Material screening equipment and screening method based on AI visual inspection

Abstract

The invention discloses material screening equipment based on AI visual inspection and a screening method, and belongs to the technical field of material screening equipment. The screening unit comprises a first side plate and a second side plate which are arranged on the two sides of the conveying unit respectively, a plurality of air outlet parts are arranged on the first side plate in the length direction of the first side plate, and a plurality of discharging ports allowing the rod-shaped materials to pass through are formed in the second side plate; the Ai visual detection unit is used for generating defect type data when identifying and judging that the bar-shaped material has defects; and the gas supply unit is communicated with the plurality of gas outlet parts and is used for supplying gas to the corresponding gas outlet parts according to the defect type data. A non-contact air-blowing sorting mode is adopted, the air supply unit controls airflow sprayed out of the corresponding air outlet parts to blow defective materials out of the corresponding discharging openings according to defect types, and damage to the surfaces of corn ears caused by a mechanical sorting mode is avoided while accurate screening of different defective bar-shaped materials is achieved; and the product quality is effectively guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of material screening equipment, in particular to a material screening equipment based on AI vision detection and a screening method. BACKGROUND

[0002] At present, screening equipment is widely used in the processing of rod-shaped materials (such as corn ears, carrots and other crops). Taking corn ear processing and screening as an example, the process usually includes: corn ears are peeled by a peeling machine and then transported to a screening equipment. Some screening equipment can first detect the quality of corn ears to identify defects such as mildewing, mixed ears or unpeeled, and finally sort and remove defective corn ears according to the detection results. However, most screening equipment usually uses mechanical methods such as mechanical prongs to screen defective corn ears, but this method can easily cause mechanical damage to the surface of the corn ears, affecting product quality. Therefore, we propose a material screening equipment based on AI vision detection and a screening method. SUMMARY

[0003] The present application aims to provide a material screening equipment based on AI vision detection and a screening method, which solves the technical problem that most screening equipment usually uses mechanical methods such as mechanical prongs to screen defective corn ears, but this method can easily cause mechanical damage to the surface of the corn ears, affecting product quality.

[0004] The present application achieves the above-mentioned purpose through the following technical solutions: A material screening equipment based on AI vision detection, comprising: a conveying unit for conveying rod-shaped materials; a screening unit comprising a first side plate and a second side plate respectively arranged on both sides of the conveying unit, a plurality of air outlets are arranged on the first side plate along the length direction thereof, a plurality of discharge ports for the rod-shaped materials to pass through are formed on the second side plate, and the discharge ports and the air outlets correspond one by one; an AI vision detection unit for collecting image data of the rod-shaped materials before they pass through the discharge ports, and identifying and judging whether the rod-shaped materials have defects according to the image data, and generating defect type data when it is identified and judged that the rod-shaped materials have defects; a gas supply unit connected to the plurality of air outlets respectively, for supplying gas to the corresponding air outlets according to the defect type data, so that the rod-shaped materials corresponding to the defect type data are discharged from the corresponding discharge ports.

[0005] Further improvements are that the conveying unit comprises several groups of conveying belt mechanisms arranged side by side, each conveying belt mechanism comprises two groups of symmetrically arranged roller bodies, and a conveying belt driven on the two roller bodies, and the first side plate and the second side plate are symmetrically arranged on both sides of the conveying belt. The two groups of roller bodies in each group of conveying belt mechanisms are fixedly sleeved on the outer walls of two shaft bodies, and the end of one shaft body is connected to a rotating device.

[0006] Further improvements are that the air outlet part comprises several air outlet holes arranged at the side of the first side plate facing the discharge port, the several air outlet holes are arranged in a direction parallel to the first side plate, the several air outlet holes are all communicated with connecting ports, the several connecting ports are all arranged on the first side plate, and the several connecting ports are all communicated with the gas supply unit.

[0007] Further improvements are that the Ai visual detection unit comprises: An image acquisition module for acquiring rod-shaped material image data; A preprocessing module for receiving rod-shaped material image data and processing the rod-shaped material image data to improve image quality; A deep learning model module for receiving preprocessed rod-shaped material image data, extracting and identifying features of the rod-shaped material image data to obtain rod-shaped material feature data, and identifying whether the rod-shaped material has defects according to the rod-shaped material feature data and generating defect type data when a defect is identified; A storage module for storing defect type data; A display module for displaying defect type data; A control module for controlling the gas supply unit to supply gas to the corresponding air outlet part according to the defect type data.

