An image processing device based on deep learning

Abstract

The application discloses a kind of image processing equipment based on deep learning, its structure includes storage cabinet, support, placement platform, processing device, box, the processing device of the present application is realized whole horizontal shift by moving wheel and is positioned, and it is matched with vacuum adsorption piece, and it is prevented from shaking by double locking of restraint clamp;Multiple independent processing components are built-in in shell, rely on the adjusting assembly of triangular support structure and cooperate telescopic frame drive image acquisition component vertical lifting, self-adapting adjustment acquisition far and near field of view;Image acquisition component is configured with symmetrical elastic guide structure and universal joint to eliminate deviation, buffer vibration, lens outside is matched with the closed protection spacer of buffer gasket, dust is blocked and lens is prevented from scratching and scratching.The present application is synchronous acquisition in multiple stations, positioning precision is high, imaging quality is stable and uniform, and high-quality standardized image data set can be continuously output, the accuracy of deep learning image recognition and defect detection is greatly improved, and it is suitable for laboratory, industrial quality inspection and other batch image acquisition scenes.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically to an image processing device based on deep learning. Background Technology

[0002] An image processor (EPP) is a device that analyzes, repairs, enhances, and synthesizes images. In image processing, image data needs to be input into the EPP, and then the processed image is output. As the demands on computer video applications increase, high-speed, convenient, and intelligent high-performance digital image processing equipment has become the future direction of video shooting equipment. Areas for improvement when using image processing equipment include: When using the image processing equipment, under normal circumstances, this deep learning-based intelligent image processing equipment can adjust the horizontal angle of the image processing equipment on the movable plate. Through the cooperation of the second hydraulic cylinder, horizontal plate, pulley, and support block, the support block can move horizontally. Through the cooperation of the support block, insert plate, pad, and baffle, it can be fixed according to the size of the image processing equipment, improving the stability of the image processing equipment and further improving the accuracy of image acquisition. When the image processing equipment is in use, the image acquisition component adopts a fixed installation structure, and the acquisition position and acquisition field of view are fixed parameters. It cannot acquire details of the target to be acquired on the stage. When it is necessary to process the target image... When performing high-precision acquisition of fine textures and local features, the device cannot adjust and zoom in for acquisition, making it difficult to capture minute details. When it is necessary to acquire a large area of ​​images or multiple target scenes on the stage, the acquisition field of view cannot be expanded, which greatly reduces the versatility and imaging integrity of the image acquisition component. Furthermore, during the raising and lowering of the image acquisition component, the lens end face has no protective isolation structure. When it descends and approaches the image / workpiece on the stage, the lens is very likely to directly rub and scratch against the image or workpiece surface, causing scratches, wear, and stains on the lens lens. This results in blurred images, color distortion, and feature extraction deviations, directly damaging the quality of the input image of the deep learning model and ultimately reducing the accuracy of image processing, recognition, and detection results. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention is achieved through the following technical solution: an image processing device based on deep learning, the structure of which includes a storage cabinet, legs, a placement platform, a processing device, and a housing. The storage cabinet is located at the bottom of the housing, and a placement platform is installed on the top of the storage cabinet. A processing device is located above the placement platform, and the top of the processing device is snapped into the top of the housing. The bottom of the housing is provided with four legs that are connected to it.

[0004] As a further optimization of the invention, the processing device includes processing components, a movable rod, a moving wheel, an adsorption sheet, a restraining clamp, a docking rod, and a housing. Multiple processing components are provided and are installed parallel to each other inside the bottom of the housing. A movable rod is vertically inserted into the middle of the top of the housing, and docking rods are symmetrically arranged on both sides of the top of the housing. Each docking rod is connected to a restraining clamp by a clamp, and a moving wheel is installed at the upper end of each docking rod. An adsorption sheet is fitted and connected to the top of the movable rod.

[0005] As a further optimization of the invention, the processing component includes an image acquisition component, a lower fixed plate, a telescopic frame, an upper fixed plate, an adjustment component, and a mounting plate. The image acquisition component is installed at the lower end of the lower fixed plate, and the lower fixed plate is parallel to the lower end of the telescopic frame. The upper fixed plate is fitted to the end of the telescopic frame away from the lower fixed plate, and an adjustment component is installed between the upper fixed plate and the lower fixed plate. Mounting plates are fixedly connected to both sides of the adjustment component, and the ends of the two mounting plates away from the adjustment component are connected to the inner walls of the telescopic frame.

