Image acquisition device for object recognition deep learning

Abstract

The utility model relates to the technical field of image acquisition, in particular to an image acquisition device for object recognition deep learning, which at least comprises a fixed base, an image acquisition component is arranged on the surface of the fixed base, the image acquisition component is used for performing image acquisition on an object, and meanwhile, the image acquisition component moves in a shooting process to acquire the image of the object. An object can be shot at multiple angles, an image of the bottom of the object is collected in the object placing process, and a movable rotating assembly is arranged on the surface of the fixed base and used for driving the object to transversely rotate; the first motor works to drive the first threaded rod to rotate, the first threaded rod rotates to drive the moving block to vertically move, the moving block moves to drive the first camera to move, and the first camera moves to change the collection range of object recognition, so that the device is suitable for image collection of object recognition of different sizes.

Description

Technical Field

[0001] The utility model relates to the technical field of image acquisition, and more specifically to an image acquisition device for deep learning of object recognition. Background Art

[0002] Deep learning is a multi-disciplinary subject involving probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory, etc. It is a special study of how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. It is the core of artificial intelligence and the fundamental way to make computers intelligent.

[0003] There are many existing technologies for image acquisition, such as:

[0004] Chinese patent publication number CN213211054U discloses an image acquisition device based on machine learning, including a tray, on which an object to be collected is placed, and then a first motor is controlled to operate to cause a horizontal plate and a collection tube to descend, and the collection tube descends to cover the object, thereby eliminating the influence of the external environment on image acquisition and making the acquisition more accurate. The third camera component can capture the bottom image of the object through the tray made of transparent glass, and the second camera component can capture the top image of the object. When the collection tube rotates, the first camera component can capture images around the object, thereby achieving the effect of being able to capture images of the object in multiple directions and in all directions, making it easy to obtain a full picture of the object. The utility model has a novel design, a simple structure, and is easy to operate. It has the advantages of being able to capture images of the object in multiple directions and in all directions.

[0005] Currently, in deep learning, it is often necessary to collect images of objects. Traditional image acquisition devices for machine learning have some shortcomings. For example, ordinary image acquisition devices can only collect images of objects at fixed angles or directions, and the objects need to be flipped during the acquisition process. The acquisition of the entire object is completed through secondary acquisition, which is rather troublesome. It is difficult to construct an accurate 3D model of the object with a single acquisition, which brings certain obstacles to deep learning and reduces learning efficiency.

[0006] In view of this, the utility model provides an image acquisition device for deep learning of object recognition. Utility Model Content

[0007] The utility model provides an image acquisition device for deep learning of object recognition, which at least includes a fixed base, and an image acquisition component is provided on the surface of the fixed base. The image acquisition component is used to acquire images of objects. At the same time, the image acquisition component can move during the shooting process to shoot objects from multiple angles, and acquire the image of the bottom of the object during the placement of the object. The surface of the fixed base is provided with a mobile rotating component, and the mobile rotating component is used to drive the object to rotate horizontally.

[0008] As a further improvement of the present technical solution, the image acquisition assembly includes a first threaded rod, the first threaded rod is rotatably connected to a fixed base, and a first motor for driving the first threaded rod to rotate is installed on the surface of the fixed base.

[0009] As a further improvement of the technical solution, a moving block is threadedly connected to the surface of the first threaded rod, and the moving block is slidably connected to the fixed base. The moving block is driven to move vertically by rotating the first threaded rod. A first camera is provided on the surface of the moving block, and a second camera is installed inside the fixed base.

[0010] As a further improvement of the present technical solution, the movable rotating assembly includes a second threaded rod, the second threaded rod is rotatably connected to the fixed base, and a second motor for driving the second threaded rod to rotate is installed on the surface of the fixed base.

[0011] As a further improvement of the technical solution, a support rod is threadedly connected to the surface of the second threaded rod, and the support rod is slidably connected to the fixed base, and the support rod is driven to move laterally by the rotation of the second threaded rod.

[0012] As a further improvement of the technical solution, the support rod surface is rotatably connected with a rotating block, the top of the fixed base is rotatably connected with a connecting disk, an elastic member is installed inside the connecting disk, and the other end of the elastic member is connected to a rubber block.

[0013] As a further improvement of the technical solution, a driving gear is meshed at the bottom of the connecting disk, and a third motor for driving the driving gear to rotate is installed on the surface of the fixed base.

[0014] Compared with the prior art, the utility model has the following beneficial effects:

[0015] 1. In the image acquisition device for deep learning of object recognition, the first motor drives the first threaded rod to rotate, the rotation of the first threaded rod drives the moving block to move vertically, the movement of the moving block drives the first camera to move, the movement of the first camera changes the acquisition range of object recognition, so that it is suitable for image acquisition of objects of different volumes, the driving gear rotates to rotate the connecting disk, the rotation of the connecting disk drives the object to rotate, so that the first camera can capture all-round images of the object except the bottom, thereby improving the image acquisition efficiency of the device for object recognition.

