A watermelon maturity grading device and method based on visual detection

Abstract

The application discloses a water caltrop maturity grading equipment and method based on visual detection, and the grading equipment comprises a feeding mechanism, a conveying mechanism, a sorting mechanism, an image collector and a controller, the feeding port of the conveying mechanism is connected with the feeding mechanism, the sorting mechanism is arranged on the two sides of the conveying mechanism, the sorting mechanism can push the water caltrop on the conveying mechanism to move to separate from other water caltrop, the image collector is arranged towards the conveying mechanism, and the controller is signal-connected with the image collector, the feeding structure, the conveying mechanism and the sorting mechanism respectively. The application solves the technical problems that the existing water caltrop maturity grading technology has low grading efficiency and low grading accuracy.

Description

Technical Field

[0001] This invention relates to the field of agricultural machinery technology, specifically to a visual inspection-based equipment and method for grading the maturity of water chestnuts. Background Technology

[0002] Water chestnuts, an aquatic vegetable with a long history of consumption, are widely cultivated in the Yangtze River basin and areas south of it in my country due to their unique taste and rich nutritional value. With the rapid development of the water chestnut industry, water chestnut products also face the challenge of standardization, especially in the deep processing stage, where the consistency of raw materials directly determines the quality and production efficiency of the final product. Water chestnuts at different stages of maturity exhibit significant differences in taste, flavor, and processing characteristics: tender water chestnuts are sweet and crisp, suitable for fresh consumption or stir-frying, while older water chestnuts have a higher starch content, suitable for boiling or making flour. Therefore, to meet diverse market demands and achieve automation in processing lines, precise and efficient sorting of water chestnuts by maturity has become an urgent need for industrial upgrading.

[0003] Current methods for sorting water chestnut maturity mainly rely on manual screening or simple density grading. However, manual screening is subject to subjective factors, resulting in low grading accuracy and efficiency; while density grading is greatly affected by the size of the water chestnut's internal cavity and its moisture content, making it difficult to reflect its biological maturity and hindering subsequent deep processing. Therefore, there is an urgent need to provide a visual inspection-based water chestnut maturity grading device and method to solve the above technical problems. Summary of the Invention

[0004] The purpose of this invention is to provide a visual detection-based water chestnut maturity grading device and method, which can solve the technical problems of low grading efficiency and low grading accuracy in existing water chestnut maturity grading technologies.

[0005] To achieve the above-mentioned technical objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a visual detection-based water chestnut maturity grading device, comprising: Feeding mechanism; A material conveying mechanism, wherein the inlet of the material conveying mechanism is connected to the feeding mechanism; The sorting mechanism is disposed on both sides of the conveying mechanism, and the sorting mechanism can push the rhombus on the conveying mechanism to move so as to separate it from other rhombuses; An image acquisition device is disposed facing the material conveying mechanism; The controller is connected to the image acquisition unit, the feeding structure, the conveying mechanism and the sorting mechanism respectively.

[0006] In some embodiments, the sorting mechanism includes a mounting base, a first driving member, and a push plate. The mounting base is fixed to the material conveying mechanism, the first driving member is disposed on the mounting base, and the push plate is pulsatorically connected to the first driving member. The push plate extends or retracts under the action of the first driving member to separate the target rhombus. The first driving member is signal-connected to the controller.

[0007] In some embodiments, the feeding mechanism includes a vibrator, a hopper, and a base plate. The outlet of the hopper is connected to one end of the base plate, and the other end of the base plate is connected to the inlet of the feeding mechanism. The vibrator is disposed at the bottom of the base plate and is used to vibrate the rhombuses in the hopper.

[0008] In some embodiments, the feeding mechanism further includes a U-shaped limiting plate and a brush. The U-shaped limiting plate is disposed on the base plate and located outside the hopper. The closed end of the U-shaped limiting plate is close to the hopper, and the brush is disposed on the base plate near the open end of the U-shaped plate.

[0009] In some embodiments, the feeding mechanism further includes partitions, and a plurality of partitions are arranged in parallel at intervals on the feeding mechanism and along the feeding direction of the feeding structure to form a plurality of transport channels.

