Plant raw material grading method and system based on image analysis

Abstract

This invention relates to the field of image processing technology, and more particularly to a method and system for grading plant raw materials based on image analysis. The method includes the following steps: acquiring an RGB image of the frozen plant raw material to be tested, converting it to the CIELAB color space to obtain a corresponding color texture map, and performing superpixel segmentation on the color texture map to obtain multiple superpixel blocks; obtaining a frost occlusion index for each superpixel block based on its brightness, gradient, and saturation features; obtaining a final quality score for each superpixel block based on the difference between its saturation and gradient features and the frost occlusion index; grouping superpixel blocks belonging to the same berry as a set of superpixel blocks; obtaining a weighted quality score for each set of superpixel blocks using the area of ​​the superpixel blocks as the weight; and grading each fruit based on the weighted quality score. This invention significantly improves the grading accuracy of frozen plant raw materials.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method and system for grading plant raw materials based on image analysis. Background Technology

[0002] In the plant extract industry, especially in the production of bioactive ingredients such as anthocyanins, the quality grading of raw materials (such as berries like blueberries and bilberries) is crucial. The integrity and maturity of the raw material's skin directly determine the extraction efficiency and product purity. In actual production, to preserve the bioactive components within the berries to the greatest extent possible, the raw materials are usually rapidly frozen immediately after harvesting and transported to the factory as frozen fruit for further processing.

[0003] Traditional visual detection algorithms rely on color space to extract the inherent color features (such as saturation) of the fruit peel to quantify ripeness and identify defects. However, the irregular ice crystals covering the surface of frozen fruit have high brightness and high frequency texture, which mask the original features of the fruit. Furthermore, the diffuse white reflection of the ice crystals causes irregular scattering of incident light, which greatly weakens the color expression of the fruit peel pigments. This not only flattens the color differences on the fruit surface but also significantly reduces the regional saturation, causing the dark features of ripe fruit to tend towards the light features of unripe fruit. At this point, the features extracted by the color space algorithm are no longer the true features of the fruit but false features after being interfered with by the ice crystals, resulting in extremely high false alarm and false negative rates.

[0004] Furthermore, the ice crystal texture formed by frost and the skin wrinkles caused by fruit shriveling defects are highly similar in image gradient features. Both show a significant increase in local gradient values, making it impossible to effectively distinguish the source of surface rough textures, ultimately causing the complete failure of fruit defect detection. Summary of the Invention

[0005] To address the technical problems that traditional color space algorithms may produce false color features due to the obscuring effect of ice crystals on the surface of frozen fruits and the interference of diffuse reflection, leading to false reports in quality assessment, and the inability to distinguish between gradient features similar to ice crystal textures and shriveled wrinkles, causing defect identification failure, this invention provides a plant raw material grading method and system based on image analysis.

[0006] In a first aspect, the present invention provides a method for grading plant raw materials based on image analysis, employing the following technical solution:

[0007] A method for grading plant raw materials based on image analysis, comprising the following steps:

[0008] RGB images of the frozen plant material to be tested are acquired and converted to the CIELAB color space to obtain the corresponding color texture map; the color texture map is then segmented into superpixels to obtain several superpixel blocks;

[0009] Based on the brightness, gradient, and color saturation of each superpixel, a frost occlusion index is obtained for each superpixel, which is used to characterize the degree to which the superpixel is covered by ice crystals; based on the difference between the color saturation and gradient of each superpixel and the frost occlusion index, a final quality score for each superpixel is obtained.

[0010] Connectivity components are labeled for all superpixel blocks to obtain each superpixel block set; the final quality score is weighted and averaged using the area of ​​each superpixel block in the superpixel block set as the weight to obtain the weighted quality score of each superpixel block set; the frozen plant raw materials corresponding to each superpixel block set are graded based on the weighted quality score.

[0011] The innovation of this invention lies in firstly constructing a frost occlusion index for each superpixel based on the characteristics of ice crystals—high brightness, high-frequency texture, and low saturation—to capture areas where the color is obscured by frost but are actually high-quality fruits, thus avoiding the misjudgment of unripe fruits due to light color in traditional algorithms. Secondly, based on the difference between the color saturation and gradient of each superpixel and the frost occlusion index, the final quality score of each superpixel is obtained. This effectively solves the problem of misjudgment caused by the similar texture of shriveled fruits and good, ice-covered fruits, and overcomes the problem that relying solely on texture or color in frozen environments cannot identify shriveled or unripe fruits, significantly improving the robustness and accuracy of the grading system.