[0006] As a further optimization of the invention, the housing is installed in the corresponding position by the support legs, and the image to be processed is placed on the placement table. The movable rod and docking rod on the processing device are connected to the upper part of the housing through the moving wheels and the suction plate. The position of the processing component is adjusted by moving the housing through the moving wheels and the docking rod, so that the processing component can be placed in the position corresponding to the image. Then, the position of the image acquisition component is adjusted by adjusting the component in conjunction with the telescopic frame, so that the image acquisition component can be magnified and reduced according to the image.

[0007] As a further optimization of the invention, a limiting groove is provided at the top of the inner cavity of the housing to accommodate the sliding of the movable wheel. The movable wheel is nested inside the limiting groove and rolls in cooperation with it. The adsorption sheet is a vacuum adsorption structure, with its top end adsorbing and adhering to the top of the inner cavity of the housing, thereby achieving the positioning and fixation of the top of the movable rod. The restraining clamp is an openable and lockable structure, with both ends of the restraining clamp hinged and fixed to the movable rod and the docking rod, respectively. The distance between the movable rod and the docking rod is locked and fixed by the restraining clamp, maintaining the overall stability of the housing during movement.

[0008] As a further optimization of the invention, the movable rod passes through the top of the housing and slides vertically therewith. Multiple processing components are snapped and fixed inside the housing in a corresponding manner, realizing synchronous positioning and acquisition of multiple image processing structures. Furthermore, the spacing between adjacent processing components is consistent, and they work independently without any operational interference.

[0009] As a further optimization of the invention, the two ends of the adjustment component respectively abut against the inner end faces of the upper fixed plate and the lower fixed plate. The mounting plate is stably connected to the telescopic frame. Through the telescopic drive of the adjustment component, in conjunction with the telescopic deformation of the telescopic frame, the lower fixed plate and the image acquisition component at the bottom are vertically displaced, thereby realizing the adaptive magnification and reduction adjustment of the image acquisition focal length and acquisition range.

[0010] As a further optimization of the invention, the adjustment component includes connecting blocks, positioning seats, extension stacks, support plates, and positioning rods. There are four connecting blocks, which are respectively connected to two extension stacks and two support plates. The two extension stacks and two support plates are all triangular in structure, with their ends inserted into the four sides of the positioning seats. The positioning rods are connected through the interior of the two extension stacks, and the positioning rods are connected to the positioning seats through the extension stacks.

[0011] As a further optimization of the invention, both of the extended stacks are internally connected with positioning rods. The two ends of the positioning rods are fixed inside the extended stacks, and the extended stacks can be extended and adjusted to a limited position through the positioning rods. The extension and retraction of the extended stacks, in conjunction with the fixed plate for limiting and guiding, realizes the overall extension and retraction adjustment of the adjustment assembly. The support plate can ensure the structural stability of the adjustment process.

[0012] As a further optimization of the invention, the image acquisition component includes a protective component, guide rods, spring-loaded components, locking components, support rods, and traction rods. The protective component has traction rods symmetrically installed on its left and right sides, and the upper ends of the two traction rods are connected to the support rods. The ends of the two support rods away from the traction rods are connected to the locking components, and spring-loaded components are connected at the angle between the traction rods and the support rods. The lower end of the locking components is connected to two guide rods, and the lower ends of the two guide rods are connected to the protective component.

[0013] As a further optimization of the invention, the locking component is used to securely connect the two support rods, ensuring that the structure is subjected to synchronous force and symmetrical linkage. The guide rod and the pull rod cooperate with the protective component to provide vertical limit guidance during the adjustment process, avoiding misalignment and displacement, effectively improving the structural stability and positioning accuracy of the image acquisition component during adjustment and operation. At the same time, physical protection is achieved by relying on the protective component to ensure stable and reliable image acquisition operation.

[0014] As a further optimization of the invention, the protective component includes a housing, a protective partition, a camera, and a universal joint. The camera is installed in the middle of the housing, and universal joints are provided on both the left and right sides of the camera. The universal joints are connected to the guide rod. The two universal joints are symmetrically arranged on both sides of the housing. The lower end of the housing is provided with a protective partition, which is fitted and connected to the lower end of the camera.

[0015] As a further optimization of the invention, the camera, as the core of image acquisition, is connected to the guide rod through universal joints on both sides, which can accommodate small swings at multiple angles during component adjustment. The bottom protective partition is fitted onto the outside of the camera lens to form a closed protective structure, effectively blocking dust, minor bumps and airflow interference.

