Mulberry leaf mass spectral image recognition device and AI grading decision-making method and system

Abstract

The invention discloses a mulberry leaf mass spectrum image recognition device and an AI grading decision-making method and system, and relates to the technical field of image recognition grading. Comprising the steps that demand item information is acquired based on a target mulberry leaf quality spectrum image recognition result, demand item items are obtained, and the demand item items are used for representing demand item information occupation data forming the mulberry leaf quality spectrum image recognition result; setting an initial weight ratio based on the demand item to obtain an initial weight item; the method comprises the following steps: acquiring demand data related to a target mulberry leaf quality spectrum image recognition result, extracting keyword features, setting a fluctuation increase and decrease value, and adjusting an initial weight through a combination result of the keyword features and the fluctuation increase and decrease value; therefore, the initial weight can be replaced and adjusted according to the actual demand data and matched with the demand, the grading result feedback of the target mulberry leaf mass spectrum image is obtained, and grading under different demand conditions is adapted.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition and grading, and in particular to a mulberry leaf quality spectrum image recognition device and an AI grading decision method and system. Background Art

[0002] The mulberry leaf quality spectral image is a visualization data of mulberry leaf quality information obtained through spectral imaging technology. It uses a spectrometer or spectral camera to capture the reflection or absorption characteristics of mulberry leaves under different wavelengths, such as visible light and near-infrared, so as to generate a multi-dimensional image reflecting the quality characteristics of mulberry leaves. The grading decision is made based on the recognition results of the mulberry leaf quality spectral image, which means that the quality of mulberry leaves is classified and evaluated based on spectral image analysis technology, and divided into different grades according to preset standards or models. The spectral information of mulberry leaves is obtained through hyperspectral imaging technology or other spectral analysis methods, and the mulberry leaves are classified into corresponding grades according to the matching degree between the recognition results and the grading standards. For example, if the spectral characteristics of a mulberry leaf meet the "high quality" standard, it will be classified as "high quality mulberry leaf".

[0003] Patent publication number CN104867159A is a digital camera sensor stain detection and grading method and device, which automatically detects the location and number of stains on industrial camera sensors and grades the current sensor; the detection process is effective and efficient. Combined with the designed lighting environment and parameter settings, the camera sensor directly images the light source embedded in the lens, with a single background and high imaging quality; at the same time, the camera exposure time is set to 100ms and the gain is 2dB; the accuracy of camera grading reaches 99.5%, which better meets the needs of production and customers.

[0004] When the above and similar technical solutions are used for grading, there are multiple key indicators related to the quality of mulberry leaves, such as chlorophyll content, water content, protein content, etc., and the judgment criteria for grading the quality of mulberry leaves are related to the specific content of the key indicators, and are judged by the comprehensive values ​​of multiple key indicators. However, different grading requirements have different requirements for key indicators, and the judgment criteria for analysis are also different. Under different grading requirements, the weights of key indicators will also be different. Therefore, it is necessary to adjust the weight ratio of key indicators according to the actual grading requirements, otherwise the quality grading results will be too general and not suitable for situations with fixed grading requirements. Summary of the invention

[0005] The purpose of the present invention is to provide a mulberry leaf quality spectrum image recognition device and an AI grading decision method and system to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: an AI grading decision method for mulberry leaf quality spectrum image, comprising:

[0007] Based on the target mulberry leaf quality spectrum image recognition result, the demand project information is obtained to obtain the demand project item, and the demand project item is used to represent the demand project information occupancy data constituting the mulberry leaf quality spectrum image recognition result;

[0008] An initial weight ratio is set based on the demand item to obtain an initial weight item, which is used to represent a weight coefficient for allocating demand item information;

[0009] Obtain demand data, extract keyword information of the demand data, and obtain specified feature items, where the specified feature items are used to represent the impact characteristics of the demand data on the demand project items;

[0010] Set the fluctuation increase or decrease value, and increase or decrease the weight of the demand project items based on the combination of the specified feature items and the fluctuation increase or decrease value to obtain the specified weight item;

[0011] Acquire the attribute information of the target mulberry leaf, and acquire the standard value of the attribute information based on the attribute information of the target mulberry leaf to obtain the attribute standard item;

[0012] Based on the attribute standard items, the demand project information is divided into at least three levels to obtain the divided level items, and scores are set based on the divided level items. Based on the matching results of the demand project items, the target score values ​​are obtained to obtain the indicator score items;

[0013] The grading range is set, and based on the combined results of the specified weight items and indicator scoring items, the grading result feedback of the target mulberry leaf quality spectral image is obtained to adapt to the grading under different demand conditions.