[0010] In some embodiments, the material conveying mechanism includes a frame, a second drive member, and a conveyor belt. The second drive member is fixed on the frame, the conveyor belt is laid on the top of the frame and is connected to the second drive member in a transmission manner, the feed inlet of the conveyor belt is connected to the feeding mechanism, and the sorting mechanism is disposed on the frame on both sides of the conveyor belt.

[0011] In some embodiments, a receiving mechanism is also included, wherein a plurality of receiving mechanisms are disposed at the discharge port of the conveying mechanism for storing water chestnuts of different maturity levels.

[0012] Secondly, the present invention provides a visual detection-based method for grading the maturity of water chestnuts, implemented using the grading device provided in the first aspect of the present invention. The grading method includes the following steps: The image of the rhombus is acquired by the image acquisition device and transmitted to the conveying mechanism by the feeding mechanism. The controller receives and processes the rhombus image; The maturity level of the processed water chestnut image is identified by the water chestnut maturity recognition model pre-stored in the controller, and the recognition result is marked on the corresponding water chestnut image. The controller controls the sorting mechanism to sort and screen the target rhombuses based on the recognition results.

[0013] In some embodiments, receiving and processing the rhombus image via the controller includes: The acquired diamond-shaped image is converted to grayscale. The median filtering method was used to smooth the grayscale image of the rhombus. The white area of ​​the smoothed rhombus image is extracted using binarization to obtain the outline of a single rhombus image.

[0014] In some embodiments, receiving and processing the rhombus image via the controller further includes: The watershed method is used to crop the obtained single rhombus image outline to obtain the processed single rhombus image.

[0015] Compared with the prior art, the beneficial effects of the present invention mainly include: This invention provides a visual inspection-based water chestnut maturity grading device. Through the coordinated operation of a feeding mechanism, conveying mechanism, sorting mechanism, and controller, water chestnuts are automatically conveyed from the feeding mechanism via the conveying mechanism. During conveying, an image acquisition device captures images of the water chestnuts, using color as a standard for maturity identification and grading. Finally, the sorting mechanism is controlled to screen the water chestnuts according to the identification results. On the one hand, this automates the water chestnut grading process, significantly improving efficiency compared to existing manual grading methods. On the other hand, the use of visual recognition greatly improves grading accuracy compared to existing density-based grading methods. In summary, this invention enables automatic maturity detection and screening of large batches of water chestnuts, offering advantages such as high accuracy, high screening efficiency, and minimal damage to the water chestnuts. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this application, the accompanying drawings used in the embodiments will be briefly described below: Figure 1 This is an overall schematic diagram of the grading device described in this invention; Figure 2 This is a top view of the grading device described in this invention; Figure 3 This is a schematic diagram of the distribution mechanism described in this invention; Figure 4 This is a flowchart of the grading method described in this invention; Figure 5 This is a control process diagram of the grading method described in this invention; Figure 6 These are the original images acquired by this invention; Figure 7 This is the grayscale image obtained according to the present invention; Figure 8 This is the image after median filtering according to the present invention; Figure 9 This is the binarized image of the present invention; Figure 10 This is the image after edge detection according to the present invention; Figure 11 This is an image after corrosion and expansion according to the present invention; Figure 12 This is the image identified by the present invention.

[0017] As shown in the figure: 100. Feeding mechanism; 110. Vibrating device; 120. Hopper; 130. Base plate; 140. U-shaped limit plate; 150. Brush; 160. Partition plate. 200. Material conveying mechanism; 210. Frame; 220. Second drive component; 230. Conveyor belt; 300, Distributor mechanism; 310, Mounting base; 320, First drive component; 330, Push plate; 400. Image acquisition device; 500. Receiving mechanism. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] This invention addresses the problems of low screening efficiency and grading accuracy in water chestnut maturity grading using existing methods such as manual screening or simple density grading, which are easily affected by the size of the internal cavity and moisture content of the water chestnut. It proposes a visual inspection-based water chestnut maturity grading device and method. The grading device is characterized by high efficiency and simple structure; the grading method, based on visual inspection technology, uses water chestnut color as the standard for maturity grading, is unaffected by the size of the internal cavity and moisture content, and has the advantage of high accuracy.