[0012] Preferably, obtaining a plurality of superpixel blocks includes:

[0013] The color texture map is divided into several superpixel blocks by performing superpixel segmentation using a linear iterative clustering algorithm.

[0014] Preferably, obtaining the frost occlusion index for each superpixel block includes:

[0015] Get the color saturation of each pixel in the i-th superpixel block;

[0016] ;

[0017] In the formula, The frost occlusion index represents the i-th superpixel block; This represents the mean value of all pixels in the i-th superpixel block in the L channel; This represents the average gradient of all pixels in the i-th superpixel block; This represents the average color saturation of all pixels in the i-th superpixel block; Represents the adjustment constant; This represents the logarithmic function.

[0018] By introducing a frost occlusion index, it can be used to reflect the degree to which superpixel blocks are covered by ice crystals, which is convenient for subsequent use in frost textures and defect textures.

[0019] Preferably, obtaining the final quality score for each superpixel block includes:

[0020] ;

[0021] In the formula, Represents the final quality score of the i-th superpixel block; This represents the average color saturation of all pixels in the i-th superpixel block; This represents the average gradient of all pixels in the i-th superpixel block; The frost occlusion index represents the i-th superpixel block; Represents the frost texture mapping coefficient; represents the normalization constant; exp() represents the exponential function with the natural constant as the base; || represents the absolute value symbol.

[0022] When high texture is accompanied by a high frost index, it is judged as ice crystal interference and its quality score is retained. When high texture is accompanied by a low frost index, it is judged as a shriveled defect and the score is significantly reduced. This effectively solves the problem of misjudgment caused by the similar texture of shriveled fruit and good fruit covered with ice.

[0023] Preferably, the step of performing connected component labeling on all superpixel blocks to obtain each superpixel block set includes:

[0024] Using superpixel blocks as the basic unit, an eight-neighbor connected component labeling algorithm is used to label all superpixel blocks, thereby obtaining a set of several superpixel blocks in the image.

[0025] Preferably, obtaining the weighted quality score for each superpixel block set includes:

[0026] ;

[0027] In the formula, The weighted quality score representing the set of the k-th superpixel blocks; This represents the final quality score of the h-th superpixel in the set of k-th superpixel blocks; This represents the total number of pixels in the h-th superpixel block within the k-th superpixel block set; This represents the number of superpixel blocks in the k-th superpixel block set.

[0028] The final quality score is weighted and averaged using the area of ​​each superpixel block in the superpixel block set as the weight, thus avoiding the influence of local noise on the overall result.

[0029] Preferably, the step of grading the frozen plant materials corresponding to each superpixel block set based on the weighted quality score includes:

[0030] Preset preferred raw material threshold If the weighted quality score of any set of superpixel blocks is greater than or equal to the preferred raw material threshold, the fruit corresponding to that set of superpixel blocks is a high-quality raw material; if the weighted quality score of any set of superpixel blocks is less than or equal to the preferred raw material threshold, the fruit corresponding to that set of superpixel blocks is a low-quality raw material.

[0031] The system sends the grading result signal of each superpixel block set to the PLC controller. When it is determined that the inferior raw material passes through the end of the conveyor belt, the controller triggers the corresponding pneumatic nozzle to open and spray compressed air to accurately blow it away from the main track. The high-quality raw material falls naturally into the collection box with the conveyor belt.

[0032] It enables the grading of frozen raw materials by quality and, in conjunction with the actuator, achieves automated rejection.

[0033] Preferably, the acquisition of RGB images of the frozen plant material to be tested includes:

[0034] An industrial camera is mounted above the raw material conveyor belt, and a diffuse ring light source is used to acquire RGB images of the frozen plant raw materials to be tested.

[0035] Preferably, obtaining the color saturation of each pixel in the i-th superpixel block includes:

[0036] The square of the value of the j-th pixel in the i-th superpixel block in channel a is taken as the first square; the square of the value of the j-th pixel in the i-th superpixel block in channel b is taken as the second square; the sum of the square root of the first square and the second square is taken as the color saturation of the j-th pixel in the i-th superpixel block.