[0014] Furthermore, the demand item information includes a chlorophyll content index, a water content index, and a protein content index, and the method for obtaining the demand item includes:

[0015] Acquire the spectral image data of the target mulberry leaf, obtain the target spectral item, denoise the target spectral item, and improve the signal-to-noise ratio;

[0016] Based on the target spectral item, an acquisition range is set, and a valid area in the target spectral item is determined based on the acquisition range to avoid edge or background interference;

[0017] Based on the target spectral item, the band information is selected to obtain the target band item, and the spectral reflectance value is extracted based on the target band item as the feature input to obtain the input feature item

[0018] Acquire sample data of known concentration to obtain target training items, perform training with the target training items, determine the relationship between the spectral characteristics of different components and their contents, and obtain relationship mapping items;

[0019] Based on the comparison results between the relationship mapping items and the target spectrum items, the demand project items are obtained.

[0020] Furthermore, the method for obtaining the initial weight item includes: obtaining a target quantity item based on the number of projects included in the demand project information, obtaining a project proportion item by evenly allocating the weight proportion based on the target quantity item, and obtaining an initial weight item by combining the project proportion items.

[0021] Furthermore, the demand data includes remark information, and the method for obtaining the specified feature items includes:

[0022] Based on the demand data, the remark information data corresponding to the target mulberry leaf is obtained to obtain the target information item;

[0023] Based on the target information item, keyword information is extracted to obtain key feature items. Based on the demand item information, feature information corresponding to the demand item information in the key feature items is obtained to obtain the target feature item, and the target feature item is set as the designated feature item.

[0024] Furthermore, the method for obtaining the specified weight item includes:

[0025] Set the judgment level, which is at least two levels, and the fluctuation increase and decrease value is at least two fixed values, which correspond to the judgment levels respectively;

[0026] Based on the judgment level, judgment factors are set, including keyword factors and tone factors. Based on the judgment factors, the level of the specified feature item is judged to obtain the feature level item, and then the feature increase or decrease value is obtained;

[0027] The demand project items are weighted and combined based on the feature increase and decrease values ​​to obtain the specified weight items.

[0028] Furthermore, the attribute information of the target mulberry leaf includes growth cycle information, and the target mulberry leaf is provided with storage information. The method for obtaining the attribute standard item includes:

[0029] Based on the stored information, the growth cycle information of the target mulberry leaf is obtained to obtain the target cycle item;

[0030] Based on the target cycle item, the average numerical information of the target mulberry leaf in the demand project information is obtained through data to obtain the attribute standard item.

[0031] Furthermore, the attribute information of the target mulberry leaf also includes average state information, and the target mulberry leaf is provided with storage information. The method for obtaining the attribute standard item includes:

[0032] Based on the stored information, the growth cycle information of the target mulberry leaves is obtained to obtain the target cycle item. Based on the target cycle item, the acquisition range is set. Based on the acquisition range, the average numerical information of the target mulberry leaves in the demand item information is obtained to obtain the attribute standard item.

[0033] Furthermore, the method for obtaining the classification level items includes:

[0034] Based on the attribute information of the target mulberry leaf, an extremely high value and an extremely low value of the attribute information are obtained to obtain an attribute extreme value item, and difference information of the attribute extreme value item is obtained to obtain an attribute difference item;

[0035] A difference partition value is set, and an attribute combination item is obtained based on a combination result of the difference partition value and the attribute difference item;

[0036] Based on the positive and negative combination results of the attribute standard item and the attribute combination item, the attribute range item is obtained, and the attribute range items are set as classification level items respectively.

[0037] Furthermore, the mulberry leaf quality spectrum image AI grading decision system uses the above-mentioned mulberry leaf quality spectrum image AI grading decision method, including:

[0038] Analysis module: based on the target mulberry leaf quality spectrum image recognition result, obtain the demand project information, obtain the demand project items, set the initial weight ratio based on the demand project items, and obtain the initial weight item;

[0039] Adjustment module: obtain demand data, extract keyword information of demand data, obtain specified feature items, set fluctuation increase or decrease values, increase or decrease the weight of demand items based on the combination of specified feature items and fluctuation increase or decrease values, and obtain specified weight items;

[0040] Matching module: obtain the attribute information of the target mulberry leaf, and at the same time obtain the standard value of the attribute information based on the attribute information of the target mulberry leaf, obtain the attribute standard item, divide the demand project information into at least three levels based on the attribute standard item, obtain the division level item, set the score based on the division level item, obtain the target score value based on the matching result of the demand project item, obtain the indicator score item, set the grading range, and obtain the grading result feedback of the target mulberry leaf quality spectrum image based on the combination result of the specified weight item and the indicator score item.