[0016] As a further optimization of the invention, the protective partition is provided with an interface, gaskets, and a fitting cover. The interface is located in the middle of the fitting cover, and the fitting cover has a square structure. Four gaskets are provided at its lower end and are connected to it. The four gaskets are provided on the four sides of the interface through the fitting cover.

[0017] As a further optimization of the invention, the fitting cover and the central interface are precisely fitted onto the outside of the camera lens, achieving alignment and protective coverage. Four sets of gaskets at the bottom of the fitting cover form a surrounding buffer structure, which can cushion and reduce shock during device adjustments, vibrations, or minor impacts, reducing lens wear and impact damage, while keeping the lens acquisition area clean, ensuring image clarity and extending the device's lifespan. Beneficial effects

[0018] This invention discloses an image processing device based on deep learning, which has the following beneficial effects: 1. This invention features a horizontally movable adjustment structure consisting of movable wheels, a top limiting slide groove, a docking rod, and a movable rod. This structure allows the housing and all internal processing components to move freely laterally and longitudinally, enabling flexible adjustment of the acquisition points based on the placement of images and workpiece samples on the platform. Combined with a vacuum adsorption plate and a lockable restraining clamp, a dual positioning and locking structure is formed. Simultaneously, multiple independent processing components with uniform spacing are arranged in parallel within the housing, allowing for simultaneous parallel acquisition of multiple images. Each acquisition operation does not obstruct or interfere with the others, significantly improving the efficiency of batch acquisition of deep learning training and detection samples by several times.

[0019] 2. This invention employs a telescopic adjustment assembly consisting of a positioning seat, a triangular structure extension stack, a support plate, and a positioning rod. This assembly, in conjunction with a telescopic frame, links the upper and lower fixed plates, driving the bottom image acquisition component to rise and fall vertically. It allows for free switching between macro magnification acquisition and panoramic wide-angle acquisition modes, significantly improving the accuracy of micro-feature positioning and ensuring the reliability of pixel-level defect detection and dimensional measurement results. The included universal joint connecting the guide rod and the camera housing adaptively releases minor angular stresses caused by lifting and vibration, preventing rigid strain deformation of the support frame. This ensures that the structural accuracy does not diminish over long-term use, resulting in greater equipment durability.

[0020] 3. This invention features an integrated protective partition on the outside of the camera lens, consisting of a housing, an interface, and bottom buffer pads, forming a closed-loop buffer structure. This protects the lens from three dimensions: dustproof, impact-proof, and shock-absorbing. The interface precisely fits tightly around the camera's outer ring, and the housing completely encloses the camera, creating a sealed isolation space that blocks dust, airflow, and stray light from polluting the lens, thus maintaining the lens's light transmittance and color reproduction capabilities over the long term. The bottom buffer pads serve as the first contact buffer layer, preventing the camera from directly rubbing or impacting external objects, ensuring consistent image quality for the deep learning model's input images. Attached Figure Description

[0021] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 A schematic diagram of the structure of an image processing device based on deep learning; Figure 2 This is a schematic diagram of the internal structure of an improved processing device.

[0022] Figure 3 This is a schematic diagram of the internal structure of an improved processing component.

[0023] Figure 4 This is a schematic diagram of the cross-sectional structure of an improved adjustment component.

[0024] Figure 5 This is a schematic diagram of the internal structure of an improved image acquisition component.

[0025] Figure 6 This is a schematic diagram of the internal structure of an improved protective component.

[0026] Figure 7 This is a schematic diagram of a three-dimensional structure of an improved protective partition.

[0027] In the diagram: 1. Locker; 2. Support leg; 3. Placement platform; 4. Processing device; 5. Box body. Processing components - 41, movable rod - 42, moving wheel - 43, adsorption plate - 44, restraint clamp - 45, docking rod - 46, housing - 47; Image acquisition component-411, lower fixing plate-412, telescopic frame-413, upper fixing plate-414, adjustment component-415, mounting plate-416; Connecting block-4151, positioning seat-4152, extension stack-up piece-4153, support plate-4154, positioning rod-4155; Protective components - 4111, guide rod - 4112, spring pull element - 4113, locking element - 4114, support rod - 4115, traction rod - 4116; Housing - 1111, Protective partition - 1112, Camera - 1113, Universal joint - 1114; Interface-1121, gasket-1122, fitting cover-1123. Detailed Implementation

[0028] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0029] Example 1 Please see Figures 1-4 , This invention provides an image processing device based on deep learning. A deep learning-based image processing device includes a storage cabinet 1, legs 2, a placement platform 3, a processing device 4, and a housing 5. The storage cabinet 1 is located at the bottom of the housing 5, and the placement platform 3 is installed on the top of the storage cabinet 1. The processing device 4 is located above the placement platform 3, and the top of the processing device 4 is snapped into the top of the housing 5. The bottom of the housing 5 is provided with four legs 2 connected to it.

