchanged.

[0020] Camera 3 is located above lighting device 2 and captures images through a through-hole in the center of lighting device 2. The vertical movement of camera 3 is controlled by drive device 2, which can also be a pneumatic cylinder or electric push rod. During operation, camera 3 and lighting device 2 must move synchronously, with a constant distance between them to ensure consistent lighting and angle.

[0021] The lighting device 2 is annular in shape, with a through-hole at its center. Multiple light sources are positioned around the through-hole, illuminating downwards, much like a headlamp on the front of a camera. The lighting device 2 has multiple light sources distributed around the through-hole, illuminating downwards and providing ample light for shooting. The vertical movement of the lighting device 2 is controlled by a drive unit 1, which can be a pneumatic cylinder or electric push rod, ensuring that the height of the lighting device 2 can be precisely adjusted during operation.

[0022] A plurality of photodetectors 6 are distributed on the stage 1. These photodetectors 6 are arranged in an array comprising N*M photodetectors 6, with a distance L between adjacent photodetectors 6. When illuminated by light, the photodetectors 6 emit electrical signals, which are received by the central processing unit (CPU). The photodetectors 6 have a photoelectric conversion function. When illuminated by light, they convert the optical signal into an electrical signal and transmit this electrical signal to the CPU. In this way, the CPU can obtain information about the illumination status of the photodetectors 6.

[0023] When photographing a product, the product is placed on stage 1, lighting device 2 illuminates the product, and camera 3 takes the photo. As lighting device 2 illuminates, light detectors 6 not obstructed by the product are activated. The central processing unit (CPU) determines the number of activated light detectors 6. The CPU stores the total number of light detectors 6, allowing it to calculate the number of unactivated light detectors 6 obstructed by the product. Based on the number of unactivated light detectors 6 and the distance L between adjacent light detectors 6, the area obstructed by the product can be roughly calculated, yielding the projected area of ​​the product.

[0024] The lighting device 2 and the camera 3 are both connected to the central processing unit, which controls the operation of the lighting device 2 and the camera 3. The central processing unit also controls the movement of the lighting device 2 and the camera 3 through the driving device.

[0025] Stage 1 is provided with a sample pattern 7. This is a planar structure, which can be made of paper or film in the form of a sheet or plate. It is designed to be removable and portable. Sample pattern 7 is used during machine learning to determine the optimal lighting intensity for shooting, providing a reference standard for the device's shooting under different ambient lighting conditions.

[0026] The central processing unit (CPU), the control core of the entire device, connects to lighting unit 2, camera 3, and both driver units 1 and 2. The CPU controls the operating states of lighting unit 2 and camera 3, including starting and shutting down them and adjusting their parameters. Furthermore, through driver units 1 and 2, the CPU precisely controls their vertical movement, ensuring they can be adjusted to meet specific needs.

[0027] A machine learning method, using the above-mentioned machine learning device, comprises the following steps: Step (1), learning to calculate the projected area of ​​the object: Place the object on the stage 1, turn on the lighting device 2 to illuminate the object. At this time, the light detector 6 in the area on the stage 1 that is not blocked by the object will be activated by the light, and the activated light detector 6 will send an electrical signal to the central processing unit. The central processing unit has pre-stored the total number of light detectors 6 on the stage 1, and can calculate the number of unactivated light detectors 6 by receiving the signals of the activated light detectors 6. Based on the number of unactivated light detectors 6 and the distance L between adjacent light detectors 6, the projection area of ​​the object on the stage 1 can be roughly calculated using the principle of geometric calculation.

[0028] Step (2), camera 3 shooting distance learning: select N objects with different projection areas, and for each object, camera 3 shoots at different heights (i.e., different distances between the camera and the object), and obtains a set of shooting data for each shooting. This set of data includes the projection area of ​​the object and the shooting distance of camera 3. In each set of shooting data, the best shot is selected based on certain evaluation criteria (such as clarity, detail completeness and other objective indicators), and the projection area and shooting distance of the object corresponding to the best shot are recorded, and the ratio value A1 of the two is calculated. Repeat the above process, shoot, screen and calculate the remaining objects with different projection areas, and obtain the ratio values ​​A2, A3...AN respectively. Finally, perform arithmetic average operation on these ratio values ​​to obtain the ratio value A of the projection area of ​​the object and the shooting distance. This ratio value A will serve as an important basis for subsequent shooting distance adjustment.