[0041] Furthermore, the mulberry leaf quality spectrum image recognition device uses the above-mentioned mulberry leaf quality spectrum image AI grading decision method, including:

[0042] Spectral camera: used to capture the spectral image of mulberry leaves and obtain spectral information in different wavelength ranges;

[0043] Light source: Provide stable lighting conditions and use light sources with specific wavelengths;

[0044] Camera sensor: converts light signals into electrical signals;

[0045] Spectral sensors: Sensors specific to spectral imaging that capture and separate different wavelengths of light and are used in conjunction with filters to extract information about specific wavelengths;

[0046] Central processing unit: processes the collected data and images;

[0047] Graphics Processing Unit: Improves computing efficiency when performing image processing and deep learning model training;

[0048] Hard disk: used to store captured image data and subsequently generated spectral data;

[0049] Spectral analysis software: used to process and analyze the captured spectral data;

[0050] Data visualization tools: Display analysis results in the form of charts or graphs to facilitate understanding and interpretation of data;

[0051] Display: used to view real-time images and analysis results, and monitor data quality in real time.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] The mulberry leaf quality spectral image recognition device and AI grading decision method and system obtain the target mulberry leaf quality spectral image recognition result, set the demand project information, and the demand project information includes chlorophyll content index, moisture content index, protein content index, and set the initial weight according to the demand project information. At the same time, by obtaining the demand data related to the target mulberry leaf quality spectral image recognition result, by extracting keyword features, setting the fluctuation increase and decrease value, and adjusting the initial weight through the combination result of the keyword features and the fluctuation increase and decrease value, the initial weight can be replaced and adjusted to match the demand according to the actual demand data, so as to obtain the grading result feedback of the target mulberry leaf quality spectral image, and adapt to the grading under different demand conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a schematic diagram of the overall process of the present invention;

[0055] Figure 2 A schematic diagram of obtaining demand project information of the present invention;

[0056] Figure 3 It is a schematic diagram of the target feature items of the present invention;

[0057] Figure 4 It is a schematic diagram of the judgment level of the present invention;

[0058] Figure 5 It is a schematic diagram of the attribute difference item of the present invention. DETAILED DESCRIPTION

[0059] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] The quality of mulberry leaves is not determined by a single factor, but is affected by a combination of key indicators, such as chlorophyll content, water content, protein content, crude fiber content, and the content of certain specific elements, which are closely related to the nutritional value and applicability of mulberry leaves. For example, chlorophyll content directly reflects the photosynthesis efficiency of mulberry leaves, which in turn affects the synthesis of sugars and nutrients; water content affects the freshness and shelf life of mulberry leaves, which in turn affects the appetite of silkworms; protein content is an important source of nutrition for the growth and development of silkworms, and directly affects the yield and quality of silkworms. These indicators interact and restrict each other. Together they determine the overall quality of mulberry leaves. However, different sericulture application scenarios have significantly different requirements for mulberry leaf quality. For example, when raising young silkworms, it is usually necessary to select mulberry leaves with high leaf tenderness, moderate water content, and rich protein content to meet the nutritional needs of young silkworms for rapid growth. At this time, protein content and leaf tenderness may occupy a higher weight. When raising strong silkworms, more emphasis may be placed on mulberry leaves with moderate leaf fiber content and high sugar content to promote the formation of cocoons. At this time, the importance of fiber content and sugar content may be relatively increased. For example, in the production process of mulberry leaf tea, tea leaves are often used as raw materials for tea. The taste, aroma and nutritional value of leaves are the key, therefore, more attention may be paid to indicators related to tea quality such as chlorophyll and tea polyphenols. Since mulberry leaf quality is affected by a variety of key indicators, and different grading requirements have different emphasis on these indicators, in the mulberry leaf quality grading process, the weights of key indicators must be dynamically adjusted according to actual needs, otherwise it will lead to limitations and inapplicability of the grading results. Therefore, fixed grading standards and weight distribution methods cannot meet the personalized requirements of different grading needs. The technical solution provided by itself obtains the target mulberry leaf quality spectral image recognition results, sets the demand project information, and the demand project information includes chlorophyll content indicators, moisture content indicators, and protein content indicators, and sets the initial weight according to the demand project information. At the same time, by obtaining the demand data related to the target mulberry leaf quality spectral image recognition results, extracting keyword features, setting fluctuation increase and decrease values, and adjusting the initial weight through the combination of keyword features and fluctuation increase and decrease values, the initial weight can be replaced and adjusted to match the demand according to the actual demand data, thereby obtaining the grading result feedback of the target mulberry leaf quality spectral image, and adapting to the grading under different demand conditions, such as Figure 1 As shown, steps S100-S800 are included.

[0061] Step S100: Based on the target mulberry leaf quality spectrum image recognition result, the demand project information is obtained to obtain the demand project item.

[0062] It should be noted that the requirement item is used to represent the requirement item information occupancy data constituting the mulberry leaf quality spectrum image recognition result.