[0030] The processing device 4 includes a processing component 41, a movable rod 42, a moving wheel 43, an adsorption sheet 44, a restraining clamp 45, a docking rod 46, and a housing 47. Multiple processing components 41 are provided and are installed in parallel at the bottom of the housing 47. A movable rod 42 is vertically inserted into the middle of the top of the housing 47, and docking rods 46 are symmetrically arranged on both sides of the top of the housing 47. Each docking rod 46 is connected to a restraining clamp 45, and a moving wheel 43 is installed at the upper end of each docking rod 46. An adsorption sheet 44 is fitted and connected to the top of the movable rod 42.

[0031] The processing component 41 includes an image acquisition component 411, a lower fixing plate 412, a telescopic frame 413, an upper fixing plate 414, an adjustment component 415, and a mounting plate 416. The image acquisition component 411 is installed at the lower end of the lower fixing plate 412, and the lower fixing plate 412 is parallel to the lower end of the telescopic frame 413. The upper fixing plate 414 is connected to the end of the telescopic frame 413 away from the lower fixing plate 412, and the adjustment component 415 is installed between the upper fixing plate 414 and the lower fixing plate 412. Mounting plates 416 are fixedly connected to both sides of the adjustment component 415, and the ends of the two mounting plates 416 away from the adjustment component 415 are connected to the inner walls of the telescopic frame 413.

[0032] The housing 5 is installed in the corresponding position by the support legs 2. The image to be processed is placed on the placement platform 3. The movable rod 42 and docking rod 46 on the processing device 4 are connected to the upper part of the housing 5 through the moving wheel 43 and the suction plate 44. The moving wheel 43 and the docking rod 46 drive the housing 47 to move to adjust the position of the processing component 41, so that the processing component 41 can be placed in the position corresponding to the image. Then, the position of the image acquisition component 411 is adjusted by adjusting the component 415 in cooperation with the telescopic frame 413, so that the image acquisition component 411 can be magnified and reduced according to the image.

[0033] The top of the inner cavity of the housing 5 is provided with a limiting groove adapted to the sliding of the movable wheel 43. The movable wheel 43 is nested and installed inside the limiting groove and rolls with it. The adsorption sheet 44 is a vacuum adsorption structure. The top of the adsorption sheet 44 is adsorbed and attached to the top of the inner cavity of the housing 5 to achieve positioning and fixation of the top of the movable rod 42. The restraining clamp 45 is an openable and lockable structure. The two ends of the restraining clamp 45 are respectively hinged and fixed to the movable rod 42 and the docking rod 46. The distance between the movable rod 42 and the docking rod 46 is locked by the restraining clamp 45 to maintain the overall stability of the housing 47 during movement.

[0034] The movable rod 42 passes through the top of the housing 47 and slides vertically therewith. Multiple processing components 41 are snapped and fixed inside the housing 47 in a one-to-one correspondence, realizing synchronous positioning and acquisition of multiple image processing structures. The spacing between adjacent processing components 41 is consistent, and they work independently without any operational interference.

[0035] The two ends of the adjustment component 415 respectively abut against the inner end faces of the upper fixed plate 414 and the lower fixed plate 412. The mounting plate 416 is stably connected to the telescopic frame 413. Through the telescopic drive of the adjustment component 415, in conjunction with the telescopic deformation of the telescopic frame 413, the lower fixed plate 412 and the image acquisition component 411 at the bottom are vertically displaced, thereby realizing the adaptive magnification and reduction adjustment of the image acquisition focal length and acquisition range.

[0036] The adjustment assembly 415 includes a connecting block 4151, a positioning seat 4152, an extension stack 4153, a support plate 4154, and a positioning rod 4155. There are four connecting blocks 4151, which are respectively connected to two extension stacks 4153 and two support plates 4154. The two extension stacks 4153 and the two support plates 4154 are all triangular in structure, and their ends are inserted into the four sides of the positioning seat 4152. The positioning rod 4155 is connected through the two extension stacks 4153, and the positioning rod 4155 is connected to the positioning seat 4152 through the extension stacks 4153.