[0037] Secondly, the present invention provides a plant raw material grading system based on image analysis, employing the following technical solution:

[0038] A plant raw material grading system based on image analysis includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the aforementioned plant raw material grading method based on image analysis is implemented.

[0039] By adopting the above technical solution, a computer program for the above-mentioned plant raw material grading method based on image analysis is generated and stored in a memory so that it can be loaded and executed by a processor. In this way, a terminal device can be made based on the memory and the processor for convenient use.

[0040] This invention has the following technical effects: First, based on the characteristics of ice crystals—high brightness, high-frequency texture, and low saturation—it constructs an ice frost occlusion index by non-linearly combining brightness, gradient, and saturation. This index increases dramatically and non-linearly when the fruit is covered by thick frost, thus enabling extremely sensitive detection of areas where the color is obscured but which are actually high-quality fruits. This avoids the situation where traditional algorithms misjudge unripe fruits due to lighter color. Next, by analyzing the difference between gradient values ​​and the ice frost occlusion index, the final quality score is obtained. This effectively solves the problem of misjudgment caused by the similar texture of shriveled fruits and good ice-covered fruits, and overcomes the problem that relying solely on texture or color in frozen environments cannot accurately identify shriveled or unripe fruits. This significantly improves the robustness and accuracy of the grading system. Attached Figure Description

[0041] Figure 1 This is a flowchart of a plant raw material grading method based on image analysis according to an embodiment of the present invention;

[0042] Figure 2 This diagram illustrates the distribution of different raw material samples in a two-dimensional feature space composed of surface texture gradient values ​​and color saturation values.

[0043] Figure 3 A schematic diagram illustrating the decoupling effect of the present invention after introducing the frost shielding index;

[0044] Figure 4 A comparative diagram representing the accuracy of the classification. Detailed Implementation

[0045] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0046] This invention discloses a method for grading plant raw materials based on image analysis, referring to... Figure 1 This includes steps S1-S4:

[0047] S1: Acquire RGB images of the frozen plant material to be tested.

[0048] In this embodiment of the invention, an industrial camera is mounted above the raw material conveyor belt, and a diffuse ring light source is used to acquire RGB images of the frozen plant raw materials to be tested.

[0049] S2: Convert the RGB image of the frozen plant material to be detected into the CIELAB color space to obtain a color texture map, and perform superpixel segmentation on the color texture map to obtain multiple superpixel blocks; based on the brightness features, gradient features and saturation features of each superpixel block, obtain the frost occlusion index of each superpixel block.

[0050] It should be noted that, at the physical optics level, ice crystal coverings have typical characteristics of high reflectivity, strong texture, and low saturation. The grayscale thresholding method commonly used in existing technologies only uses a single grayscale feature for discrimination, which can easily confuse white frost with targets with similar grayscale features, such as light-colored fruits, resulting in insufficient detection accuracy.

[0051] To address this deficiency, this invention requires converting the RGB image of the frozen plant material to be tested into the CIELAB color space to obtain a color texture map. The color information is then decomposed into three independent dimensions: the L channel represents the brightness dimension (which can accurately quantify the high reflectivity of the ice crystal cover layer), and the a and b channels represent the red-green and yellow-blue color dimensions, respectively (the combination of the two can directly calculate the color saturation). Therefore, based on this dimensional separation feature, the brightness and saturation features required by this invention can be extracted. Next, the color texture map is divided into several superpixel blocks with similar color and texture features through superpixel segmentation. By constructing a nonlinear combination model that integrates the brightness, gradient, and saturation of each superpixel block, the frost occlusion index of each superpixel block is obtained, ultimately making the frost region highly distinguishable in the index space.

[0052] In this embodiment of the invention, the RGB image of the frozen plant material to be detected is converted into the CIELAB color space to obtain a color texture map of the frozen plant material to be detected; the color texture map is divided into several superpixel blocks by performing superpixel segmentation using the linear iterative clustering (SLIC) algorithm.

[0053] To obtain the color saturation of each pixel in the i-th superpixel block: take the square of the value of the j-th pixel in the a channel of the i-th superpixel block as the first square; take the square of the value of the j-th pixel in the b channel of the i-th superpixel block as the second square; take the sum of the square root of the first square and the second square as the color saturation of the j-th pixel in the i-th superpixel block.