[0063] Specifically, the required project information includes chlorophyll content index, moisture content index, and protein content index. The method for obtaining the required project items includes: obtaining the spectral image data of the target mulberry leaf to obtain the target spectral item. A spectrometer is needed to collect the spectrum of the mulberry leaf. Usually, the suitable wavelength range is 400-2500nm, including visible light and near-infrared regions. Then, the target spectral item is denoised to improve the signal-to-noise ratio. The data of each wavelength is smoothed using a Savitzky-Golay filter, which can retain the characteristics of the spectrum and remove random noise; based on the target spectral item, an acquisition range is set, and the set acquisition range is a circular area with a radius of 1 cm at the center of the target mulberry leaf. Based on the acquisition range, the effective area in the target spectral item is determined to avoid edge or background interference; based on the target spectral item, the band information is selected to obtain the target band item. Due to the required project information It includes chlorophyll content index, water content index and protein content index. The absorption peaks of chlorophyll usually at 680nm and 750nm can be used to estimate the chlorophyll content. The absorption characteristics of water near 1450nm and 1900nm are related to the water content. Protein is generally related to the absorption characteristics in the range of 1700-1900nm. Therefore, the selected band information is: spectral values ​​of 680nm, 750nm, 1450nm, 1700nm and 1900nm are used as features, covering the above three. The spectral reflectance value is extracted based on the target band item as the feature input to obtain the input feature item. The sample data of known concentration is obtained to obtain the target training item. The target training item is used for training to determine the relationship between the spectral characteristics of different components and their content to obtain the relationship mapping item. Based on the comparison results of the relationship mapping item and the target spectral item, the demand project item is obtained.

[0064] Embodiment 1

[0065] In the specific implementation process, 10 mulberry leaf samples were collected in a laboratory, and their chlorophyll content, water content, and protein content were measured in mg / g, and the results are shown in Table 1;

[0066] Table 1

[0067]

[0068] The spectral image data of 10 mulberry leaf samples were obtained using a spectrometer, and the target spectral items were obtained, as shown in Table 2;

[0069] Table 2

[0070]

[0071] Use the Savitzky-Golay filter to smooth the data of each wavelength, which can retain the characteristics of the spectrum and remove random noise. Take the scip library in Python as an example, use:

[0072] The scipy.signal.savgol_filter function is used for smoothing. The window size is set to 5 and the order is set to 2. The same smoothing process is performed on the data of each wavelength to ensure that the spectral value of each point is filtered in the same way to maintain data consistency. The two wavelengths of 680nm and 750nm are mainly related to the absorption of the photosynthetic pigment chlorophyll. Chlorophyll has obvious absorption of red light and near-infrared light in the visible light range. This makes these two wavelengths the key features for estimating chlorophyll content. Water has a characteristic absorption peak in the near-infrared region. 1450nm and 1900nm are representative wavelengths of water content. The absorption characteristics of water molecules at these wavelengths make them important indicators for measuring the water content in samples. 1700nm is the characteristic absorption range of proteins and amino acids. In the mid-infrared region, many functional groups such as C=O, NH, etc. have absorption peaks near 1750nm. Therefore, choosing this wavelength is also very important for estimating protein content. Therefore, the selected target band items are: 680nm and 750nm for chlorophyll, 1450nm and 1900nm for water, and 1700nm for protein. The partial least squares regression model is established, and the spectral characteristics are regressed with the known chlorophyll, water and protein contents to obtain the following simple linear regression equation, that is, the relationship mapping items are:

[0073] Chlorophyll content (mg / g) = 2*spectral value (680nm) + 3*spectral value (750nm);

[0074] Moisture content (%) = 95*spectral value (1450nm) + 90*spectral value (1900nm);

[0075] Protein content (%) = 80*spectral value (1700nm);

[0076] like Figure 2 As shown, the target spectrum items obtained are 0.65, 0.45, 0.35, 0.2, and 0.55 at 680nm, 750nm, 1450nm, 1700nm, and 1900nm, respectively. At this time, according to the relationship mapping items, the required items are calculated as follows:

[0077] Chlorophyll content = 2*0.65+3*0.45=1.3+1.38=2.68mg / g;

[0078] Moisture content = 85*0.35+70*0.55=29.75+38.5=68.25%;

[0079] Protein content = 80*0.2 = 16%.

[0080] Step S200: setting initial weight ratios based on demand items to obtain initial weight items.

[0081] It should be noted that the initial weight item is used to represent the weight coefficient for allocating demand item information. The method for obtaining the initial weight item includes: obtaining the target quantity item based on the number of items included in the demand item information. Since the demand item information includes chlorophyll content index, moisture content index, and protein content index, there are three target quantity items. The weight proportions are evenly allocated based on the target quantity items to obtain the project proportion items. The project proportion items are combined to obtain the initial weight item, so the project proportion items are all 33.33%.

[0082] Step S300: Acquire demand data, extract keyword information of the demand data, and obtain designated feature items.

[0083] like Figure 3 As shown, it should be noted that the designated feature item is used to represent the impact characteristics of the demand data on the demand project item. The demand data includes remark information. The method for obtaining the designated feature item includes: based on the demand data, obtaining the remark information data corresponding to the target mulberry leaf to obtain the target information item; based on the target information item, extracting keyword information to obtain the key feature item; based on the demand project information, obtaining the feature information corresponding to the demand project information in the key feature item to obtain the target feature item, and setting the target feature item as the designated feature item.

