Laser image feature recognition instrument based on convolutional network model

Abstract

The utility model belongs to the technical field of image recognition, and particularly relates to a laser image feature recognition instrument based on a convolutional network model, which comprises two support plates, a sorting plate arranged between the two support plates, a laser image acquisition device arranged above the sorting plate, a foreign matter cleaning execution assembly arranged on the surface of the support plates, and a laser image recognition device arranged on the surface of the foreign matter cleaning execution assembly. The foreign matter cleaning execution assembly comprises a U-shaped support arranged above the supporting plate and two side edge plates fixedly connected to the bottom face of the U-shaped support. The multiple motors, the gears and the toothed plates are matched, so that the clamping hopper accurately moves to the position above the foreign matter to be positioned, the multiple motors, the sliding blocks and the mechanical clamping assembly cooperate to rapidly clamp and transfer the foreign matter, and the pumpkin seed sorting purity is improved.

Description

Technical Field

[0001] This invention belongs to the field of image recognition technology, specifically relating to a laser image feature recognition instrument based on a convolutional network model. Background Technology

[0002] The laser image feature recognition instrument utilizes the reflection and scattering characteristics generated after laser interacts with an object to acquire laser images containing information about the object's surface. The device's built-in convolutional network model processes these images. First, by sliding the convolutional kernels in the convolutional layer, local features in the laser image, such as the shape and texture of the light spot, are extracted. Then, the pooling layer downsamples the feature map, reducing the amount of data while retaining key features. Finally, the fully connected layer classifies and identifies the processed features to determine the object's category, state, and other information. The entire process is based on deep learning principles such as convolutional operations and weight sharing, achieving efficient and accurate analysis of laser images.

[0003] According to the public announcement (CN222469762U), a sorting machine for pumpkin seed processing is disclosed. This technology discloses "a technical solution including two support plates, on which a vibration mechanism and a reset component are provided; and a sorting plate is fixedly connected between the two support plates, which solves the problem that existing sorting machines for pumpkin seed processing are difficult to achieve accurate grading during use".

[0004] Although this design solves the technical problem of existing pumpkin seed processing sorting machines failing to achieve accurate grading during use, traditional sorting methods mostly rely on simple physical screening or manual visual identification in terms of identifying and removing foreign objects. Manual identification is inefficient and prone to omissions, resulting in impurities still being present in the finished pumpkin seeds, affecting product quality and market competitiveness.

[0005] To address this issue, a laser image feature recognition instrument based on a convolutional network model was designed. Utility Model Content

[0006] To address the problems mentioned in the background section, this invention provides a laser image feature recognition instrument based on a convolutional network model, which can effectively solve the problem of impurities still existing in finished pumpkin seeds, affecting product quality.

[0007] To achieve the above objectives, this utility model provides the following technical solution: a laser image feature recognition device based on a convolutional network model, comprising two support plates, a sorting plate disposed between the two support plates, a laser image acquisition device disposed above the sorting plate, a foreign object cleaning execution component disposed on the surface of the support plates, the foreign object cleaning execution component comprising a U-shaped bracket disposed above the support plates and two side plates fixedly connected to the bottom surface of the U-shaped bracket, an image preprocessing module mounted on the surface of the laser image acquisition device, a convolutional network model processing module disposed on one side of the support plates, a second drive motor mounted on the surface of one of the side plates, and a first motor mounted on one end of the U-shaped bracket.

[0008] As a preferred embodiment of the laser image feature recognition instrument based on the convolutional network model of this utility model, each of the side plates is fixedly connected to a first T-shaped block at its bottom end. The first T-shaped block is inserted into a T-shaped groove opened on the upper surface of the support plate, and the first T-shaped block and the support plate are slidably connected.

[0009] In a preferred embodiment of the laser image feature recognition instrument based on the convolutional network model of this utility model, a gear is fixedly connected to the end of the output shaft of the second drive motor, and a toothed plate is meshed with the surface of the gear, and the toothed plate is fixedly connected to one side of one of the support plates.