[0008] Further improvements are that the gas supply unit comprises a mounting frame, a protective shell arranged on the mounting frame, a gas source distributor arranged in the protective shell, and several pipe groups connected to the output end of the gas source distributor, the input end of the gas source distributor is connected to an external gas source, and the several pipe groups are one-to-one communicated with the several air outlet parts; The pipe group comprises several connecting pipes, and the several connecting pipes are one-to-one communicated with the several connecting ports.

[0009] Further improvements are that the connecting pipe is provided with a control valve electrically connected with the control module.

[0010] Further improvements are that the Ai visual detection unit further comprises: A size data extraction unit for receiving preprocessed rod-shaped material image data and processing the rod-shaped material image data to obtain material length data; The control module is further configured to receive material length data, and when the deep learning model module determines that the rod-shaped material has defects, control the number of open control valves on the connecting pipeline corresponding to the air outlet according to a preset length interval threshold, wherein the material length and the number of open control valves are in a positive correlation.

[0011] Further improvements are that the image acquisition module includes a support frame assembly, an image acquisition device arranged on the support frame assembly for acquiring image data of the rod-shaped material, and a light source device arranged on the support frame assembly for providing illumination for the image acquisition module. The support frame assembly includes two vertical frames, a horizontal frame movably arranged between the two vertical frames, the image acquisition device is movably arranged on the horizontal frame, and the light source device is movably arranged between the two vertical frames and on the side of the horizontal frame facing the conveying unit.

[0012] Further improvements are that the screening device further includes a collection unit, and the collection unit includes: A first collection hopper for receiving the rod-shaped material output from the output end of the conveying unit; A plurality of second collection hoppers corresponding one-to-one to the plurality of discharge ports for receiving the rod-shaped material discharged from the discharge ports.

[0013] Further improvements are that the screening device further includes a rack main body, and the conveying unit, the gas supply unit, the Ai visual detection unit, the collection unit, and the screening unit are all arranged on the rack main body, and the conveying unit further includes a guide opening obliquely arranged on the rack main body and corresponding to the input end of the conveying unit.

[0014] A material screening method based on AI visual detection, using the above-mentioned screening device, includes the following steps: S1: conveying the rod-shaped material through the conveying unit; S2: the Ai visual detection unit acquires image data of the rod-shaped material, and identifies whether the rod-shaped material has defects according to the image data of the rod-shaped material, and generates defect type data when it is identified that the rod-shaped material has defects; S3: the gas supply unit supplies gas to the corresponding air outlet according to the defect type data, and when the rod-shaped material is conveyed to the corresponding discharge port position by the conveying unit, the airflow generated by the air outlet causes the rod-shaped material corresponding to the defect type data to be discharged from the corresponding discharge port to complete the screening.

[0015] The beneficial effects of the present application are that: the present application adopts a non-contact air blowing sorting mode, detects the presence of defects in the rod-shaped material through the Ai visual detection unit when the conveying unit is conveying the rod-shaped material, and according to the defect type, the air supply unit controls the air flow of the corresponding air outlet to blow the defective material out of the corresponding discharge port, realizing accurate screening of different defective rod-shaped materials while avoiding mechanical damage to the surface of the corn ear caused by the traditional mechanical sorting method, effectively protecting the product quality, and the Ai visual detection unit can also control the working number of the air outlet hole of the corresponding air outlet according to the length of the rod-shaped material, not only avoiding sorting failure or damage to the rod-shaped material, but also effectively optimizing the energy utilization rate and saving the use cost. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 Structure schematic view of the rod-shaped material screening device of the present application; Figure 2 Structure schematic view of the rod-shaped material screening device of the present application; Figure 1 Structure schematic view of the rod-shaped material screening device of the present application; Figure 3 Structure schematic view of the rod-shaped material screening device of the present application; Figure 1 Structure schematic view of the rod-shaped material screening device of the present application; Figure 4 Structure schematic view of the rod-shaped material screening device of the present application; Figure 5 Structure schematic view of the rod-shaped material screening device of the present application; Figure 4 Structure schematic view of the rod-shaped material screening device of the present application; Figure 6 Principle block diagram of the Ai visual detection unit of the present application; Figure 7 Working flow chart of the Ai visual detection unit of the present application; Figure 8 Structure schematic view of the rod-shaped material screening device of the present application; Figure 9 Structure schematic view of the rod-shaped material screening device of the present application; Figure 8 Structure schematic view of the rod-shaped material screening device of the present application.