[0084] Embodiment 2

[0085] In the specific implementation process, the remark information data corresponding to the target mulberry leaves are obtained: mainly consider the chlorophyll content, and the crude fiber content should also be appropriately increased. At this time, the target information item obtained is to mainly consider the chlorophyll content and appropriately increase the crude fiber content. Keyword extraction is performed based on the target information item, and the extracted key special items are "mainly chlorophyll" and "appropriately increase crude fiber". At this time, since the set demand item information includes chlorophyll content index, moisture content index, and protein content index, the characteristic information corresponding to the demand item information in the key feature item is "mainly chlorophyll", and then the target feature item is "mainly chlorophyll", and the specified feature item is obtained.

[0086] Step S400: Setting the fluctuation increase or decrease value, and based on the combination of the designated feature item and the fluctuation increase or decrease value, weighting the demand item to obtain the designated weight item.

[0087] It should be noted that the method for obtaining the specified weight item includes: setting a judgment level, the judgment level is at least two levels, and the fluctuation increase or decrease value is at least two fixed values, which correspond to the judgment level respectively; based on the judgment level, setting judgment factors, the judgment factors include keyword factors and tone factors, based on the judgment factors, performing level judgment on the specified feature items, obtaining feature level items, and then obtaining feature increase or decrease values; based on the feature increase or decrease values, weighting the demand project items to obtain the specified weight item.

[0088] Specifically, the judgment level is set to three levels, namely, mild, medium, and high, and the fluctuation increase and decrease values ​​are also set to three fixed values. The three fluctuation increase and decrease values ​​are set to 10%, 15%, and 20%, respectively, corresponding to the judgment level, and the judgment factors are set. Since the judgment factors include keyword factors and tone factors, the tone factors can be judged by punctuation marks. For example, an exclamation mark can strengthen the tone factor, and multiple exclamation marks can further strengthen the tone factor.

[0089] Embodiment 3

[0090] like Figure 4 As shown, in the specific implementation process, the designated feature item obtained is "chlorophyll is main". At this time, according to the judgment factors, the judgement point is the keyword factor, which is "main", the keyword factor is low, and there is a lack of tone factors. At this time, the judgment level of the designated feature item is set to slight, so the matching fluctuation increase or decrease value is 10%. Another designated feature item is "chlorophyll is the most important!". At this time, according to the judgment factors, the judgement point is the keyword factor and tone privacy. The keyword factor is "most important" and the tone factor is "!". Therefore, the judgment level of the designated feature item is set to high, and the matching fluctuation increase or decrease value is 20%.

[0091] Step S500: Acquire the attribute information of the target mulberry leaf, and acquire the standard value of the attribute information based on the attribute information of the target mulberry leaf to obtain the attribute standard item.

[0092] It should be noted that the attribute information of the target mulberry leaf includes growth cycle information. Storage information is set on the target mulberry leaf. The method for obtaining the attribute standard item includes: based on the storage information, obtaining the growth cycle information of the target mulberry leaf to obtain the target cycle item; based on the target cycle item, obtaining the average numerical information of the target mulberry leaf in the demand item information through data to obtain the attribute standard item.

[0093] In the specific implementation process, the storage information set on the target mulberry leaf indicates that the mulberry leaf is in the mature stage. The average values ​​of the chlorophyll content index, moisture content index and protein content index of the target mulberry leaf are obtained through data: the chlorophyll content of the mature mulberry leaf increases significantly, with an average value of 2.5 mg / g, the moisture content of the mature leaf decreases slightly, with an average value of 70%, the leaf texture is relatively tough, and the protein content of the mature leaf decreases, with an average value of 18%. At this time, the attribute standard item can be obtained.

[0094] It should be noted that the attribute information of the target mulberry leaf also includes average state information. The target mulberry leaf is provided with storage information. The method for obtaining the attribute standard item includes: based on the storage information, obtaining the growth cycle information of the target mulberry leaf to obtain the target cycle item; based on the target cycle item, setting the acquisition range; the acquisition range is to take the target mulberry leaf position as its actual feature point; setting the radius of 100m as the acquisition radius; the obtained circular area is the acquisition range; based on the acquisition range, the average numerical information of the target mulberry leaf in the demand item information is obtained to obtain the attribute standard item.

[0095] In the specific implementation process, the storage information set on the target mulberry leaf indicates that the mulberry leaf is in the mature stage. The average values ​​of the chlorophyll content index, moisture content index and protein content index of the mulberry leaves that are in the mature stage at the same time are obtained according to the acquisition range. The average chlorophyll content is 2.3 mg / g, the average moisture content is 65%, and the average protein content is 15%. At this time, the attribute standard items can be obtained.

[0096] Step S600: Based on the attribute standard items, the demand project information is divided into at least three levels to obtain the division level items.