[0054] Get the frost occlusion index for each superpixel:

[0055] ;

[0056] In the formula, The frost occlusion index represents the i-th superpixel block; This represents the mean value of all pixels in the i-th superpixel block in the L channel; This represents the average gradient of all pixels in the i-th superpixel block; This represents the average color saturation of all pixels in the i-th superpixel block; Represents the adjustment constant; Representing a logarithmic function, it suppresses the influence of extreme texture values, resulting in a smooth, non-linear growth.

[0057] In this embodiment of the invention, a preset adjustment constant is used. This is used to prevent calculation overflow caused by a denominator of 0, and at the same time to adjust the sensitivity to high saturation regions to avoid numerical instability due to extremely low saturation.

[0058] Because frost typically has a high gloss, while ripe berries or shriveled fruit have a low gloss, therefore A larger value indicates a more severe impact of frost on the i-th superpixel block, resulting in a higher frost occlusion index for the i-th superpixel block. Frost cover can obscure fruit color, leading to reduced saturation. Dried fruits have a duller color due to oxidation, resulting in lower saturation, while ripe fruits have high saturation. The smaller the value, the greater the frost occlusion index of the i-th superpixel block; This represents the roughness of the surface of the i-th superpixel. The larger the value, the more likely the i-th superpixel is covered by frost or is a dried-up fruit.

[0059] Therefore when and Simultaneously increase, and When decreasing, The index increases dramatically and non-linearly, meaning it can capture the typical frosting features that are bright and rough but whose colors are masked with extreme sensitivity.

[0060] S3: Based on the difference between the saturation features and gradient features of each superpixel block and the frost occlusion index, obtain the final quality score of each superpixel block; use the connected component labeling algorithm to classify superpixel blocks belonging to the same berry as each superpixel block set; use the area of ​​the superpixel block as the weight to obtain the weighted quality score of each superpixel block set.

[0061] It should be noted that the ripening process of fruit is accompanied by the synthesis and accumulation of pigments such as carotenoids and anthocyanins. The more mature the fruit, the more abundant its epidermal pigment content, and the higher the color saturation of the corresponding superpixel block. Therefore, the color saturation of the superpixel block can directly quantify the ripeness of the fruit and provide a basis for quality scoring. However, because frost can cover the fruit color, resulting in a decrease in color saturation, and shriveled fruit has a dull color due to oxidation, resulting in a lower color saturation, it is impossible to score the quality of each pixel block by relying solely on saturation.

[0062] Similarly, the gradient mean can only characterize the roughness of the fruit surface, but cannot distinguish whether the high texture of the fruit surface is caused by frost coverage or by the fruit's shriveling defect. Therefore, this invention introduces a frost occlusion index. Frost-derived textures are positively correlated with the frost index, while the feature difference between shriveling-derived textures and the frost index is unrelated. This constructs a texture attribute discrimination mechanism: if both the gradient texture and the frost occlusion index of a superpixel block are large, the texture is determined to be a frost interference texture, and a quality score is given based on the color saturation of the superpixel block; if the gradient texture of a superpixel block is large but the frost occlusion index is small, the texture is determined to be a real defect texture caused by the shriveling of the material, and a quality score is given after reducing the color saturation of the superpixel block, thereby accurately identifying inferior fruits.

[0063] In this embodiment of the invention, the final quality score of each superpixel block is obtained:

[0064] ;

[0065] In the formula, Represents the final quality score of the i-th superpixel block; This represents the average color saturation of all pixels in the i-th superpixel block; This represents the average gradient of all pixels in the i-th superpixel block; The frost occlusion index represents the i-th superpixel block; Represents the frost texture mapping coefficient; The symbol represents the normalization constant; exp() represents the exponential function with the natural constant as the base; || represents the absolute value symbol.

[0066] In this embodiment of the invention, a preset In other embodiments, implementers may pre-determine specific implementation methods. Value; Preset The frost texture mapping coefficient represents the coefficient that maps the frost occlusion index back to the gradient space; in other embodiments, implementers can preset the coefficient according to the specific implementation method. The value;

[0067] Its function is to identify the source of gradient textures, distinguishing between textures caused by frost and textures caused by defects such as shriveling. If the i-th superpixel belongs to a high-quality fruit area covered by frost, then the greater the surface roughness of the i-th superpixel and the greater the frost occlusion index, the better. The value is close to 0. The value is close to 1, and the final quality score of the i-th superpixel block is determined by... Decide;

[0068] If the i-th superpixel belongs to a region of shriveled fruit without frost cover, then the surface roughness of the i-th superpixel is greater, and the frost occlusion index is smaller. The larger the value, the better. The value is close to 0, for The value is reduced to accurately identify inferior products.