[0010] In a preferred embodiment of the laser image feature recognition instrument based on the convolutional network model of this utility model, a lead screw is fixedly connected to the end of the output shaft of the first motor. The two ends of the lead screw pass through the two ends of the U-shaped bracket respectively. The lead screw and the U-shaped bracket are rotatably connected. A slider is provided on the surface of the U-shaped bracket. The lead screw is inserted into a threaded hole opened inside the slider. The slider and the lead screw are threadedly connected.

[0011] As a preferred embodiment of the laser image feature recognition instrument based on the convolutional network model of this utility model, the surface of the slider is provided with a mechanical gripping assembly. The mechanical gripping assembly includes a first electric push rod installed on the bottom surface of the slider and a rectangular frame fixedly connected to the end of the telescopic rod of the first electric push rod. Two second T-shaped blocks are inserted inside the rectangular frame. The second T-shaped blocks and the rectangular frame are slidably connected. A second electric push rod is provided between the two second T-shaped blocks. One end of the second electric push rod is fixedly connected to one of the second T-shaped blocks, and the end of the telescopic rod of the second electric push rod is fixedly connected to the other second T-shaped block.

[0012] As a preferred embodiment of the laser image feature recognition instrument based on the convolutional network model of this utility model, each of the second T-shaped blocks is fixedly connected to a clamping plate at its bottom end, the top of the clamping plate is in contact with the surface of the rectangular frame, and clamping buckets are fixedly connected to adjacent sides of the two clamping plates respectively.

[0013] In a preferred embodiment of the laser image feature recognition instrument based on the convolutional network model of this invention, a limiting plate is fixedly connected to the surface of one of the side plates, and a storage box is fixedly connected to the bottom end of the limiting plate.

[0014] Compared with the prior art, the beneficial effects of this utility model are as follows: A foreign object cleaning execution component is added to this application. Through the cooperation of multiple motors, gears, and toothed plates, the clamping bucket is precisely moved to a position above the foreign object. Multiple motors, sliders, and mechanical clamping components work together to quickly clamp and transfer the foreign object, improving the purity of pumpkin seed sorting. Simultaneously, a mechanical clamping component is added. A first electric push rod pushes the rectangular frame to precisely position it at the height of the foreign object, and a second electric push rod drives the clamping bucket to move to both sides of the foreign object and clamp it. This special design ensures reliable clamping, thereby quickly transferring the foreign object to the storage box and ensuring sorting efficiency. Attached Figure Description

[0015] The accompanying drawings are provided to further illustrate the present invention and form part of the specification. They are used together with the embodiments of the present invention to explain the present invention, but do not constitute a limitation thereof. In the drawings:

[0016] Figure 1 This is a schematic diagram of the overall structure of this utility model;

[0017] Figure 2 This is a schematic diagram of the slider and U-shaped bracket in this utility model;

[0018] Figure 3 This is a schematic diagram of the gear and gear plate in this utility model;

[0019] Figure 4 This is a schematic diagram of the side plate and the first T-shaped block in this utility model;

[0020] Figure 5 This is a schematic diagram of the structure of the clamping plate and clamping bucket in this utility model;

[0021] Figure 6 This is a schematic diagram of the structure of the first electric push rod and the rectangular frame in this utility model;

[0022] In the picture:

[0023] 1. Support plate; 2. Sorting plate; 3. Laser image acquisition device;

[0024] 4. Foreign object cleaning execution component; 41. Convolutional network model processing module; 42. Image preprocessing module; 43. U-shaped bracket; 44. First motor; 45. Lead screw; 46. Side plate; 47. First T-block; 48. T-shaped slide; 49. Second drive motor; 410. Toothed plate; 411. Gear; 412. Slider; 413. Wire hole;

[0025] 5. Mechanical clamping assembly; 51. First electric push rod; 52. Rectangular frame; 53. Second T-block; 54. Second electric push rod; 55. Clamping plate; 56. Clamping bucket; 57. Limiting plate; 58. Storage box. Detailed Implementation

[0026] The technical solutions of the present utility model will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present utility model, and not all embodiments. Based on the embodiments of the present utility model, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of the present utility model.