[0037] Both of the extended stacks 4153 have a positioning rod 4155 connected through them. The two ends of the positioning rod 4155 are fixed inside the extended stack 4153, and the extended stack 4153 can be extended and adjusted to a limited position through the positioning rod 4155. The extension and retraction of the extended stack 4153 cooperates with the fixed plate 412 for limiting and guiding, so as to realize the extension and retraction adjustment of the entire adjustment component 415. The support plate 4154 can ensure the structural stability of the adjustment process.

[0038] The working principle of the above technical solution is explained below: In use, the equipment is positioned and prepared for loading by four sets of support legs 2 at the bottom of the housing 5. The height of the support legs is adjusted to keep the housing level. The image to be tested, the workpiece, and the sample are laid flat on the placement platform 3 inside the housing. The storage cabinet 1 at the bottom of the housing can store cleaning tools, spare lenses, and wiring accessories without occupying the acquisition work space. The top of the housing has a pre-fabricated through-type limiting slide groove, which provides a horizontal moving track for the processing device 4. In the initial state of horizontal movement and alignment of the processing device 4, the restraint clamp 45 is in the released and unlocked state, and the movable rod 42 and the two side docking rods 46 have no rigid constraints. The moving wheel 43 at the top of the docking rod 46 is embedded in the limiting slide groove at the top of the housing, driving the moving wheel at the top of the docking rod 46 to roll horizontally and vertically along the slide groove. Simultaneously, the lower housing 47 is moved horizontally, and multiple parallel processing components 41 inside the housing move in sync. Operators can precisely align each processing component 41 with its corresponding acquisition area based on the distribution of multiple target samples on the placement platform 3. The multiple processing components are uniformly spaced and independent, allowing for the simultaneous acquisition of multiple image samples without obstruction or operational interference, significantly improving the efficiency of batch deep learning sample acquisition. After the housing 47 is horizontally adjusted, the adsorption plate 44 at the top of the movable rod 42 tightly adheres to the inner top surface of the housing 5, and the two ends of the restraint clamp 45 respectively grip the movable rod 42 and the docking rod 46, locking the relative distance between the three and constraining the left-right and front-back offset of the housing 47 to ensure stable imaging. The adjustment component drives vertical lifting. The adaptive zoom acquisition rod 42 slides vertically with the top of the housing 47. Each processing component 41 inside the housing has a built-in independent adjustment component 415 and a telescopic frame 413. The positioning rod 4155 inside the adjustment component 415 horizontally passes through two sets of extension stacks 4153, serving as a telescopic guide reference. This drives the extension stacks 4153 to extend outwards along the positioning rod 4155, increasing their overall height. The two ends of the extension stacks 4153 push against the upper fixing plate 414 and the lower fixing plate 412 respectively via connecting blocks 4151, causing the upper and lower fixing plates to move away from each other. Simultaneously, the telescopic frame 413 extends downwards. The lower fixing plate 412 causes the bottom image acquisition component 411 to descend, shortening the vertical distance between the camera and the sample image, thus magnifying image texture and minor defects. Detailed features meet the high-precision feature extraction requirements of deep learning models; the two triangular support plates 4154 open synchronously with the extended stack to form a stable triangular support structure, offsetting the force on one side and preventing structural tilting during the lifting process. The extended stack 4153 is controlled to fold and shrink inward along the positioning rod 4155, shortening the overall height. The upper and lower fixed plates are pulled closer to each other, the telescopic frame 413 folds and retracts upward, and the image acquisition component 411 is raised synchronously to increase the distance between the camera and the sample, expand the shooting field of view, and capture the entire image and a large-scale workpiece image at once. The high-definition image acquired by the camera is transmitted to the back-end deep learning computing module in real time to complete the algorithm processing such as image classification, defect detection, feature recognition, and image repair.Single or multiple processing components can simultaneously output multiple channels of image data, enabling parallel processing and improving overall processing speed.

[0039] Example 2 Please see Figures 5-7 , This invention provides an image processing device based on deep learning. The image acquisition component 411 includes a protective component 4111, guide rods 4112, spring-loaded components 4113, locking components 4114, support rods 4115, and traction rods 4116. The protective component 4111 has symmetrically installed traction rods 4116 on its left and right sides, and the upper ends of the two traction rods 4116 are connected to the support rods 4115. The ends of the two support rods 4115 away from the traction rods 4116 are connected to the locking components 4114. Spring-loaded components 4113 are connected at the angle between the traction rods 4116 and the support rods 4115. The lower end of the locking components 4114 is connected to two guide rods 4112, and the lower ends of the two guide rods 4112 are connected to the protective component 4111.