[0069] It should be noted that superpixel segmentation only divides the local regions of the image and cannot be directly associated with the physical entity of a single fruit. Therefore, this invention uses a connected component labeling algorithm to aggregate all superpixel blocks belonging to the same fruit into a superpixel block set, and each superpixel set uniquely corresponds to an independent fruit.

[0070] In this embodiment of the invention, superpixel blocks are used as basic units, and an eight-neighbor connected component labeling algorithm is used to label all superpixel blocks to obtain a set of several superpixel blocks in the image.

[0071] Obtain the weighted quality score for each set of superpixel blocks:

[0072] ;

[0073] In the formula, The weighted quality score representing the set of the k-th superpixel blocks; This represents the final quality score of the h-th superpixel in the set of k-th superpixel blocks; This represents the total number of pixels in the h-th superpixel block within the k-th superpixel block set; each superpixel block set forms a region corresponding to a fruit. The larger the value, the higher the quality of the raw material corresponding to the k-th superpixel block set.

[0074] S4: Grade each fruit based on the weighted quality score of each superpixel block set.

[0075] In this embodiment of the invention, a preset preferred raw material threshold is used. If the weighted quality score of any superpixel block set is greater than or equal to the preferred raw material threshold, it means that the fruit corresponding to the superpixel block set is plump and mature, and the fruit corresponding to the pixel block is a high-quality raw material.

[0076] If the weighted quality score of any superpixel block set is less than or equal to the preferred raw material threshold, it indicates that the fruit corresponding to the superpixel block set has severe shriveling or immaturity characteristics, and the fruit corresponding to the superpixel block set is a low-quality raw material.

[0077] The system sends the grading result signal of each superpixel block set to the PLC controller. When it is determined that the inferior raw material passes through the end of the conveyor belt, the controller triggers the corresponding pneumatic nozzle to open and spray compressed air to accurately blow it away from the main track. The high-quality raw material falls naturally into the collection box with the conveyor belt.

[0078] Figure 2 This diagram illustrates the distribution of different raw material samples in a two-dimensional feature space composed of surface texture gradient values ​​and color saturation values. High-quality raw materials are distributed in the upper left corner (low texture, high saturation), while raw materials covered by frost and shriveled defective raw materials are severely mixed and overlapped in the lower right corner (high texture, low saturation). This visually reveals that it is impossible to distinguish whether the surface roughness is due to icing (good fruit) or shriveling (bad fruit) based solely on texture and color, leading to a large number of misjudgments.

[0079] Figure 3 This diagram illustrates the decoupling effect of introducing the frost shading index in this invention. The frost-covered raw material sample points are clearly distributed diagonally along the frost influence trend line, showing a significant linear correlation. This indicates that the frost shading index accurately captures the frost characteristics. Conversely, the shriveled defect raw material sample points are isolated in the upper left corner, far from the frost influence trend line. This change in distribution pattern allows for the differentiation between frost-covered raw materials and shriveled defect raw materials by calculating the difference between the surface texture gradient value of the sample points and the frost shading index.

[0080] Figure 4 A comparative diagram representing the accuracy of the classification. Figure 4 Existing scoring methods often result in frost-covered raw materials being penalized below the acceptable threshold (misjudged as substandard). However, the scoring system of this invention shows that all good fruit covered in frost has its scores rebounded, rising above the acceptable threshold and being correctly identified as high-quality. Simultaneously, genuinely defective raw materials remain in the low score range. This clearly demonstrates that this invention can significantly recover high-quality raw materials that have been mistakenly damaged, without overlooking any defective fruit.

[0081] This invention also discloses a plant raw material grading system based on image analysis, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a plant raw material grading method based on image analysis provided by this invention.