[0017] In the figure: 100, conveying unit; 101, conveying belt mechanism; 1011, roller body; 1012, conveying belt; 102, material guide port; 200, air supply unit; 201, mounting frame; 202, protective shell; 203, air source distributor; 204, pipeline group; 2041, connecting pipeline; 300, Ai visual detection unit; 301, support frame assembly; 3011, vertical frame; 3012, horizontal frame; 302, image acquisition device; 303, light source device; 400, rack main body; 500, collection unit; 501, first collection hopper; 502, second collection hopper; 600, screening unit; 601, first side plate; 602, second side plate; 603, air outlet hole; 604, connecting port; 605, discharge port. DETAILED DESCRIPTION

[0018] The application will be further described in detail below with reference to the drawings. It is necessary to point out here that the following specific embodiments are only used to further illustrate the application and cannot be understood as limiting the scope of protection of the application. Those skilled in the art can make some non-essential improvements and adjustments to the application according to the above application content.

[0019] Below, please refer to the drawings Figures 1-9 The rod-shaped material screening device according to the present embodiment is described, wherein the rod-shaped material includes but is not limited to corn cobs (corn ears), carrots, etc. In the following description of the present application, corn ears are taken as an example for illustration. Of course, those skilled in the art can understand that the present rod-shaped material screening device can also be used for screening other rod-shaped materials and is not limited to the above. In actual situations, there are moldy corn ears, mixed ears (non-target ears, for example, the preset target corn ears are yellow, and the corn ears transported are colored, which are non-target ears), and corn ears without husking, etc. Preferably, the present rod-shaped material screening device can be arranged upstream of the existing husking device, that is, the corn ears are husked by the husking device and then enter the present rod-shaped material screening device for screening.

[0020] Embodiment 1 Please refer to the drawings Figures 1-5 The present application is a material screening device based on AI visual detection, which comprises the following structures: The conveying unit 100 is used for conveying rod-shaped materials. Corn ears are input from the input end of the conveying belt mechanism 101 to the output end, and are screened in the conveying process, facilitating efficient large-batch screening operations. Preferably, the width of the conveying unit 100 is adapted to the width of the rod-shaped material, ensuring that the rod-shaped materials are conveyed one by one. The screening unit 600 comprises a first side plate 601 and a second side plate 602 arranged on the two sides of the conveying unit 100, respectively. The first side plate 601 and the second side plate 602 effectively limit the corn ears on the conveying unit 100, so that they do not fall from the side of the conveying unit 100. A plurality of air outlets are arranged on the first side plate 601 along the length direction thereof. A plurality of discharge ports 605 for the rod-shaped materials are formed on the second side plate 602. The discharge ports 605 and the air outlets one-to-one correspond. Optionally, the second side plate 602 is a plate body, and the discharge ports 605 are formed on the plate body. Further, the second side plate 602 comprises a plurality of plate bodies, and the discharge ports 605 are formed between adjacent two plate bodies. In the present embodiment, the discharge ports 605 and the air outlets are both provided at least two. The discharge ports 605 and the air outlets at different positions cooperate to control the discharge of corn ears with different defects. For example, as shown in the drawings Figure 4As shown, one discharge port 605 close to the output end of the conveying unit 100 is discharge port A, and one discharge port 605 far from the output end of the conveying unit 100 is discharge port B. If the defect of the conveyed rod-shaped material is mildew, the air outlet corresponding to discharge port A works, so that the rod-shaped material is conveyed to discharge port A and discharged from discharge port A under the action of the air outlet. If the defect of the conveyed rod-shaped material is unpeeling, the air outlet corresponding to discharge port B works, so that the rod-shaped material is conveyed to discharge port B and discharged from discharge port B under the action of the air outlet. Different types of corn ears with defects can be accurately guided to the corresponding discharge port 605 for discharge, realizing efficient and accurate multi-stage classification and screening. The screening device adopts a non-contact air blowing sorting method, avoiding mechanical damage to the surface of the corn ear caused by the traditional mechanical sorting method, and effectively ensuring the product quality. The Ai visual detection unit 300 is used to collect rod-shaped material image data before the rod-shaped material passes through the discharge port 605 (i.e., the corn ear is detected by the Ai visual detection unit 300 before passing through the discharge port 605 during conveying by the conveying unit 100, and the corresponding air outlet is controlled to work according to the detection result to realize screening), and to identify and judge whether the rod-shaped material has defects according to the rod-shaped material image data. When it is identified and judged that the rod-shaped material has defects, defect type data is generated. It should be noted that if it is identified and judged that the rod-shaped material has no defects, it will be normally conveyed to the output end of the conveying unit 100 and not discharged from the discharge port 605. The gas supply unit 200 is connected to a plurality of air outlets, respectively, and is used to supply gas to the corresponding air outlet according to the defect type data, so that the rod-shaped material corresponding to the defect type data is discharged from the corresponding discharge port 605. For example, when the corn ear is conveyed by the conveying unit 100 and detected by the Ai visual detection unit 300 to determine that the corn ear has a mildew defect, a signal is sent to the gas supply unit 200, the air outlet of the discharge port 605 for discharging the mildew defect corn ear of the gas supply unit 200 is used to blow air, and when the corn ear is conveyed by the conveying unit 100 to the discharge port 605 for discharging the mildew defect corn ear, it is discharged from the discharge port 605 under the action of the corresponding air outlet. Optionally, the embodiment can be provided with a sensor (such as a photoelectric sensor or a proximity switch, not shown in the figure) corresponding to each outlet on the first side plate 601 in actual conditions, and the sensors work in linkage with the AI vision detection unit. For example, when the AI vision detection unit determines that the corn ear is moldy, only the sensor corresponding to the moldy outlet 605 is activated. When the activated sensor detects that the target corn ear reaches the predetermined position during the conveying process, the air supply unit 200 is triggered to start blowing air to the corresponding outlet. The remaining unassociated sensors remain in a dormant state. This way can not only ensure that the defective corn ear is accurately sorted, but also avoid misoperation and energy waste.