[0029] Learning the illumination intensity of the lighting device 2: During the product shooting process, the presence of ambient light will affect the shooting effect, so the illumination intensity of the lighting device 2 needs to be adjusted according to the changes in the ambient light. In the specific operation, first place the sample pattern 7 on the stage 1, and determine the shooting distance of the camera 3 according to the above steps (1) and (2). Then, adjust the illumination intensity of the lighting device 2 and shoot the sample pattern 7 under different illumination intensities. Compare the photos obtained from each shot with the sample patterns pre-stored in the central processor. The comparison process can use an image similarity algorithm to calculate the similarity value between each shot and the stored sample pattern. Find the shot with the highest similarity value. This picture is the best shooting pattern, and its corresponding illumination intensity is selected as the shooting light intensity under the current environment. Since the ambient light in different locations and different time periods is different, it is generally necessary to adjust the illumination intensity before each shooting to ensure the stability and reliability of the shooting effect.

[0030] Analysis of beneficial effects of technical solutions: 1. Automatic adjustment of shooting distance: By calculating and learning the ratio A between the projected area of ​​the object and the shooting distance using machine learning, the device accurately calculates the appropriate shooting distance based on the actual size (i.e., projected area) of the product being photographed, combined with the predetermined ratio A. During the actual shooting process, the central processing unit controls drive unit 2 to adjust the height of camera 3 to ensure it is at the optimal shooting distance. This prevents image distortion and loss of detail caused by inappropriate shooting distance, thereby improving the quality and accuracy of the captured photos.

[0031] 2. Automatic adjustment of light intensity: Using sample pattern 7 to learn illumination intensity, by comparing and analyzing photos taken at different illumination intensities with stored sample patterns, the optimal illumination intensity for the current ambient light conditions can be determined. In practical applications, regardless of ambient light variations, the device can automatically adjust the illumination intensity of illumination device 2 to an appropriate value, ensuring that parameters such as brightness and contrast of the captured image are ideal. This effectively avoids issues such as dimming or overexposure caused by insufficient or excessive illumination, further improving the quality of captured images and meeting the user's demand for high-quality photography.

[0032] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A machine learning device, characterized in that: It includes a stage, a lighting device is provided on the upper side of the stage, a camera is provided on the upper side of the lighting device, the lighting device and the camera can be adjusted in the up and down directions, a through hole is provided in the center of the lighting device, and the camera takes pictures through the through hole.

2. A machine learning device according to claim 1, characterized in that: A plurality of light detectors are distributed on the object carrier, and the light detectors can emit electrical signals after being illuminated by light.

3. The machine learning device according to claim 2, wherein: The utility model also includes a central processing unit, and the light detector, the lighting device, and the camera are all connected to the central processing unit.

4. The machine learning device according to claim 3, wherein: The lighting device is in the shape of a ring, and a plurality of light sources are arranged on the lighting device around the through hole.

5. The machine learning device according to claim 4, wherein: It also includes a driving device 1 for driving the lighting device to move up and down, and a driving device 2 for driving the camera to move up and down. Both the driving device 1 and the driving device 2 are controlled by a central processing unit.

6. The machine learning device according to claim 5, wherein: An example pattern is provided on the stage.

7. The machine learning device according to claim 7, wherein: A lens is provided in the through hole.

8. A machine learning method, characterized in that Using the machine learning device according to claim 7 comprises the following steps: Step (1) is to learn how to calculate the projected area of ​​the object. The object is placed on the stage, and the lighting device illuminates the object. The light detectors in the area on the stage that is not blocked by the object are activated by the light and send out signals. In this way, the total number of light detectors on the stage and the number of activated light detectors are known, and the number of unactivated light detectors is calculated. The projected area of ​​the object is calculated based on the number of unactivated light detectors. Step (2) is to learn the camera shooting distance. Use objects with different projection areas to shoot. The same object is shot at different heights by the camera, and the optimal photo is selected. The projection area and shooting distance of the object are stored. By shooting objects with different projection areas and storing the optimal shooting distance, the ratio value A of the projection area of ​​the object to the shooting distance is obtained.

9. A machine learning method according to claim 8, characterized in that: Conduct illumination intensity learning for the lighting device; photograph the sample pattern on the stage, run steps (1) and (2) in sequence, determine the camera shooting distance, adjust different illumination intensities, and take photos; compare the photos taken at different illumination intensities with the sample patterns stored in the central processing unit to obtain the best shooting pattern, and the illumination intensity corresponding to the best shooting pattern is selected as the shooting illumination intensity.

Images