[0016] 2. In the image acquisition device for deep learning of object recognition, the second camera does not stop working. The bottom of the object is not blocked during placement, so the second camera can capture the bottom of the object and use the elastic force generated by the elastic member to push the rubber block upward, so that there is a gap between the object and the bottom of the connecting plate, so that the second camera has more time to recognize and process the data, avoiding the need for a second flip to complete the acquisition of the object. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The present invention is described in more detail below by way of example with reference to the accompanying drawings, in which:

[0018] Figure 1 It is a schematic diagram of the overall structure of the utility model;

[0019] Figure 2 It is a schematic diagram of the connection disk of the utility model;

[0020] Figure 3 is a schematic diagram of the second camera of the present utility model;

[0021] Figure 4 It is a schematic diagram of the driving gear of the utility model.

[0022] The meaning of each number in the figure is:

[0023] 11.Fix the base;

[0024] 20. Image acquisition component; 21. First threaded rod; 22. First motor; 23. Moving block; 24. First camera; 25. Second camera;

[0025] 30. Mobile rotating assembly; 31. Second threaded rod; 32. Second motor; 33. Support rod; 34. Rotating block; 35. Third motor; 36. Driving gear; 37. Connecting plate; 38. Elastic member; 39. Rubber block. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the utility model to clearly and completely describe the technical solutions in the embodiments of the utility model. Obviously, the described embodiments are only part of the embodiments of the utility model, not all of the embodiments. Based on the embodiments of the utility model, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the utility model.

[0027] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present invention.

[0028] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present utility model, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0029] Deep learning is a multi-disciplinary subject involving probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory, etc. It is a special study of how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. It is the core of artificial intelligence and the fundamental way to make computers intelligent.

[0030] like Figure 1-4 As shown, the device includes at least a fixed base 11, and an image acquisition component 20 is provided on the surface of the fixed base 11. The image acquisition component 20 is used to acquire images of the object. At the same time, the image acquisition component 20 can move during the shooting process to shoot the object at multiple angles, and acquire the image of the bottom of the object during the placement of the object. A mobile rotating component 30 is provided on the surface of the fixed base 11, and the mobile rotating component 30 is used to drive the object to rotate horizontally.

[0031] In deep learning, it is often necessary to collect images of objects. Traditional image acquisition devices for machine learning have some shortcomings. For example, ordinary image acquisition devices can only collect images of objects at fixed angles or fixed directions. Therefore, the image acquisition component 20 includes a first threaded rod 21, which is rotatably connected to the fixed base 11. A first motor 22 for driving the first threaded rod 21 to rotate is installed on the surface of the fixed base 11. A moving block 23 is threadedly connected to the surface of the first threaded rod 21. The moving block 23 is slidably connected to the fixed base 11. The moving block 23 is driven to move vertically by the rotation of the first threaded rod 21. A first camera 24 is provided on the surface of the moving block 23, and a second camera 25 is installed inside the fixed base 11.

[0032] Depend on Figure 1-Figure 3 As shown, when it is necessary to collect images for object recognition, the object is placed on the surface of the connection plate 37, and then the first motor 22 is started to work according to the volume of the object. The first motor 22 drives the first threaded rod 21 to rotate, and the rotation of the first threaded rod 21 drives the moving block 23 to move vertically. The movement of the moving block 23 drives the first camera 24 to move. The movement of the first camera 24 changes the collection range of object recognition, so that it is suitable for image collection of objects of different volumes.

[0033] Data acquisition: The camera captures image data through the lens.

[0034] Data preprocessing: Before feeding images to a deep learning model, the data usually needs to be preprocessed. This may include image resizing, color space conversion, noise removal, image enhancement, etc. to ensure that the data is suitable for model input.

[0035] Feature extraction: In deep learning models, structures such as convolutional neural networks (CNNs) are often used to extract features from images. These features can capture various information in the image, such as edges, textures, shapes, etc.

[0036] Feature recognition: The extracted features are passed to the deep learning model, which uses these features to recognize objects, faces, actions, etc. in the image. This may involve different tasks such as classification tasks, object detection, instance segmentation, etc.

[0037] Output results: The model generates corresponding output results based on the content of the input image. These results may be the object category, location information, action classification, etc.

[0038] Feedback and Adjustment: Based on the output results, feedback and adjustments can be made to the model to improve its performance and accuracy. This may involve adjusting the model's parameters, retraining the model, collecting more data, etc.