[0021] A material screening method based on AI vision detection, using the above screening device, comprising the following steps: S1: conveying rod-shaped materials through the conveying unit 100; S2: the AI vision detection unit 300 collects rod-shaped material image data and identifies whether the rod-shaped material has defects according to the rod-shaped material image data. When it is identified that the rod-shaped material has defects, generate defect type data; S3: the air supply unit 200 supplies air to the corresponding outlet according to the defect type data. When the rod-shaped material is conveyed to the corresponding outlet 605 position by the conveying unit 100, the airflow generated by the outlet makes the rod-shaped material corresponding to the defect type data discharged from the corresponding outlet 605 to complete the screening.

[0022] Please refer to the accompanying Figures 1-5 As a preferred, the conveying unit 100 of the embodiment includes several groups of side-by-side arranged conveying belt mechanisms 101. The conveying belt mechanism 101 includes two groups of symmetrically arranged roller bodies 1011, and a conveying belt 1012 driven on the two roller bodies 1011. The first side plate 601 and the second side plate 602 are symmetrically arranged on both sides of the conveying belt 1012. Optionally, the conveying belt mechanism 101 in the embodiment is provided with sixteen groups to meet the high-batch corn ear detection and screening work and improve the work efficiency. The two roller bodies 1011 in each group of the conveying belt mechanism 101 are respectively fixedly sleeved on the outer walls of two shaft bodies, and the end of one shaft body is connected to a rotating device. The rotating device includes a rotating motor and a speed reducer. Further, when this arrangement is adopted, the structure design between part of the adjacent two groups of conveying belt mechanisms 101 can be optimized. For example, by sharing one first side plate 601, the number of components is reduced, but the air outlet holes 603 are provided on both sides of the first side plate 601 to respectively correspond to the air supply requirements of the conveying belt mechanisms 101 on both sides, so as to balance the space utilization rate and the functional integrity. This will not be described in detail. The outlets 605 at the adjacent two groups of conveying belt mechanisms 101 are staggered to prevent interference when the materials fall. Please refer to the attached Figure 2 , Figures 5-6 As preferred, the air outlet part of the embodiment includes a plurality of air outlet holes 603 arranged on the side of the first side plate 601 facing the discharge port 605, the plurality of air outlet holes 603 are arranged in a direction parallel to the first side plate 601, the plurality of air outlet holes 603 are all communicated with a connecting port 604, the plurality of connecting ports 604 are all arranged on the first side plate 601, and the plurality of connecting ports 604 are all communicated with the air supply unit 200. Optionally, the air outlet holes 603 in the embodiment are provided with eight groups, and the length of the air outlet part is matched with the length of the discharge port 605. In the embodiment, the number of the connecting ports 604 is the same as the number of the air outlet holes 603, that is, if the air outlet holes 603 are eight groups, the number of the connecting ports 604 is also eight groups, and each air outlet hole 603 is controlled by an independent air path, so as to flexibly adjust the number of the opened air outlet holes 603 according to the size of the corn ear and the type of the defect, and realize the precise and controllable air flow sorting effect.