[0040] The locking component 4114 is used to securely connect the two side support rods 4115, ensuring that the structure is subjected to synchronous force and symmetrical linkage. The guide rod 4112 and the pull rod 4116 cooperate with the protective component 4111 to restrain and guide the adjustment process vertically, avoid misalignment and displacement, effectively improve the structural stability and positioning accuracy of the image acquisition component during adjustment and operation, and at the same time rely on the protective component to achieve physical protection, ensuring the stable and reliable image acquisition operation.

[0041] The protective component 4111 includes a housing 1111, a protective partition 1112, a camera 1113, and a universal joint 1114. The camera 1113 is installed in the middle of the housing 1111, and universal joints 1114 are provided on both the left and right sides of the camera 1113. The universal joints 1114 are connected to the guide rod 4112. The two universal joints 1114 are symmetrically arranged on both sides of the inside of the housing 1111. The protective partition 1112 is provided at the lower end of the housing 1111, and the protective partition 1112 is fitted and connected to the lower end of the camera 1113.

[0042] The camera 1113 serves as the core of image acquisition. It is connected to the guide rod 4112 via universal joints 1114 on both sides, which can accommodate small swings at multiple angles during component adjustment. The bottom protective partition 1112 is fitted onto the outside of the lens of the camera 1113 to form a closed protective structure, effectively blocking dust, minor bumps and airflow interference.

[0043] The protective partition 1112 is provided with an interface 1121, a gasket 1122, and a fitting cover 1123. The interface 1121 is located in the middle of the interior of the fitting cover 1123, and the fitting cover 1123 has a square structure. Four gaskets 1122 are provided at its lower end and are connected to it. The four gaskets 1122 are provided on the four sides of the interface 1121 through the fitting cover 1123.

[0044] The fitting cover 1123 and the central interface 1121 are precisely fitted onto the outside of the lens of the camera 1113, achieving alignment and protective coverage. The four sets of gaskets 1122 at the bottom of the fitting cover 1123 form a surrounding buffer structure, which can buffer and reduce shock during equipment fine-tuning, vibration, or slight bumps, reducing lens wear and impact damage, while keeping the lens acquisition area clean, ensuring image acquisition clarity and equipment lifespan.

[0045] The working principle of the above technical solution is explained below: In use, the image acquisition component 411 is mounted on the lower end of the lower fixed plate 412 and moves synchronously with the adjustment component 415. The locking component 4114 is symmetrically hinged to two support rods 4115. The lower ends of the support rods 4115 are connected to diagonally arranged pull rods 4116, and the bottom ends of the pull rods 4116 are connected to the outer shell of the protective component 4111. A spring-loaded element 4113 is installed at the angle between the support rods 4115 and the pull rods 4116, continuously providing elastic tension to ensure that the left and right rods always maintain synchronous opening and closing. The locking component 4114 rigidly connects the two support rods 4115, ensuring that the displacement of the left and right rods is fully controlled during lifting and lowering. In accordance with the design, two vertical guide rods 4112 are vertically connected to the locking component and the protective assembly shell, ensuring that the camera 1113 inside the protective assembly 4111 is always vertically aligned with the sample below, improving the accuracy of data acquisition and positioning. Universal joints 1114 are symmetrically mounted on the left and right sides of the camera 1113 inside the protective assembly 4111. The outer side of the universal joint is hinged to the lower end of the guide rod 4112. When the adjustment component 415 extends or retracts, the equipment vibrates slightly, or there is a slight height difference between the sample on the placement platform, the guide rod 4112 will generate a slight angular pulling force; the universal joint 1114 can realize small-amplitude, multi-angle adaptive swing of the camera 1113, releasing the rod and the camera. The mechanical internal stress between the images 1113 ensures that the lens acquisition surface of the camera 1113 remains parallel to the sample plane. A protective spacer 1112 is fitted to the bottom of the lens on the camera 1113. The interface 1121 at the center of the protective spacer 1112 precisely fits onto the outer ring of the lens on the camera 1113. The housing 1123 completely encloses the lens, forming a closed isolation cavity that effectively prevents external dust, airflow, and debris from adhering to the lens of the camera 1113. Four independent gaskets 1122, made of soft cushioning material, are integrally connected to the bottom four sides of the housing 1123. If the image acquisition component descends too close... When the equipment experiences bumps or vibrations, or when the camera 1113 is moved down due to human error, the pad 1122 will preferentially contact the surface of the image and workpiece below, preventing the camera 1113 from directly rubbing or impacting the sample. This eliminates the problem of lens scratches, chipping, and wear. The protective partition 1112 completely covers the lens. After the acquisition is completed and the camera 1113 is raised, the protective partition 1112 remains on the outside of the lens of the camera 1113. When the machine is stopped and stored, it continuously protects the lens of the camera 1113, extends the service life of the camera, and continuously provides scratch-free, high-definition, and color-uniform original image samples for deep learning algorithms.