[0020] like Figure 1-3As shown, in a first aspect, the present invention provides a visual detection-based water chestnut maturity grading device, comprising a feeding mechanism 100, a conveying mechanism 200, a sorting mechanism 300, an image acquisition device 400, and a controller. The feeding mechanism 100 is used to store water chestnuts to be graded and to convey water chestnuts to the conveying mechanism 200, the inlet of the conveying mechanism 200 being connected to the feeding mechanism 100. The sorting mechanism 300 is disposed on both sides of the conveying mechanism 200, and the sorting mechanism 300 is capable of pushing the water chestnuts on the conveying mechanism 200 to move and separate them from other water chestnuts. The image acquisition device 400 is disposed facing the conveying mechanism 200. The controller is signal-connected to the image acquisition device 400, the feeding mechanism 100, the conveying mechanism 200, and the sorting mechanism 300.

[0021] Thus, the grading equipment provided by this invention can control the feeding mechanism 100 to convey water chestnuts to the conveying mechanism 200 via a controller. Simultaneously, while the conveying mechanism 200 is conveying the water chestnuts, an image acquisition device 400 acquires images of the water chestnuts. The controller then uses color comparison to identify the maturity level of the acquired images. Finally, based on the identification result, the controller controls the sorting mechanism 300 to group water chestnuts of the same level together, achieving the purpose of water chestnut grading. In this process, the grading equipment provided by this invention achieves automated grading, is simple to operate, and greatly improves work efficiency.

[0022] In one preferred embodiment, the feeding mechanism 100 includes a vibrator 110, a hopper 120, and a base plate 130. The outlet of the hopper 120 is connected to one end of the base plate 130, and the other end of the base plate 130 is connected to the inlet of the conveying mechanism 200. The vibrator 110 is disposed at the bottom of the base plate 130 and is used to vibrate the rhombuses in the hopper 120.

[0023] It is understood that in this embodiment, the vibrating feeder 110 is a vibrating feeder, the base plate 130 is a rectangular plate, one end of the base plate 130 is fixed to the discharge port of the hopper 120, and the other end of the base plate 130 is connected to the feed inlet of the conveying mechanism 200. Thus, when the vibrating feeder 110 is started, it drives the rhombuses flowing from the hopper 120 onto the base plate 130 to vibrate. After vibrating onto the conveying mechanism 200, the conveying mechanism 200 continues to convey the rhombuses.

[0024] The adjustable vibration frequency range of the vibrator 110 is 90-120Hz. Z 100H is preferred ZThe water chestnuts are conveyed on the hopper 120, and the U-shaped limiting plate 140 makes the water chestnuts conveyed in two rows; the brush 150 adjusts its height so that the water chestnuts fall onto the conveyor belt 230 individually and at a frequency, and the speed range of the conveyor belt 230 is 0.15-0.25m / s, preferably 0.2m / s.

[0025] In one preferred embodiment, the feeding mechanism 100 further includes a U-shaped limiting plate 140 and a brush 150. The U-shaped limiting plate 140 is disposed on the bottom plate 130 and located outside the hopper 120. The closed end of the U-shaped limiting plate 140 is close to the hopper 130, and the brush 150 is disposed on the bottom plate 130 on the side close to the open end of the U-shaped plate 140.

[0026] In one preferred embodiment, the feeding mechanism 100 further includes partitions 160, and a plurality of partitions 160 are arranged in parallel at intervals on the conveying mechanism 200 and along the feeding direction of the conveying mechanism 200 to form multiple transport channels.

[0027] Thus, in this embodiment, the rhombuses coming out of the hopper 120 are automatically divided into two columns under the action of the U-shaped limiting plate 140. When they pass through the brush 150, they can be basically sorted into two columns of single-layer rhombuses by the brush 150 and continue to be conveyed forward on the conveying mechanism 200 through the two transport channels formed by the partition 160.