[0023] Please refer to the attached Figures 6-7 As preferred, the Ai visual detection unit 300 of the embodiment includes: An image acquisition module for acquiring rod-shaped material image data; A preprocessing module for receiving rod-shaped material image data and processing the rod-shaped material image data, for example, the preprocessing module performs denoising, cropping, normalization and other operations on the acquired rod-shaped material image data to improve the image quality; A deep learning model module for receiving the preprocessed rod-shaped material image data, extracting and identifying the features of the rod-shaped material image data to obtain rod-shaped material feature data, and judging whether the rod-shaped material has defects according to the rod-shaped material feature data, and generating defect type data when it is judged as a defect, such as mold defect, mixed ear defect, and unpeeled defect. The deep learning model module belongs to the conventional technology in the field. Deep learning is an important branch of machine learning, which can automatically learn features from data by constructing a multi-layer neural network model without manual feature extraction. In the field of image recognition, deep learning models such as convolutional neural networks (CNN) have shown excellent performance and can effectively identify complex patterns and features in images. The embodiment preferentially adopts a small sample deep learning model based on meta-learning, which is composed of a feature extractor and a classifier. The feature extractor uses an improved ResNet architecture to extract the texture (such as irregular texture of moldy area), color (such as abnormal color difference of mixed ear), and morphological features (such as the contour features of unpeeled corn) of the corn ear, generating a high-dimensional feature vector. The classifier uses a fully connected neural network combined with prototype metric learning. By calculating the similarity between the input features and the pre-trained defect feature prototypes (including mold, mixed ear, unpeeled, etc.), the softmax function outputs the probability distribution of each defect type. When the probability of a certain defect type exceeds a certain threshold and is significantly higher than other categories, it is determined that the corn ear has the corresponding defect. Then it is classified and the corresponding control signal is sent to the control module. Secondly, the deep learning model module training stage adopts a task division training mechanism, which divides the labeled data set into multiple independent tasks. Each task contains a support set (used to construct a defect prototype) and a query set (used to verify the classification effect). Training multiple tasks enables the model to learn general feature representation and classification methods, thereby quickly and accurately identifying defects in corn ears. The storage module is used to store defect type data. Optionally, the storage module in the embodiment uses a high-speed SSD memory to record real-time defect type data (including defect category, similarity, image frame number, and processing time) with a time stamp, and realizes fast query and historical data backtracking through database indexing. The display module is used to display the defect type data. Optionally, the display module in the embodiment integrates an industrial-grade touch screen to visually display the current detection results (including defect classification images, statistical charts, and sorting efficiency, etc. Key indicators) in real time, while providing an audible and visual alarm function to alert serious defects. The control module is used to control the air supply unit 200 to supply air to the corresponding air outlet according to the defect type data. Optionally, the control module in the embodiment is a PLC controller. The control module receives the defect type data and accurately controls the corresponding air outlet in the air supply unit 200 according to the pre-set defect-air outlet mapping relationship table. For example, when moldy corn ears are detected, the air outlet at the moldy corn ear discharge outlet 605 is controlled to work. As shown in FIG. 6, the corn ear sorting system 600 includes a detection module 100, a control module 300, a storage module 400, a display module 500, and an air supply unit 200. Figure 6As shown, if the defect A is a mildew defect, the defect B is a mixed ear defect, and the defect C is an unpeeled defect, the outlet A is used to remove the mildew defect corn ear, the outlet B is used to remove the mixed ear defect corn ear, and the outlet C is used to remove the unpeeled defect corn ear; The corn ear to be detected is conveyed by the conveying unit 100, and the image acquisition module sends the image data to the deep learning model module after acquiring the image data. After being processed and recognized by the deep learning model module, it is determined that the corn ear has a mildew defect. Then, a signal is sent to the control module to control the air outlet part of the corresponding outlet A to work, so that the mildew defect corn ear is removed from the outlet A during continuous conveying. If the corn ear does not have a defect, it is normally conveyed to the output end of the conveying unit 100.