[0046] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the scope of the invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0047] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A deep learning-based image processing device, comprising a storage cabinet (1), legs (2), a placement platform (3), a processing device (4), and a housing (5), wherein the storage cabinet (1) is located at the bottom of the housing (5), and a placement platform (3) is installed at the top of the storage cabinet (1), the processing device (4) is located above the placement platform (3), and the top of the processing device (4) is engaged with the top of the housing (5), and the bottom of the housing (5) is provided with four legs (2) connected thereto, characterized in that: The processing device (4) is provided with a processing component (41), a movable rod (42), a moving wheel (43), an adsorption sheet (44), a restraining clamp (45), a docking rod (46), and a housing (47). There are multiple processing components (41), and multiple processing components (41) are installed in parallel at the bottom of the housing (47). A movable rod (42) is vertically inserted into the middle of the top of the housing (47), and docking rods (46) are symmetrically arranged on both sides of the top of the housing (47). A restraining clamp (45) is connected between each docking rod (46) and the movable rod (42), and a moving wheel (43) is installed on the upper end of each docking rod (46). An adsorption sheet (44) is connected to the top of the movable rod (42). The processing component (41) includes an image acquisition component (411), a lower fixing plate (412), a telescopic frame (413), an upper fixing plate (414), an adjustment component (415), and a mounting plate (416). The image acquisition component (411) is installed at the lower end of the lower fixing plate (412), and the lower fixing plate (412) is parallel to the lower end of the telescopic frame (413). The upper fixing plate (414) is connected to the end of the telescopic frame (413) away from the lower fixing plate (412), and an adjustment component (415) is installed between the upper fixing plate (414) and the lower fixing plate (412). Mounting plates (416) are fixedly connected to both sides of the adjustment component (415), and the ends of the two mounting plates (416) away from the adjustment component (415) are connected to the inner walls of the telescopic frame (413). The housing (5) is installed in the corresponding position by the support legs (2). The image to be processed is placed on the placement platform (3). The movable rod (42) and docking rod (46) on the processing device (4) are connected to the upper part of the housing (5) through the moving wheel (43) and the suction plate (44). The position of the processing component (41) is adjusted by moving the housing (47) through the moving wheel (43) and docking rod (46), so that the processing component (41) can be placed in the position corresponding to the image. Then, the position of the image acquisition component (411) is adjusted by adjusting the component (415) and the telescopic frame (413), so that the image acquisition component (411) can be magnified and reduced according to the image.

2. The image processing device based on deep learning according to claim 1, characterized in that: The top of the inner part of the housing (5) is provided with a limiting groove adapted to the sliding of the movable wheel (43). The movable wheel (43) is nested and installed inside the limiting groove and rolls with it. The adsorption sheet (44) is a vacuum adsorption structure. The top of the adsorption sheet (44) is adsorbed and attached to the top of the inner part of the housing (5) to realize the positioning and fixation of the top of the movable rod (42). The restraining clamp (45) is an openable and lockable structure. The two ends of the restraining clamp (45) are respectively hinged and fixed to the movable rod (42) and the docking rod (46). The distance between the movable rod (42) and the docking rod (46) is locked and fixed by the restraining clamp (45) to maintain the overall stability of the housing (47) during the movement process.

3. The image processing device based on deep learning according to claim 1, characterized in that: The movable rod (42) passes through the top of the housing (47) and slides vertically therewith. Multiple processing components (41) are fixed inside the housing (47) in a corresponding manner, realizing synchronous positioning and acquisition of multiple image processing structures. The spacing between adjacent processing components (41) is consistent, and they work independently without interference.