[0028] It should be noted that the installation height of the brush 150 is adjustable to limit the output frequency of the rhombus.

[0029] In one preferred embodiment, the material conveying mechanism 200 includes a frame 210, a second drive member 220, and a conveyor belt 230. The frame 210 is constructed by welding steel profiles together. The second drive member 220 is fixed to the frame 210. The conveyor belt 230 is laid on the top of the frame 210 and is connected to the second drive member 220 for transmission. The feed inlet of the conveyor belt 230 is connected to the feeding mechanism 100, specifically to the base plate 130. The sorting mechanism 300 is disposed on the frame 210 on both sides of the conveyor belt 230.

[0030] In one preferred embodiment, the sorting mechanism 300 includes a mounting base 310, a first driving member 320, and a push plate 330. The mounting base 310 is fixed to the material conveying mechanism 200. The first driving member 320 is disposed on the mounting base 310. The push plate 330 is pulsatorically connected to the first driving member 320. The push plate 330 extends or retracts under the action of the first driving member 320 to separate the target rhombus. The first driving member 320 is signal-connected to the controller.

[0031] In one preferred embodiment, the mounting base 310 is an L-shaped plate. One end of the mounting base 310 is fixed to the frame 210, and the other end of the mounting base 310 is located directly above the conveyor belt 230 for fixing the first driving member 320. The execution end of the first driving member 320 is connected to the push plate 330. Under the drive of the first driving member 320, the push plate can push the target material receiving and changing transport channel to the corresponding receiving mechanism 500. When the diamond is discharged from the conveying mechanism 200, it falls exactly into the corresponding receiving mechanism 500. The speed of the push plate 330 is preferably 165 mm / s.

[0032] In one preferred embodiment, the grading device further includes a receiving mechanism 500, and a plurality of receiving mechanisms 500 are disposed at the discharge port of the conveying mechanism 200 for storing water chestnuts of different maturity levels.

[0033] In one preferred embodiment, the controller includes a host computer and a slave computer. The slave computer is connected to the host computer via the Modbus RTU communication protocol to receive commands from the host computer to control the extension and retraction of the push plate 330 and the speed of the conveyor belt 230.

[0034] It should be noted that, since there is a time difference between the time when the conveyor belt 230 delivers the material to the pusher plate 330 and the time when the model gives the signal, the pusher plate 330 is set to start with a delay. The delay times for the two pushers 330 are 0.15s and 0.4s, respectively.

[0035] In addition, the image acquisition device 400 is an industrial camera, which is mounted on one side of the material conveying mechanism 200 via a bracket, and is used to acquire images of the rhombuses on the material conveying mechanism 200; furthermore, the bracket is adjustable to adjust the height and angle of the image acquisition device 400. The bracket and the image acquisition device 400 are mature technologies in the field, and will not be described in detail here.

[0036] With the development of large-scale algorithm models and visual inspection technologies, water chestnuts of different maturity levels can be classified according to color standards using deep learning models and visual inspection techniques. Based on this, this invention provides a visual inspection-based method for classifying water chestnut maturity.

[0037] like Figure 4 and Figure 5 As shown, in a second aspect, the present invention provides a visual detection-based method for grading the maturity of water chestnuts, implemented using the grading device provided in the first aspect of the present invention. The grading method includes the following steps: Step S1: The image acquisition device 400 acquires the image of the rhombus (e.g., the rhombus image transmitted from the feeding mechanism 100 to the conveying mechanism 200) via the image acquisition device 400. Figure 6 (as shown) Step S2, the controller receives and processes the rhombus image (e.g., Figure 7-11 (as shown) Step S3: The maturity level of the processed water chestnut image is identified using the water chestnut maturity recognition model pre-stored in the controller, and the recognition result is labeled on the corresponding water chestnut image (e.g., ...). Figure 12 (as shown) Step S4: The controller controls the sorting mechanism 300 to sort and screen the target diamonds based on the recognition results.