[0024] Please refer to the accompanying Figures 8-9 As preferred, the air supply unit 200 of the embodiment includes a mounting frame 201 (vertical section in U shape), a protective shell 202 (one side of the protective shell 202 is provided with a shell door for regular maintenance of the internal structure) arranged on the mounting frame 201, an air source distributor 203 (the air source distributor 203 is used for centralized distribution of compressed air, has a plurality of output ends, and is widely used in the field, which will not be described in detail here) arranged in the protective shell 202, and a plurality of pipe groups 204 (the number of the pipe groups 204 is the same as that of the outlets 605 and the air outlets, so as to supply air to the air outlets 603 of different outlets 605) connected to the output ends of the air source distributor 203. The input end of the air source distributor 203 is connected to an external air source (such as a gas tank for supplying compressed gas, etc.). The plurality of pipe groups 204 are respectively in communication with the plurality of air outlets. Among them, the pipe group 204 includes a plurality of connection pipes 2041 (the number of the connection pipes 2041 is the same as that of the connection ports 604 and the air outlets 603 in the corresponding air outlet), and the plurality of connection pipes 2041 are in one-to-one communication with the plurality of connection ports 604. In this way, the independence and stability of air supply of each air outlet 603 are ensured, and the system maintenance and precise control of the number of air outlets 603 are facilitated.

[0025] Please refer to the accompanying Figures 1-2 As preferred, the image acquisition module of the embodiment includes a support frame assembly 301, an image acquisition device 302 (such as an industrial camera, etc.) arranged on the support frame assembly 301 for acquiring image data of rod-shaped materials, and a light source device 303 (such as an LED light source, etc.) arranged on the support frame assembly 301 for providing illumination for the image acquisition module. The support frame assembly 301 comprises two groups of vertical frames 3011, a horizontal frame 3012 movably arranged between the two groups of vertical frames 3011, and optionally, vertical sliding grooves are arranged on the side walls of the vertical frames 3011, the horizontal frame 3012 is connected to the vertical sliding grooves at both ends through sliding blocks to realize height adjustment, and the position is fixed through locking bolts; the image acquisition device 302 is movably arranged on the horizontal frame 3012, and the light source device 303 is movably arranged between the two groups of vertical frames 3011 and located on the side of the horizontal frame 3012 facing the conveying unit 100, and the setting mode can adopt the same connection mode as the horizontal frame 3012 and the vertical frame 3011. It should be noted that the conveying belt mechanism 101 has several groups, and the corresponding image acquisition module also has several groups.

[0026] Please refer to the accompanying drawings Figure 4 As preferred, the screening device of the embodiment further comprises a collecting unit 500, and the collecting unit 500 comprises: A first collecting hopper 501 for receiving the rod-shaped materials output from the output end of the conveying unit 100; A plurality of second collecting hoppers 502 corresponding to the plurality of discharge ports 605, for receiving the rod-shaped materials discharged from the discharge ports 605; Through the above arrangement, the first collecting hopper 501 is used to receive the rod-shaped materials without defects, and the second collecting hopper 502 is used to receive the rod-shaped materials sorted and discharged, so as to realize the orderly collection of the rod-shaped materials without defects and the rod-shaped materials with different defects (including mildew, mixed spikes or unpeeled), and improve the operation convenience of the subsequent operator in processing the materials.

[0027] Please refer to the accompanying drawings Figures 1-3 As preferred, the screening device of the embodiment further comprises a rack main body 400, and the conveying unit 100, the gas supply unit 200, the Ai visual detection unit 300, the collecting unit 500 and the screening unit 600 are arranged on the rack main body 400. Specifically, the conveying belt mechanism 101 (shaft body, rotating device) in the conveying unit 100, the mounting frame 201 in the gas supply unit 200, the first collecting hopper 501 and the second collecting hopper 502 in the collecting unit 500, and the first side plate 601 and the second side plate 602 in the screening unit 600 are all connected to the rack main body 400, and the conveying unit 100 further comprises a guide port 102 obliquely arranged on the rack main body 400 and corresponding to the input end of the conveying unit 100, as shown in the accompanying drawings Figures 1-3 As can be seen, one end of the guide port 102 facing the input end of the conveying unit 100 is lower than the other end, and the other end of the guide port 102 can be connected to the output end of the upstream peeling device, so that the peeled corn ears can naturally slide onto the conveying unit 100 by gravity for conveying, detection and screening.

[0028] Please refer to the attached Figure 4 As preferred, the discharge port 605 of the present embodiment is inclinedly provided with a folded edge on the top of the outer wall on the side away from the input end of the conveying belt mechanism 101, as shown in the attached Figure 4 The tip of the folded edge extends outward, which facilitates the smooth sliding of the rod-shaped material to the discharge port 605 when the air outlet part discharges air.