[0097] It should be noted that the method for obtaining the classification level items includes: obtaining the extremely high value and the extremely low value of the attribute information based on the attribute information of the target mulberry leaf to obtain the attribute extreme value item, obtaining the difference information of the attribute extreme value item, and obtaining the attribute difference item; setting the difference division value, the difference division item is 25%, and obtaining the attribute combination item based on the combination result of the difference division value and the attribute difference item; obtaining the attribute range item based on the positive and negative combination results of the attribute standard item and the attribute combination item, and setting the attribute range items as classification level items respectively.

[0098] Embodiment 4

[0099] like Figure 5In the specific implementation process, the attribute standard items of the target mulberry leaves are obtained. Since the average numerical information of the target mulberry leaves in the demand project information is obtained through data, the items in the demand project information will have different high and low situations. According to the actual chlorophyll content, 2.5 mg / g, the moisture content is 70%, and the protein content is 18%. At the same time, when obtaining the attribute standard items, the extremely high and extremely low values ​​of the chlorophyll content are 2.7 and 2.3 mg / g, respectively, the extremely high and extremely low values ​​of the moisture content are 77% and 65%, respectively, and the extremely high and extremely low values ​​of the protein content are 22% and 13%, respectively. At this time, the result of the attribute difference item is: the chlorophyll content attribute difference item is 0.4 mg / g, the moisture content attribute difference item is 12%, and the protein content attribute difference item is 9%. The set difference division item is 25%. At this time, the difference division item and the attribute difference item are The combination results are: 0.1mg / g, 3%, 2.25%. At this time, the positive and negative combination results of the attribute standard item and the attribute combination item can obtain the attribute range item, that is, the attribute range items are: 2.5±0.1=(2.4,2.6), that is, the attribute range item of chlorophyll content is 2.3--2.4, 2.4--2.6, 2.6--2.7; 70%±3%=(67%, 73%), that is, the attribute range item of moisture content is 65%--67%, 67%--73%, 73%--77%; 18%±2.25%=(15.75%, 20.25%), that is, the attribute range item of protein content is 13%--15.75%, 15.75%--20.25%, 20.25%--22%. Therefore, the demand project information is divided into three levels, namely low level, medium level and high level.

[0100] Step S700: setting scores based on the classification level items, obtaining target score values ​​based on the matching results of the demand project items, and obtaining indicator score items.

[0101] It should be noted that since there are three classification items, the set scores are 1, 2, and 3, corresponding to low, medium, and high levels respectively. According to the requirements, the chlorophyll content in the project is 2.5 mg / g, the moisture content is 70%, and the protein content is 18%. At this time, 2.5 mg / g is between 2.4 and 2.6, 70% is between 67% and 73%, and 18% is between 15.75% and 20.25%, so the corresponding scores in the classification items are all 2.

[0102] Step S800: Set the grading range, and obtain the grading result feedback of the target mulberry leaf quality spectrum image based on the combination result of the specified weight item and the index scoring item.

[0103] It should be noted that since the demand project information is divided into three levels, namely low, medium and high, and the corresponding scores are 1, 2 and 3 respectively, the grading ranges are 0-1, 1-2 and 2-3 respectively.

[0104] Embodiment 5

[0105] In the specific implementation process, the designated feature item obtained is "chlorophyll is mainly". At this time, according to the judgment factors, the point that can be judged is the keyword factor, which is "mainly". The keyword factor is low and lacks the tone factor. At this time, the judgment level of the designated feature item is set to slight, so the matching fluctuation increase or decrease value is 10%, and in the initial weight items, the item proportion items are all 33.33%. Therefore, among the chlorophyll content index, moisture content index, and protein content index, the weight of the chlorophyll content index is 43.33%, and the weights of the moisture content index and protein content are 28.33%. According to the requirement items, the chlorophyll content is 2.5 mg / g, the moisture content is 70%, and the protein content is 18%. At this time, 2.5 mg / g is between 2.4-2.6, 70% is between 67%-73%, and 18% is between 15.75%-20.25%. Therefore, the scores in the corresponding classification items are all 2. Therefore, the grading result feedback of the target mulberry leaf quality spectral image is: 43.33%*2+28.33%*2+28.33%*2=0.8666+0.5666+0.5666=1.9998, and the feedback result is intermediate.

[0106] The mulberry leaf quality spectrum image recognition device and AI grading decision system use the above-mentioned mulberry leaf quality spectrum image recognition device and AI grading decision method, including: an analysis module: based on the target mulberry leaf quality spectrum image recognition result, obtain the demand project information, obtain the demand project item, set the initial weight ratio based on the demand project item, and obtain the initial weight item; an adjustment module: obtain the demand data, extract the keyword information of the demand data, obtain the specified feature item, set the fluctuation increase and decrease value, and increase and decrease the weight of the demand project item based on the combination result of the specified feature item and the fluctuation increase and decrease value to obtain the specified weight item; a matching module: obtain the attribute information of the target mulberry leaf, and at the same time obtain the standard value of the attribute information based on the attribute information of the target mulberry leaf to obtain the attribute standard item, divide the demand project information into at least three levels based on the attribute standard item to obtain the division level item, set the score based on the division level item, obtain the target score value based on the matching result of the demand project item, obtain the index score item, set the classification range, and obtain the classification result feedback of the target mulberry leaf quality spectrum image based on the combination result of the specified weight item and the index score item.