[0038] Thus, the grading method provided by the present invention is based on visual detection. Specifically, the image acquisition device 400 acquires images of the water chestnuts on the conveying mechanism 200, and then the controller processes the acquired images and compares the processed images with the models pre-stored in the controller to identify water chestnuts of different maturity levels. Finally, the sorting mechanism 300 is controlled to organize water chestnuts of the same maturity level into the same transport channel to achieve grading. Compared with the prior art, the grading method of the present invention is obviously more accurate.

[0039] In the above technical solution, after the target water chestnuts are screened by the sorting mechanism 300, water chestnuts of different maturity levels are graded and stored by different receiving structures 500 for subsequent deep processing.

[0040] Further, in step S2, processing the rhombus image through the controller specifically includes: Step S21, as follows Figure 7 As shown, the acquired rhombus image is converted to grayscale to enhance image contrast and make the image's feature information clearer. The grayscale conversion method is a weighted average method. , In the formula These are the weights for the R-value, G-value, and B-value, respectively. Step S22, as follows Figure 8 As shown, in order to suppress noise while preserving image edge information, median filtering is used to smooth the grayscale image of the rhombus to avoid interference from isolated noise points. , In the formula, Represents the original image; This represents the image after median filtering.

[0041] Step S23, as Figure 9 As shown, the white region of the smoothed rhombus image is extracted using binarization to obtain the outline of a single rhombus image. The white region (rhombus) is extracted using binarization, and the Otsu algorithm is used to achieve automatic segmentation of the rhombus image. , , ， ， ， , , In the formula Indicates the size of the image; This indicates the number of pixels whose grayscale value is lower than the threshold. Indicates the number of pixels that are higher than or equal to the threshold; This indicates the proportion of foreground pixels in the image; This indicates the proportion of background pixels to the total number of pixels in the image. This represents the variance between classes.

[0042] Step S24, as Figure 10 As shown, the Canny operator is used to capture edge information in an image. Its processing flow mainly includes four stages: First, the image is smoothed using a Gaussian filter to suppress high-frequency noise; then, the magnitude and direction of the gradient are calculated to locate potential edges; then, non-maximum suppression is used to remove false edges and refine lines; finally, a double threshold algorithm is used to connect and filter out the final edges.

[0043] Step S25, as Figure 11 As shown, erosion and dilation: Erosion and dilation, also known as opening and closing operations, mainly connect the internal regions of the same rhombus, remove noise, and fill in image holes.

[0044] Step S26: Use the watershed method to crop the water chestnut outline obtained from the image preprocessing to obtain the original image of a single water chestnut. Call the model training results to identify the maturity of the single water chestnut image and label the identification results on the original image.

[0045] Afterwards, the controller sends an execution command to the sorting mechanism 300 to perform the screening based on the identification results; and the water chestnuts with different maturity levels are collected into different receiving mechanisms 500 to achieve the grading of water chestnuts.

[0046] Furthermore, in step S3, the pre-existing diamond maturity recognition model in the controller is any one of four models: VGG19, ResNet50, DETR, and YOLO11, preferably the YOLO11 model.

[0047] Studies have shown that as water chestnuts mature, the degree of lignification of their outer shell deepens; that is, the harder the water chestnut, the darker its outer shell, and the more mature the water chestnut. Therefore, to improve the accuracy of water chestnut grading, this invention also fully considers the hardness of the water chestnut when training the water chestnut maturity recognition model pre-stored in the controller.

[0048] Specifically, the training process of the water chestnut maturity recognition model is as follows: Before training the model, the collected raw image data was amplified and divided into training, validation, and test sets in a 6:2:2 ratio. The dataset was then trained using four models: VGG19, ResNet50, DETR, and YOLO11.

[0049] To verify the correlation between the image features and hardness features of water chestnuts at different maturity levels, and considering the limitations of visual features on critical samples, this paper references the application of multimodal information fusion in agricultural product detection and introduces the physical hardness value of water chestnuts as supplementary evidence to verify the model based on the DS evidence theory multimodal fusion decision algorithm.