[0082] The system also includes other components well-known to those skilled in the art, such as communication buses and communication interfaces, the setup and functions of which are known in the art and will not be described in detail here. In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0083] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for grading plant raw materials based on image analysis, characterized in that, include: RGB images of the frozen plant material to be tested are acquired and converted to the CIELAB color space to obtain the corresponding color texture map; The color texture map is segmented into superpixels to obtain several superpixel blocks; Based on the brightness, gradient, and color saturation of each superpixel block, the frost occlusion index of each superpixel block is obtained. The frost occlusion index is used to characterize the degree to which the superpixel block is covered by ice crystals. The method for obtaining the index is as follows: Get the color saturation of each pixel in the i-th superpixel block; ； In the formula, The frost occlusion index represents the i-th superpixel block; This represents the mean value of all pixels in the i-th superpixel block in the L channel; This represents the average gradient of all pixels in the i-th superpixel block; This represents the average color saturation of all pixels in the i-th superpixel block; Represents the adjustment constant; Represents the logarithmic function; The final quality score for each superpixel is obtained based on the difference between the color saturation and gradient of each superpixel and the frost occlusion index. Connectivity labeling is performed on all superpixel blocks to obtain a set of each superpixel block; The final quality score is weighted and averaged using the area of ​​each superpixel block in the superpixel block set as the weight, to obtain a weighted quality score for each superpixel block set; the frozen plant raw materials corresponding to each superpixel block set are graded based on the weighted quality score.

2. The method for grading plant raw materials based on image analysis according to claim 1, characterized in that, The acquisition of several superpixel blocks includes: The color texture map is divided into several superpixel blocks by performing superpixel segmentation using a linear iterative clustering algorithm.

3. The method for grading plant raw materials based on image analysis according to claim 1, characterized in that, The process of obtaining the final quality score for each superpixel block includes: ； In the formula, Represents the final quality score of the i-th superpixel block; This represents the average color saturation of all pixels in the i-th superpixel block; This represents the average gradient of all pixels in the i-th superpixel block; The frost occlusion index represents the i-th superpixel block; Represents the frost texture mapping coefficient; represents the normalization constant; exp() represents the exponential function with the natural constant as the base; || represents the absolute value symbol.

4. The method for grading plant raw materials based on image analysis according to claim 1, characterized in that, The process of performing connected component labeling on all superpixel blocks to obtain each superpixel block set includes: Using superpixel blocks as the basic unit, an eight-neighbor connected component labeling algorithm is used to label all superpixel blocks, thereby obtaining a set of several superpixel blocks in the image.

5. The method for grading plant raw materials based on image analysis according to claim 1, characterized in that, The process of obtaining the weighted quality score for each superpixel block set includes: ； In the formula, The weighted quality score representing the set of the k-th superpixel blocks; This represents the final quality score of the h-th superpixel in the set of k-th superpixel blocks; This represents the total number of pixels in the h-th superpixel block within the k-th superpixel block set; This represents the number of superpixel blocks in the k-th superpixel block set.

6. The method for grading plant raw materials based on image analysis according to claim 1, characterized in that, The grading process for the frozen plant materials corresponding to each superpixel block set based on the weighted quality score includes: Preset preferred raw material threshold If the weighted quality score of any set of superpixel blocks is greater than or equal to the preferred raw material threshold, the fruit corresponding to that set of superpixel blocks is a high-quality raw material; if the weighted quality score of any set of superpixel blocks is less than or equal to the preferred raw material threshold, the fruit corresponding to that set of superpixel blocks is a low-quality raw material. The system sends the grading result signal of each superpixel block set to the PLC controller. When it is determined that the inferior raw material passes through the end of the conveyor belt, the controller triggers the corresponding pneumatic nozzle to open and spray compressed air to accurately blow it away from the main track. The high-quality raw material falls naturally into the collection box with the conveyor belt.

7. The method for grading plant raw materials based on image analysis according to claim 1, characterized in that, The acquisition of RGB images of the frozen plant material to be tested includes: An industrial camera is mounted above the raw material conveyor belt, and a diffuse ring light source is used to acquire RGB images of the frozen plant raw materials to be tested.

8. The method for grading plant raw materials based on image analysis according to claim 1, characterized in that, The step of obtaining the color saturation of each pixel in the i-th superpixel block includes: The square of the value of the j-th pixel in the i-th superpixel block in channel a is taken as the first square; the square of the value of the j-th pixel in the i-th superpixel block in channel b is taken as the second square; the sum of the square root of the first square and the second square is taken as the color saturation of the j-th pixel in the i-th superpixel block.

9. A plant raw material grading system based on image analysis, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement a plant raw material grading method based on image analysis according to any one of claims 1-8.

Images