4. The image processing device based on deep learning according to claim 1, characterized in that: The two ends of the adjustment component (415) respectively abut against the inner end face of the upper fixed plate (414) and the lower fixed plate (412). The mounting plate (416) is stably connected to the telescopic frame (413). Through the telescopic drive of the adjustment component (415), in conjunction with the telescopic deformation of the telescopic frame (413), the lower fixed plate (412) and the image acquisition component (411) at the bottom are vertically displaced, thereby realizing the adaptive magnification and reduction adjustment of the image acquisition focal length and acquisition range.

5. The image processing device based on deep learning according to claim 1, characterized in that: The adjustment assembly (415) includes a connecting block (4151), a positioning seat (4152), an extension stack (4153), a support plate (4154), and a positioning rod (4155). There are four connecting blocks (4151), which are respectively connected to two extension stacks (4153) and two support plates (4154). The two extension stacks (4153) and the two support plates (4154) are all triangular in structure, and their ends are inserted into the four sides of the positioning seat (4152). The positioning rod (4155) is connected through the interior of the two extension stacks (4153), and the positioning rod (4155) is connected to the positioning seat (4152) through the extension stacks (4153). Both of the extended stacks (4153) are internally connected with positioning rods (4155). The two ends of the positioning rods (4155) are fixed inside the extended stacks (4153). The extended stacks (4153) can be extended and adjusted to a limited position through the positioning rods (4155). The extension and retraction of the extended stacks (4153) cooperates with the fixed plate (412) for limiting and guiding, so as to realize the overall extension and retraction adjustment of the adjustment component (415). The support plate (4154) can ensure the structural stability of the adjustment process.

6. The image processing device based on deep learning according to claim 1, characterized in that: The image acquisition component (411) is provided with a protective component (4111), guide rods (4112), spring pullers (4113), locking components (4114), support rods (4115), and pull rods (4116). Pull rods (4116) are symmetrically installed on the left and right sides of the protective component (4111), and the upper ends of the two pull rods (4116) are connected to the support rods (4115). The ends of the two support rods (4115) away from the pull rods (4116) are connected to the locking components (4114), and spring pullers (4113) are connected at the angle between the pull rods (4116) and the support rods (4115). The lower end of the locking components (4114) is connected to two guide rods (4112), and the lower ends of the two guide rods (4112) are connected to the protective component (4111). The locking component (4114) is used to securely connect the two side support rods (4115), ensuring that the structure is subjected to synchronous force and symmetrical linkage. The guide rod (4112) and the pull rod (4116) cooperate with the protective component (4111) to provide vertical limit guidance for the adjustment process, avoid misalignment and displacement, effectively improve the structural stability and positioning accuracy of the image acquisition component during adjustment and operation, and at the same time rely on the protective component to achieve physical protection, ensuring the stable and reliable image acquisition operation.

7. The image processing device based on deep learning according to claim 6, characterized in that: The protective component (4111) includes a housing (1111), a protective partition (1112), a camera (1113), and a universal joint (1114). The camera (1113) is installed in the middle of the housing (1111), and universal joints (1114) are provided on both the left and right sides of the camera (1113). The universal joints (1114) are connected to the guide rod (4112). The two universal joints (1114) are symmetrically arranged on both sides of the inside of the housing (1111), and the lower end of the housing (1111) is provided with a protective partition (1112). The protective partition (1112) is fitted and connected to the lower end of the camera (1113). The camera (1113) serves as the core of image acquisition. It is connected to the guide rod (4112) via the universal joints (1114) on both sides. It can accommodate small swings at multiple angles during component adjustment. The bottom protective partition (1112) is fitted onto the outside of the camera (1113) lens to form a closed protective structure, effectively blocking dust, minor bumps and airflow interference.

8. The image processing device based on deep learning according to claim 7, characterized in that: The protective partition (1112) is provided with an interface (1121), a gasket (1122), and a fitting cover (1123). The interface (1121) is located in the middle of the interior of the fitting cover (1123), and the fitting cover (1123) is square in shape. Four gaskets (1122) are provided at its lower end and are connected to it. The four gaskets (1122) are provided on the four sides of the interface (1121) through the fitting cover (1123). The sleeve cover (1123) and the middle interface (1121) are precisely fitted onto the outside of the camera lens (1113) to achieve alignment and protective coverage. The four sets of pads (1122) at the bottom of the sleeve cover (1123) form a surrounding buffer structure, which can buffer and reduce shock when the equipment is finely adjusted, vibrated, or slightly bumped, reducing lens wear and impact damage, while keeping the lens acquisition area clean, ensuring image acquisition clarity and equipment lifespan.

Images