[0050] To transform the scalar hardness values ​​into probabilistic evidence compatible with neural network outputs, statistical modeling was performed on the hardness distributions of the "young" and "old" samples in the training set. Assuming the hardness values ​​of both classes follow a Gaussian distribution, the mean and standard deviation of the hardness for old and young water chestnuts were estimated based on the training set data.

[0051] For the hardness value h of the sample to be tested, calculate the likelihood probability of it belonging to each class, and perform normalization to construct the basic probability distribution of the hardness modes, denoted as . m hard Considering the potential noise in physical measurements, a variance scaling factor is introduced to adjust the width of the distribution, and a discount factor is also introduced. hard To control the weight of hardness evidence in the final decision:

[0052] In the formula This part represents a reservation about uncertainty.

[0053] After acquiring both visual and physical evidence, decision fusion is performed using the DS evidence theory synthesis rule. First, the classification logits output by the model backbone network are subjected to a softmax transformation with a temperature coefficient T to generate the probability distribution of the visual modality, and a visual discount factor is also introduced. img Construct a visual BPA, denoted as mimg Subsequently, using Dempster's synthesis rules, m img and img Perform orthogonalization and fusion to calculate joint trust degree m fused :

[0054]

[0055] Where K is the conflict coefficient, which measures the degree of contradiction between two modal pieces of evidence.

[0056] To achieve optimal fusion performance, this paper employs a grid search strategy on the validation set, considering factors such as temperature coefficient T and visual weights. img Hardness weight hard The optimal combination of hyperparameters is determined by jointly optimizing the variance scaling factor and the variance scaling factor.

[0057] Finally, the four models were trained using the DS evidence theory. After introducing the hardness features of water chestnuts through the DS evidence theory, the accuracy of the VGG19, ResNet50, DETR and YOLO11 models on the test set was almost the same as that of the water chestnut image prediction. This shows that the appearance features of water chestnuts are consistent with the hardness of water chestnuts, and also shows that it is feasible to predict the maturity of water chestnuts through the image features of water chestnuts.

[0058] In this embodiment, since there may be a few diamonds stacked together when they enter the conveying mechanism 200, the present invention segments the acquired diamond image through the above-mentioned preprocessing process to obtain a clear single diamond image, so as to facilitate identification and improve the accuracy of grading judgment.

[0059] In summary, the visual inspection-based water chestnut maturity grading device and method provided by this invention have the following two advantages: (a) Grading equipment 1. Simple and compact structure, high efficiency: The grading equipment of the present invention is mainly composed of three major modules connected in sequence: a feeding mechanism 100, a conveying mechanism 200, and a sorting mechanism 300. The overall structure is simple and easy to manufacture and maintain. Meanwhile, through the combined design of vibrating feeder 110, U-shaped limiting plate 140 and brush 150, automatic sorting of diamonds, two-column conveying and single fixed-frequency feeding are realized, preventing stacking and blockage, creating stable conditions for subsequent visual inspection and significantly improving processing efficiency. 2. Highly modular and adjustable: The frequency of the vibrator 110 involved in this invention can be adjusted between 90-120 Hz, and the speed of the conveyor belt 230 can be adjusted between 0.15-0.25 m / s. Combined with the adjustable height of the brush 150, it can adapt to different varieties, sizes and quantities of water chestnuts. (II) Grading Method 1. High accuracy and scientific basis: This method uses the external visual features of water chestnuts, such as color and texture, as the main criteria for judgment, and verifies that these external features are highly consistent with the internal hardness (degree of lignification) of water chestnuts. Compared with the density method, which relies on buoyancy (affected by the size of the cavity), visual detection can more directly and accurately reflect biological maturity. 2. A complete image preprocessing workflow: By employing grayscale conversion (enhancing contrast), median filtering (denoising and preserving edges), Otsu's binarization (automatic segmentation), Canny edge detection, morphological operations (denoising and hole filling), and watershed cropping, we ensure the accurate extraction of individual rhomboid contours from cluttered backgrounds, providing high-quality input for model recognition and improving recognition accuracy. 3. Non-destructive testing: The entire inspection process is based on machine vision, requiring no physical contact or damage to the rhombuses, minimizing material damage and preserving the integrity and marketability of the rhombuses.