[0107] The mulberry leaf quality spectral image recognition device uses the above-mentioned mulberry leaf quality spectral image AI grading decision method, including: spectral camera: used to capture the spectral image of mulberry leaves and obtain spectral information in different wavelength ranges; light source: provide stable lighting conditions and use a light source of a specific wavelength; camera sensor: convert light signals into electrical signals; spectral sensor: a sensor specific to spectral imaging, which captures and separates light of different wavelengths and is used in conjunction with filters to extract information of specific wavelengths; central processing unit: processes the collected data and images; graphics processing unit: improves computing efficiency when performing image processing and deep learning model training; hard disk: used to store captured image data and subsequently generated spectral data; spectral analysis software: used to process and analyze captured spectral data; data visualization tool: displays the analysis results in the form of charts or graphs to facilitate understanding and interpretation of the data; display: used to view real-time images and analysis results, and monitor data quality in real time.

[0108] It should be noted that in actual use, check the working status of the spectral camera, light source, camera sensor and spectral sensor to ensure that they are operating normally, confirm that the drivers of the central processing unit and graphics processing unit have been correctly installed and updated, connect the spectral camera to the computing device, usually through USB, HDMI or a specific interface, turn on the light source, ensure that the light it emits is uniform and covers the entire shooting area, place the mulberry leaves in front of the spectral camera, and adjust the camera focus to ensure a clear image, use the spectral camera to shoot the mulberry leaves, obtain spectral information at different wavelengths, and try to capture multiple images for subsequent analysis, transfer the captured image data to the computing device for processing and storage, and store the data on the hard disk, use spectral analysis software to pre-process the image, including denoising, color correction and image enhancement, etc., to improve the accuracy of the analysis, extract spectral features in the spectral analysis software, usually including the intensity values ​​of each wavelength and other related features, use data visualization tools to visualize the analysis results, generate charts or graphs to more intuitively display the recognition results, view real-time feedback on the display, and adjust the data display method as needed.

[0109] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is limited by the attached embodiments and their equivalents.

Claims

1. Mulberry leaf quality spectral image AI grading decision method, characterized by: include: Based on the target mulberry leaf quality spectrum image recognition result, the demand project information is obtained to obtain the demand project item, and the demand project item is used to represent the demand project information occupancy data constituting the mulberry leaf quality spectrum image recognition result; An initial weight ratio is set based on the demand item to obtain an initial weight item, which is used to represent a weight coefficient for allocating demand item information; Obtain demand data, extract keyword information of the demand data, and obtain specified feature items, where the specified feature items are used to represent the impact characteristics of the demand data on the demand project items; Set the fluctuation increase or decrease value, and increase or decrease the weight of the demand project items based on the combination of the specified feature items and the fluctuation increase or decrease value to obtain the specified weight item; Acquire the attribute information of the target mulberry leaf, and acquire the standard value of the attribute information based on the attribute information of the target mulberry leaf to obtain the attribute standard item; Based on the attribute standard items, the demand project information is divided into at least three levels to obtain the divided level items, and scores are set based on the divided level items. Based on the matching results of the demand project items, the target score values ​​are obtained to obtain the indicator score items; The grading range is set, and based on the combined results of the specified weight items and indicator scoring items, the grading result feedback of the target mulberry leaf quality spectral image is obtained to adapt to the grading under different demand conditions.

2. The AI ​​grading decision method for mulberry leaf quality spectrum image according to claim 1 is characterized by: The demand item information includes a chlorophyll content index, a water content index, and a protein content index. The method for obtaining the demand item includes: Acquire the spectral image data of the target mulberry leaf, obtain the target spectral item, denoise the target spectral item, and improve the signal-to-noise ratio; Based on the target spectral item, an acquisition range is set, and a valid area in the target spectral item is determined based on the acquisition range to avoid edge or background interference; Based on the target spectral item, the band information is selected to obtain the target band item, and the spectral reflectance value is extracted based on the target band item as the feature input to obtain the input feature item Acquire sample data of known concentration to obtain target training items, perform training with the target training items, determine the relationship between the spectral characteristics of different components and their contents, and obtain relationship mapping items; Based on the comparison results between the relationship mapping items and the target spectrum items, the demand project items are obtained.

3. The AI ​​grading decision method for mulberry leaf quality spectrum image according to claim 1 is characterized by: The method for obtaining the initial weight item includes: obtaining a target quantity item based on the number of items included in the demand project information, obtaining a project proportion item by evenly allocating the weight proportion based on the target quantity item, and obtaining an initial weight item by combining the project proportion items.