[0027] like Figures 1 to 6 As shown;

[0028] The laser image feature recognition device based on the convolutional network model includes two support plates 1, a sorting plate 2 is set between the two support plates 1, and a laser image acquisition device 3 is set above the sorting plate 2.

[0029] In this implementation plan: According to the public announcement (CN222469762U), a sorting machine for pumpkin seed processing is disclosed. Although this design solves the problem that existing sorting machines for pumpkin seed processing are difficult to achieve accurate grading during use, traditional sorting methods mostly rely on simple physical screening or manual visual identification in terms of identifying and removing foreign objects. Manual identification is inefficient and prone to omissions, resulting in impurities still existing in the finished pumpkin seeds, affecting product quality and market competitiveness. To solve this technical problem, a foreign object cleaning execution component 4 and a mechanical clamping component 5 are added on this basis.

[0030] Furthermore:

[0031] In summary: The surface of the support plate 1 is equipped with a foreign object cleaning execution component 4. This component 4 includes a U-shaped bracket 43 positioned above the support plate 1 and two side plates 46 fixedly connected to the bottom surface of the U-shaped bracket 43. An image preprocessing module 42 is mounted on the surface of the laser image acquisition device 3. A convolutional network model processing module 41 is located on one side of the support plate 1. A second drive motor 49 is mounted on the surface of one of the side plates 46. A first motor 44 is mounted at one end of the U-shaped bracket 43. A gear 411 is fixedly connected to the end of the output shaft 49. A toothed plate 410 is meshed with the surface of the gear 411. The toothed plate 410 is fixedly connected to one side of one of the support plates 1. A lead screw 45 is fixedly connected to the end of the output shaft of the first motor 44. The two ends of the lead screw 45 pass through the two ends of the U-shaped bracket 43 respectively. The lead screw 45 and the U-shaped bracket 43 are rotatably connected. A slider 412 is provided on the surface of the U-shaped bracket 43. The lead screw 45 is inserted into the threaded hole 413 opened inside the slider 412. The slider 412 and the lead screw 45 are threadedly connected.

[0032] In this implementation scheme: the second drive motor 49, through the meshing of gear 411 and toothed plate 410, can drive the U-shaped bracket 43 and side plate 46 to move horizontally along the support plate 1, thereby achieving lateral positioning of the foreign object. The output shaft of the first motor 44 is connected to the lead screw 45. When the lead screw 45 rotates, the slider 412, which is threadedly connected to the lead screw 45, moves on the U-shaped bracket 43 to complete the longitudinal positioning of the foreign object, ensuring that it can be accurately moved above the location of the foreign object, thus preparing for subsequent cleaning operations.

[0033] An image preprocessing module 42 is installed on the surface of the laser image acquisition device 3 to ensure that the acquired images can be preprocessed in a timely manner, such as noise reduction and enhancement, and quickly transmitted to the convolutional network model processing module 41 for analysis and recognition. This improves the efficiency and accuracy of image recognition, allowing the device to quickly determine the location and type of foreign objects and provide accurate information for foreign object removal.

[0034] Furthermore:

[0035] In an optional embodiment, a first T-shaped block 47 is fixedly connected to the bottom end of each side plate 46. The first T-shaped block 47 is inserted into a T-shaped groove 48 opened on the upper surface of the support plate 1, and the first T-shaped block 47 and the support plate 1 are slidably connected.