[0060] The specific embodiments of the present invention described above do not constitute a limitation on the scope of protection of the present invention. Any other corresponding changes and modifications made in accordance with the technical concept of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A visual inspection-based water chestnut maturity grading device, characterized in that, include: Feeding mechanism; A material conveying mechanism, wherein the inlet of the material conveying mechanism is connected to the feeding mechanism; The sorting mechanism is disposed on both sides of the conveying mechanism, and the sorting mechanism can push the rhombus on the conveying mechanism to move so as to separate it from other rhombuses; An image acquisition device is disposed facing the material conveying mechanism; The controller is connected to the image acquisition unit, the feeding structure, the conveying mechanism and the sorting mechanism respectively.

2. The visual inspection-based water chestnut maturity grading device according to claim 1, characterized in that, The sorting mechanism includes a mounting base, a first driving member and a push plate. The mounting base is fixed on the material conveying mechanism. The first driving member is disposed on the mounting base. The push plate is connected to the first driving member in a transmission manner. The push plate extends or retracts under the action of the first driving member to separate the target rhombus. The first driving component is signal-connected to the controller.

3. The visual inspection-based water chestnut maturity grading device according to claim 1, characterized in that, The feeding mechanism includes a vibrator, a hopper, and a base plate. The outlet of the hopper is connected to one end of the base plate, and the other end of the base plate is connected to the inlet of the feeding mechanism. The vibrator is located at the bottom of the base plate and is used to vibrate the rhombuses in the hopper.

4. The visual inspection-based water chestnut maturity grading device according to claim 3, characterized in that, The feeding mechanism also includes a U-shaped limiting plate and a brush. The U-shaped limiting plate is disposed on the base plate and located outside the hopper. The closed end of the U-shaped limiting plate is close to the hopper, and the brush is disposed on the base plate near the open end of the U-shaped plate.

5. The visual inspection-based water chestnut maturity grading device according to claim 4, characterized in that, The feeding mechanism also includes partitions, and multiple partitions are arranged in parallel at intervals on the feeding mechanism and along the feeding direction of the feeding structure to form multiple transport channels.

6. The visual inspection-based water chestnut maturity grading device according to claim 1, characterized in that, The material conveying mechanism includes a frame, a second drive component, and a conveyor belt. The second drive component is fixed on the frame. The conveyor belt is laid on the top of the frame and is connected to the second drive component for transmission. The feed inlet of the conveyor belt is connected to the feeding mechanism. The sorting mechanism is set on the frame on both sides of the conveyor belt.

7. The visual inspection-based water chestnut maturity grading device according to claim 1, characterized in that, It also includes a receiving mechanism, with multiple receiving mechanisms disposed at the discharge port of the conveying mechanism for storing water chestnuts of different maturity levels.

8. A method for grading the maturity of water chestnuts based on visual detection, implemented using the grading device described in any one of claims 1-7, characterized in that, The grading method includes the following steps: The image of the rhombus is acquired by the image acquisition device and transmitted to the conveying mechanism by the feeding mechanism. The controller receives and processes the rhombus image; The maturity level of the processed water chestnut image is identified by the water chestnut maturity recognition model pre-stored in the controller, and the recognition result is marked on the corresponding water chestnut image. The controller controls the sorting mechanism to sort and screen the target rhombuses based on the recognition results.

9. The method for grading the maturity of water chestnuts based on visual detection according to claim 8, characterized in that, The process of receiving and processing the rhombus image via the controller includes: The acquired diamond-shaped image is converted to grayscale. The median filtering method was used to smooth the grayscale image of the rhombus. The white area of ​​the smoothed rhombus image is extracted using binarization to obtain the outline of a single rhombus image.

10. The method for grading the maturity of water chestnuts based on visual detection according to claim 9, characterized in that, The step of receiving and processing the rhombus image via the controller also includes: The watershed method is used to crop the obtained single rhombus image outline to obtain the processed single rhombus image.

Images