[0039] During the acquisition process, the object needs to be turned over, and the acquisition of the whole object is completed through secondary acquisition, which is rather troublesome. It is difficult to build an accurate 3D model of the object through a single acquisition, which brings certain obstacles to deep learning and reduces the learning efficiency. Therefore, the mobile rotating component 30 includes a second threaded rod 31, the second threaded rod 31 is rotatably connected to the fixed base 11, and a second motor 32 for driving the second threaded rod 31 to rotate is installed on the surface of the fixed base 11. A support rod 33 is threadedly connected to the surface of the second threaded rod 31, and the support rod 33 is slidably connected to the fixed base 11, and the support rod 33 is driven to move laterally by the rotation of the second threaded rod 31;

[0040] The surface of the support rod 33 is rotatably connected with a rotating block 34, the top of the fixed base 11 is rotatably connected with a connecting plate 37, an elastic member 38 is installed inside the connecting plate 37, the other end of the elastic member 38 is connected with a rubber block 39, the bottom of the connecting plate 37 is meshed with a driving gear 36, and the surface of the fixed base 11 is installed with a third motor 35 for driving the driving gear 36 to rotate;

[0041] Depend on Figure 2-Figure 4 As shown, when the object is placed on the surface of the connection plate 37, the second camera 25 does not stop working, and the bottom of the object is not blocked during placement, so the second camera 25 collects the bottom of the object, and the elastic member 38 generates elastic force to push the rubber block 39 upward, so that there is a gap between the object and the bottom of the connection plate 37, so that the second camera 25 has more time to identify and process the data, avoiding the need for a second flip to complete the collection of the object;

[0042] At the same time, according to the volume of the object, the second motor 32 is started to work, and the second motor 32 drives the second threaded rod 31 to rotate. The rotation of the second threaded rod 31 causes the support rod 33 to move. The movement of the support rod 33 drives the rotating block 34 to move, so that the rotating block 34 fits the side of the object to prevent the object from falling during rotation. Then the third motor 35 is started to work, and the third motor 35 drives the driving gear 36 to rotate. The rotation of the driving gear 36 causes the connecting disk 37 to rotate. The rotation of the connecting disk 37 drives the object to rotate, so that the first camera 24 can collect all-round images of the object except the bottom, thereby improving the image collection efficiency of the device for object recognition.

[0043] The above shows and describes the basic principle, main features and advantages of the utility model. Those skilled in the art should understand that the utility model is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the utility model and are not used to limit the utility model. Without departing from the spirit and scope of the utility model, the utility model may have various changes and improvements, which fall within the scope of the utility model to be protected. The scope of protection of the utility model is defined by the attached claims and their equivalents.

Claims

1. An image acquisition device for deep learning of object recognition, characterized in that: The invention comprises at least one fixed base (11), wherein an image acquisition component (20) is provided on the surface of the fixed base (11), wherein the image acquisition component (20) is used to acquire images of an object, and wherein the image acquisition component (20) can be moved during the shooting process to shoot the object at multiple angles, and can acquire an image of the bottom of the object during the object placement process, and wherein a moving rotation component (30) is provided on the surface of the fixed base (11), wherein the moving rotation component (30) is used to drive the object to rotate horizontally.

2. The image acquisition device for deep learning of object recognition according to claim 1, characterized in that: The image acquisition assembly (20) comprises a first threaded rod (21), the first threaded rod (21) being rotatably connected to a fixed base (11), and a first motor (22) for driving the first threaded rod (21) to rotate is mounted on a surface of the fixed base (11).

3. The image acquisition device for deep learning of object recognition according to claim 2, characterized in that: A moving block (23) is threadedly connected to the surface of the first threaded rod (21); the moving block (23) is slidably connected to the fixed base (11); the moving block (23) is driven to move vertically by the rotation of the first threaded rod (21); a first camera (24) is provided on the surface of the moving block (23); and a second camera (25) is installed inside the fixed base (11).

4. The image acquisition device for deep learning of object recognition according to claim 1, characterized in that: The movable rotating assembly (30) comprises a second threaded rod (31), the second threaded rod (31) is rotatably connected to a fixed base (11), and a second motor (32) for driving the second threaded rod (31) to rotate is mounted on the surface of the fixed base (11).

5. The image acquisition device for deep learning of object recognition according to claim 4, characterized in that: A support rod (33) is threadedly connected to the surface of the second threaded rod (31), and the support rod (33) is slidably connected to the fixed base (11), and the support rod (33) is driven to move laterally by the rotation of the second threaded rod (31).

6. The image acquisition device for deep learning of object recognition according to claim 5, characterized in that: The surface of the support rod (33) is rotatably connected to a rotating block (34), the top of the fixed base (11) is rotatably connected to a connecting disk (37), an elastic member (38) is installed inside the connecting disk (37), and the other end of the elastic member (38) is connected to a rubber block (39).

7. The image acquisition device for deep learning of object recognition according to claim 6, characterized in that: A driving gear (36) is meshed at the bottom of the connecting plate (37), and a third motor (35) for driving the driving gear (36) to rotate is installed on the surface of the fixed base (11).

Images