[0036] In this implementation scheme: the first T-shaped block 47 cooperates with the T-shaped groove 48 on the support plate 1 to provide precise guidance for the movement of the side plate 46. When the second drive motor 49 works to drive the side plate 46 to move, the T-shaped groove 48 can constrain the movement trajectory of the first T-shaped block 47, so that the side plate 46 can only move in a straight line along the direction of the T-shaped groove 48, ensuring the accuracy of the foreign object cleaning execution component 4 when moving in the horizontal direction, thereby ensuring that the laser image acquisition device 3, the image preprocessing module 42 and the subsequent cleaning structure can accurately align with the position of the foreign object;

[0037] The side plate 46 is slidably connected to the support plate 1 via the first T-block 47, which enhances the stability of the entire foreign object cleaning execution component 4 during movement. Under conditions such as motor drive and equipment operation causing vibration, the cooperation of the T-structure can effectively prevent the side plate 46 from shaking or shifting, ensuring the stable operation of components such as the first motor 44 and lead screw 45 installed on the U-shaped bracket 43, and providing a stable foundation for accurately gripping foreign objects.

[0038] Furthermore:

[0039] In an optional embodiment, a mechanical gripping assembly 5 is provided on the surface of the slider 412. The mechanical gripping assembly 5 includes a first electric push rod 51 installed on the bottom surface of the slider 412 and a rectangular frame 52 fixedly connected to the end of the telescopic rod of the first electric push rod 51. Two second T-shaped blocks 53 are inserted inside the rectangular frame 52 and are slidably connected to the rectangular frame 52. A second electric push rod 54 is provided between the two second T-shaped blocks 53. One end of the second electric push rod 54 is fixedly connected to one of the second T-shaped blocks 53, and the end of the telescopic rod of the second electric push rod 54 is fixedly connected to the other second T-shaped block 53. A clamping plate 55 is fixedly connected to the bottom end of each second T-shaped block 53. The top of the clamping plate 55 is in contact with the surface of the rectangular frame 52. A clamping bucket 56 is fixedly connected to the adjacent side of the two clamping plates 55 respectively.

[0040] In this implementation scheme: the slider 412 is threadedly connected to the lead screw 45 and can move up and down along the lead screw 45, driving the mechanical gripping component 5 installed on its surface to move closer to or away from the foreign object. The first electric push rod 51 is installed on the bottom surface of the slider 412. By controlling the extension and retraction of its telescopic rod, the height of the rectangular frame 52 can be further flexibly adjusted, so that the mechanical gripping component 5 can accurately position the height of the foreign object and adapt to the gripping needs of foreign objects at different heights.

[0041] The two second T-shaped blocks 53 inside the rectangular frame 52 are slidably connected to the rectangular frame 52. Under the action of the second electric push rod 54, they can move smoothly towards or away from each other along the track of the rectangular frame 52, so that the clamping plate 55 and the clamping bucket 56 can be moved stably to both sides of the foreign object and reliably clamp the foreign object.

[0042] Furthermore:

[0043] In an optional embodiment, a limiting plate 57 is fixedly connected to the surface of one of the side plates 46, and a storage box 58 is fixedly connected to the bottom end of the limiting plate 57.

[0044] In this implementation scheme: the limiting plate 57 is fixed on the surface of the side plate 46, providing a precise installation position for the storage box 58. When the mechanical gripping component 5 grips the foreign object and moves it above the storage box 58, it can accurately release the foreign object into the storage box 58, thereby achieving precise collection of the foreign object.

[0045] Working principle: Connect the power supply to the equipment and initialize the settings of each component, including calibrating the position of the laser image acquisition device 3 to ensure that it can accurately cover the sorting plate 2 area, checking the running status of the image preprocessing module 42 and the convolutional network model processing module 41, loading the trained convolutional network model and related parameters, and resetting the foreign object cleaning execution component 4 and the mechanical gripping component 5 to put each component in the initial position and prepare for subsequent work.