[0029] Embodiment 2 On the basis of embodiment 1, the connecting pipeline 2041 of the present embodiment is provided with a control valve (not shown in the figure) electrically connected with the control module. Optionally, the control valve in the present embodiment can adopt a high-frequency electronic valve, of course, it is not limited to this. Due to the fact that the air supply of one air outlet hole 603 is controlled by one connecting pipeline 2041, the number of air outlet holes 603 can be controlled by the control valve, or the air outlet of air outlet holes 603 at different positions can be realized. Further, the control valve can also be programmed by PLC or a special controller to enable millisecond-level delay triggering of multiple control valves, so that each air outlet hole 603 acts in a preset order, which will not be described in detail here.

[0030] Please refer to the attached Figure 5 As preferred, the Ai visual detection unit 300 of the present embodiment further comprises: a size data extraction unit for receiving the pre-processed rod-shaped material image data and processing the rod-shaped material image data to obtain material length data; for example, in the present embodiment, after the size data receives the image data processed by the pre-processing module, the material main contour is extracted by an edge detection algorithm, the pixel length is calculated by using the minimum circumscribed rectangle method (for the curved shape, the central axis length is obtained by combining the skeletonization algorithm), and then the actual material length data is obtained according to the pre-calibrated pixel-physical size conversion coefficient, of course, it is not limited to this way; The control module is also configured to receive the material length data, and when the deep learning model module judges that the rod-shaped material has defects, the opening number of the control valve on the connecting pipeline 2041 corresponding to the air outlet part is controlled according to the preset length interval threshold, wherein the material length and the opening number of the control valve are in a positive correlation (i.e., according to the length of the corn ear to determine the number of air outlet holes, the longer the corn, the more air outlet holes are opened, and the opening is performed according to the preset setting, for example, if the air outlet hole 603 of the air outlet part is eight groups, then the connecting pipeline 2041 connected with the air outlet part is eight groups, and there are eight control valves, when the length of the corn ≤15 cm → 4 control valves are opened, 15 cm < the length of the corn ≤25 cm → 6 control valves are opened, and the length of the corn ≥26 cm → 8 control valves are opened).

[0031] By the above-mentioned mode, the airflow coverage range is accurately matched with the material length, which not only avoids the sorting failure of short corn ears due to insufficient airflow or the damage of long corn ears due to excessive airflow, but also effectively optimizes the energy utilization rate. According to the material length, a necessary number of air outlet holes 603 are opened as needed, which effectively reduces the energy consumption and saves the use cost compared with the traditional full opening mode.

[0032] The above-mentioned embodiments only express several embodiments of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent of the present application. It should be pointed out that for ordinary skilled persons in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which belong to the protection scope of the present application.

Claims

1. An AI vision detection-based material screening device, characterized in that, The application relates to a rod-shaped material screening device. The rod-shaped material screening device comprises a conveying unit (100) for conveying rod-shaped materials, a screening unit (600) comprising a first side plate (601) and a second side plate (602) respectively arranged on the two sides of the conveying unit (100), a plurality of air outlets are arranged on the first side plate (601) along the length direction of the first side plate (601), a plurality of discharge ports (605) for the rod-shaped materials to pass through are formed on the second side plate (602), and the discharge ports (605) and the air outlets are in one-to-one correspondence. An Ai visual detection unit (300) is arranged for collecting rod-shaped material image data before the rod-shaped materials pass through the discharge ports (605), identifying and judging whether the rod-shaped materials have defects according to the rod-shaped material image data, and generating defect type data when it is judged that the rod-shaped materials have defects. A gas supply unit (200) is arranged for being communicated with the plurality of air outlets respectively, for supplying gas to the corresponding air outlets according to the defect type data, so that the rod-shaped materials corresponding to the defect type data are discharged from the corresponding discharge ports (605). The conveying unit (100) comprises a plurality of groups of conveying belt mechanisms (101) arranged side by side, each conveying belt mechanism (101) comprises two groups of roller bodies (1011) arranged symmetrically, and a conveying belt (1012) arranged in transmission on the two roller bodies (1011), and the first side plate (601) and the second side plate (602) are symmetrically arranged on the two sides of the conveying belt (1012).