4. The AI ​​grading decision method for mulberry leaf quality spectrum image according to claim 1 is characterized in that: The demand data includes remark information, and the method for obtaining the specified feature items includes: Based on the demand data, the remark information data corresponding to the target mulberry leaf is obtained to obtain the target information item; Based on the target information item, keyword information is extracted to obtain key feature items. Based on the demand item information, feature information corresponding to the demand item information in the key feature items is obtained to obtain the target feature item, and the target feature item is set as the designated feature item.

5. The AI ​​grading decision method for mulberry leaf quality spectrum image according to claim 1 is characterized in that: The method for obtaining the specified weight item includes: Set the judgment level, which is at least two levels, and the fluctuation increase and decrease value is at least two fixed values, which correspond to the judgment levels respectively; Based on the judgment level, judgment factors are set, including keyword factors and tone factors. Based on the judgment factors, the level of the specified feature item is judged to obtain the feature level item, and then the feature increase or decrease value is obtained; The demand project items are weighted and combined based on the feature increase and decrease values ​​to obtain the specified weight items.

6. The AI ​​grading decision method for mulberry leaf quality spectrum image according to claim 1 is characterized by: The attribute information of the target mulberry leaf includes growth cycle information, and the target mulberry leaf is provided with storage information. The method for obtaining the attribute standard item includes: Based on the stored information, the growth cycle information of the target mulberry leaf is obtained to obtain the target cycle item; Based on the target cycle item, the average numerical information of the target mulberry leaf in the demand project information is obtained through data to obtain the attribute standard item.

7. The AI ​​grading decision method for mulberry leaf quality spectrum image according to claim 1 is characterized by: The attribute information of the target mulberry leaf also includes average state information. The target mulberry leaf is provided with storage information. The method for obtaining the attribute standard item includes: Based on the stored information, the growth cycle information of the target mulberry leaves is obtained to obtain the target cycle item. Based on the target cycle item, the acquisition range is set. Based on the acquisition range, the average numerical information of the target mulberry leaves in the demand item information is obtained to obtain the attribute standard item.

8. The AI ​​grading decision method for mulberry leaf quality spectrum image according to claim 1 is characterized by: The method for obtaining the classification level items includes: Based on the attribute information of the target mulberry leaf, an extremely high value and an extremely low value of the attribute information are obtained to obtain an attribute extreme value item, and difference information of the attribute extreme value item is obtained to obtain an attribute difference item; A difference partition value is set, and an attribute combination item is obtained based on a combination result of the difference partition value and the attribute difference item; Based on the positive and negative combination results of the attribute standard item and the attribute combination item, the attribute range item is obtained, and the attribute range items are set as classification level items respectively.

9. Mulberry leaf quality spectrum image AI grading decision system, characterized by: The mulberry leaf quality spectrum image AI grading decision method according to any one of claims 1 to 8 is used, comprising: Analysis module: based on the target mulberry leaf quality spectrum image recognition result, obtain the demand project information, obtain the demand project items, set the initial weight ratio based on the demand project items, and obtain the initial weight item; Adjustment module: obtain demand data, extract keyword information of demand data, obtain specified feature items, set fluctuation increase or decrease values, increase or decrease the weight of demand items based on the combination of specified feature items and fluctuation increase or decrease values, and obtain specified weight items; Matching module: obtain the attribute information of the target mulberry leaf, and at the same time obtain the standard value of the attribute information based on the attribute information of the target mulberry leaf, obtain the attribute standard item, divide the demand project information into at least three levels based on the attribute standard item, obtain the division level item, set the score based on the division level item, obtain the target score value based on the matching result of the demand project item, obtain the indicator score item, set the grading range, and obtain the grading result feedback of the target mulberry leaf quality spectrum image based on the combination result of the specified weight item and the indicator score item.

10. Mulberry leaf quality spectrum image recognition device, characterized in that: The mulberry leaf quality spectrum image AI grading decision method according to any one of claims 1 to 8 is used, comprising: Spectral camera: used to capture the spectral image of mulberry leaves and obtain spectral information in different wavelength ranges; Light source: Provide stable lighting conditions and use light sources with specific wavelengths; Camera sensor: converts light signals into electrical signals; Spectral sensors: Sensors specific to spectral imaging that capture and separate different wavelengths of light and are used in conjunction with filters to extract information about specific wavelengths; Central processing unit: processes the collected data and images; Graphics Processing Unit: Improves computing efficiency when performing image processing and deep learning model training; Hard disk: used to store captured image data and subsequently generated spectral data; Spectral analysis software: used to process and analyze the captured spectral data; Data visualization tools: Display analysis results in the form of charts or graphs to facilitate understanding and interpretation of data; Display: used to view real-time images and analysis results, and monitor data quality in real time.

Images