[0046] Pumpkin seeds are conveyed to sorting plate 2. As the sorting machine runs, the pumpkin seeds begin to move on sorting plate 2. During this process, the laser image acquisition device 3 continuously emits laser beams to irradiate the pumpkin seeds and any foreign objects that may be present, according to the preset working mode.

[0047] After the laser irradiates the pumpkin seeds and foreign objects, the reflected laser image is collected by the laser image acquisition device 3. The acquired image is quickly transmitted to the image preprocessing module 42 installed on its surface through a high-speed data transmission line. The image preprocessing module 42 starts working and uses preset noise reduction algorithms to remove noise interference in the image, such as Gaussian filtering to remove Gaussian noise. It also uses image enhancement algorithms to improve the contrast and clarity of the image, such as histogram equalization, so that the details in the image are more prominent, so as to facilitate subsequent feature extraction and recognition work.

[0048] After preprocessing, the image is transmitted to the convolutional network model processing module 41 via the data bus. The convolutional network model is based on deep learning algorithms and performs layer-by-layer feature extraction and analysis on the image. The convolutional layers inside the model slide on the image through convolutional kernels to extract local features in the image, such as shape and texture. The pooling layer downsamples the feature map to reduce the amount of data and retain key features. The fully connected layer integrates the extracted features and compares them with the features of the trained samples to determine whether there are foreign objects in the image and the type and location of the foreign objects.

[0049] Once the convolutional network model processing module 41 identifies a foreign object, it immediately generates a corresponding control signal. This signal is transmitted to the second drive motor 49 in the foreign object cleaning execution component 4. The second drive motor 49 starts running, and the gear 411 at the end of its output shaft rotates with the motor. Since the gear 411 meshes with the toothed plate 410 fixed on one side of the support plate 1, the gear 411 and the toothed plate 410 work together to drive the side plate 46, the U-shaped bracket 43, and the laser image acquisition device 3 and image preprocessing module 42 installed on the U-shaped bracket 43 to move synchronously, so that the lead screw 45 on the U-shaped bracket 43 moves to directly above the foreign object.

[0050] When the lead screw 45 moves above the foreign object, the first motor 44 starts, and the lead screw 45 at the end of its output shaft begins to rotate. Since the lead screw 45 is inserted into the threaded hole 413 inside the slider 412, and the slider 412 and the lead screw 45 are threadedly connected, the rotation of the lead screw 45 causes the slider 412 to move along the direction of the lead screw 45. A mechanical gripping assembly 5 is installed on the surface of the slider 412. As the slider 412 moves, the mechanical gripping assembly 5 gradually approaches the foreign object.

[0051] The first electric push rod 51 in the mechanical gripping assembly 5 starts to work, and its telescopic rod end pushes the rectangular frame 52 to move downward. Two second T-shaped blocks 53 are inserted inside the rectangular frame 52. During the descent of the rectangular frame 52, the second electric push rod 54 is activated. The telescopic rod end of the second electric push rod 54 pushes the two second T-shaped blocks 53 to move towards the middle. The clamping plate 55 and the clamping bucket 56, which are fixedly connected to the bottom of the two second T-shaped blocks 53, also move accordingly until the two clamping buckets 56 move to both sides of the foreign object. Then, the second electric push rod 54 continues to work, further pushing the two second T-shaped blocks 53 closer together, so that the two clamping buckets 56 clamp the foreign object.

[0052] After the clamping bucket 56 clamps the foreign object, the first electric push rod 51 retracts, driving the rectangular frame 52, the clamping bucket 56, and the foreign object to move upward. After moving to a suitable position, the first motor 44 starts, and the lead screw 45 at the end of its output shaft begins to rotate, so that when the clamping bucket 56 and the foreign object move above the storage box 58, the second electric push rod 54 runs in the opposite direction, causing the two second T-blocks 53 to move to both sides, and the two clamping buckets 56 separate accordingly. The clamped foreign object loses its clamping force and falls into the storage box 58 fixed at the bottom of the limiting plate 57 for storage. Thus, one foreign object identification and cleaning operation is completed. As the pumpkin seeds move continuously on the sorting plate 2, the above steps are repeated to achieve continuous identification and cleaning of foreign objects in the pumpkin seeds on the sorting plate 2.