2. The screening apparatus of claim 1, wherein, The two groups of roller bodies (1011) in the plurality of groups of conveying belt mechanisms (101) are fixedly sleeved on the outer walls of two shaft bodies respectively, and the end portion of one shaft body is connected with a rotating device. The air outlet comprises a plurality of air outlet holes (603) arranged at the side of the first side plate (601) facing the discharge ports (605), the plurality of air outlet holes (603) are arranged in parallel with the first side plate (601), the plurality of air outlet holes (603) are all communicated with connecting ports (604), the plurality of connecting ports (604) are all arranged on the first side plate (601), and the plurality of connecting ports (604) are all communicated with the gas supply unit (200).

3. The screening apparatus of claim 1, wherein, The Ai visual detection unit (300) comprises:

4. The screening apparatus of claim 3, wherein, An image acquisition module is arranged for acquiring rod-shaped material image data; A pretreatment module is arranged for receiving the rod-shaped material image data and processing the rod-shaped material image data to improve the image quality; A deep learning model module is arranged for receiving the pretreated rod-shaped material image data, extracting and identifying the features of the rod-shaped material image data to obtain rod-shaped material feature data, judging whether the rod-shaped material has defects according to the rod-shaped material feature data, and generating defect type data when it is judged that the rod-shaped material has defects; A storage module is arranged for storing the defect type data; A display module is arranged for displaying the defect type data; A control module is arranged for controlling the gas supply unit (200) to supply gas to the corresponding air outlets according to the defect type data. ​ 5. The screening apparatus of claim 4, wherein, The air supply unit (200) comprises a mounting frame (201), a protective shell (202) arranged on the mounting frame (201), an air source distributor (203) arranged in the protective shell (202), and a plurality of pipeline groups (204) connected to the output end of the air source distributor (203); the input end of the air source distributor (203) is connected to an external air source, and the plurality of pipeline groups (204) are respectively in one-to-one communication with a plurality of air outlets. The pipeline group (204) comprises a plurality of connecting pipelines (2041), and the plurality of connecting pipelines (2041) are in one-to-one communication with a plurality of connecting ports (604).

6. The screening apparatus of claim 5, wherein, The connecting pipeline (2041) is provided with a control valve electrically connected to the control module.

7. The screening apparatus of claim 6, wherein, The Ai visual detection unit (300) further comprises: A size data extraction unit configured to receive the preprocessed rod-shaped material image data and process the rod-shaped material image data to obtain material length data. The control module is further configured to receive the material length data and control the number of open control valves on the connecting pipeline (2041) of the corresponding air outlet when the deep learning model module determines that the rod-shaped material has a defect, wherein the material length and the number of open control valves are in a positive correlation.

8. The screening apparatus of claim 4, wherein, The image acquisition module comprises a support frame assembly (301), an image acquisition device (302) arranged on the support frame assembly (301) for acquiring rod-shaped material image data, and a light source device (303) arranged on the support frame assembly (301) for providing illumination for the image acquisition module. The support frame assembly (301) comprises two groups of vertical frames (3011), a horizontal frame (3012) movably arranged between the two groups of vertical frames (3011), the image acquisition device (302) is movably arranged on the horizontal frame (3012), and the light source device (303) is movably arranged between the two groups of vertical frames (3011) and located on one side of the horizontal frame (3012) facing the conveying unit (100).

9. The screening apparatus of claim 1, wherein, The screening device further comprises a collecting unit (500), and the collecting unit (500) comprises: A first collecting hopper (501) for receiving the rod-shaped material output from the output end of the conveying unit (100); A plurality of second collecting hoppers (502) corresponding to a plurality of discharge ports (605) for receiving the rod-shaped material discharged from the discharge ports (605).

10. The screening apparatus of claim 9, wherein, The screening device further comprises a rack main body (400), and the conveying unit (100), the air supply unit (200), the Ai visual detection unit (300), the collecting unit (500), and the screening unit (600) are arranged on the rack main body (400), and the conveying unit (100) further comprises a guide port (102) arranged obliquely on the rack main body (400) and corresponding to the input end of the conveying unit (100).

11. A material screening method based on AI vision detection, using the screening device according to any one of claims 1-10, characterized in that, The method comprises the following steps: S1: conveying the rod-shaped material through the conveying unit (100); S2: The Ai visual inspection unit (300) collects rod material image data, and identifies whether the rod material has defects according to the rod material image data. When it is identified that the rod material has defects, defect type data is generated; S3: The gas supply unit (200) supplies gas to the corresponding gas outlet according to the defect type data. When the rod material is transported to the corresponding discharge port (605) position by the conveying unit (100), the airflow generated by the gas outlet causes the rod material corresponding to the defect type data to be discharged from the corresponding discharge port (605), completing the screening.

Images