[0053] Finally, it should be noted that the above description is merely a preferred embodiment of this utility model and is not intended to limit the utility model. Although the utility model has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this utility model should be included within the protection scope of this utility model.

Claims

1. A laser image feature recognition instrument based on a convolutional network model, comprising two support plates (1), a sorting plate (2) is arranged between the two support plates (1), and a laser image acquisition device (3) is arranged above the sorting plate (2), characterized in that: The surface of the support plate (1) is provided with a foreign matter cleaning execution assembly (4), the foreign matter cleaning execution assembly (4) includes a U-shaped support (43) arranged above the support plate (1) and two side plates (46) fixedly connected to the bottom surface of the U-shaped support (43), the surface of the laser image acquisition device (3) is provided with an image preprocessing module (42), one side of the support plate (1) is provided with a convolution network model processing module (41), the surface of one of the side plates (46) is provided with a second driving motor (49), and one end of the U-shaped support (43) is provided with a first motor (44). 2.The laser image feature recognition instrument based on a convolution network model according to claim 1, characterized in that: The bottom end of each side plate (46) is fixedly connected with a first T-shaped block (47), the first T-shaped block (47) is inserted into a T-shaped sliding groove (48) formed on the upper surface of the support plate (1), and the first T-shaped block (47) is in sliding connection with the support plate (1). 3.The laser image feature recognition instrument based on a convolution network model according to claim 1, characterized in that: The end of the output shaft of the second driving motor (49) is fixedly connected with a gear (411), the surface of the gear (411) is engagedly connected with a toothed plate (410), and the toothed plate (410) is fixedly connected to one side of one of the support plates (1). 4.The laser image feature recognition instrument based on a convolution network model according to claim 1, characterized in that: The end of the output shaft of the first motor (44) is fixedly connected with a lead screw (45), the two ends of the lead screw (45) penetrate through the two ends of the U-shaped support (43), the lead screw (45) is in rotational connection with the U-shaped support (43), the surface of the U-shaped support (43) is provided with a sliding block (412), the lead screw (45) is inserted into a lead hole (413) formed in the sliding block (412), and the sliding block (412) is in threaded connection with the lead screw (45). 5.The laser image feature recognition instrument based on a convolution network model according to claim 4, characterized in that: The surface of the sliding block (412) is provided with a mechanical clamping assembly (5), the mechanical clamping assembly (5) includes a first electric push rod (51) arranged on the bottom surface of the sliding block (412) and a rectangular frame (52) fixedly connected to the end of the telescopic rod of the first electric push rod (51), two second T-shaped blocks (53) are inserted into the rectangular frame (52), the second T-shaped blocks (53) are in sliding connection with the rectangular frame (52), a second electric push rod (54) is arranged between the two second T-shaped blocks (53), one end of the second electric push rod (54) is fixedly connected to one of the second T-shaped blocks (53), and the end of the telescopic rod of the second electric push rod (54) is fixedly connected to the other second T-shaped block (53). 6.The laser image feature recognition instrument based on a convolution network model according to claim 5, characterized in that: The bottom end of each second T-shaped block (53) is fixedly connected with a clamping plate (55), the top of the clamping plate (55) is attached to the surface of the rectangular frame (52), and the adjacent sides of the two clamping plates (55) are fixedly connected with clamping hoppers (56). 7.The laser image feature recognition instrument based on a convolutional network model according to claim 1, characterized in that: The surface of one of the side plates (46) is fixedly connected with a limiting plate (57), and the bottom end of the limiting plate (57) is fixedly connected with a storage box (58).

Images