Nondestructive detection method and system for pineapple maturity based on deep convolutional neural network

Abstract

The invention discloses a nondestructive detection method and system for pineapple maturity based on a deep convolutional neural network, and the invention relates to the technical field of fruit and vegetable detection. The method comprises the following steps: obtaining a pineapple appearance image with known surface texture characteristic parameters and color characteristic parameters, and preprocessing the image to form a sample image. A convolutional neural network prediction model is established based on the sample image, and the pineapple appearance image is input and the corresponding characteristic parameters are output. The appearance image of the pineapple to be detected is collected, and after preprocessing, the image is input into the prediction model to obtain the characteristic parameters, and the maturity judgment coefficient is calculated. The ethylene and carbon dioxide concentrations of the pineapple to be detected are detected, and the maturity judgment coefficient is corrected to obtain an accurate judgment coefficient. The judgment threshold is dynamically corrected, and the accurate judgment coefficient is compared with the dynamic threshold, and different maturity judgment results are issued, providing a more accurate and comprehensive maturity evaluation, and providing a more efficient and accurate decision support.

Description

Technical Field

[0001] The present invention relates to the technical field of fruit and vegetable detection, and in particular to a non-destructive detection method and system for pineapple maturity based on a deep convolutional neural network. Background Art

[0002] With the continuous development of the global fruit market, pineapple, as a popular tropical fruit, is favored by more and more consumers. Consumers' demand for pineapples is not limited to quantity, but they pay more attention to its quality and maturity. The maturity of pineapples directly affects its flavor, taste and nutritional value. Therefore, how to accurately and quickly judge the maturity of pineapples has become a technical problem that needs to be solved in the circulation and sales of agricultural products. At present, traditional pineapple maturity detection methods mostly rely on manual evaluation, which is not only time-consuming and labor-intensive, but also difficult to guarantee the accuracy and consistency of the test results due to subjective factors. In addition, manual detection also faces problems such as low efficiency and high cost, which makes it difficult to meet the needs of modern agricultural production and market.

[0003] At present, image-based pineapple maturity assessment technology has been gradually studied and applied. Some studies have attempted to predict the maturity of pineapples by analyzing the color, texture, morphology and other features of the pineapple's appearance and combining it with machine learning algorithms. For example, researchers extracted features from the color features of pineapples (such as the distribution of green, yellow or orange) and surface textures (such as spots, lines, etc.), and then input these features into traditional machine learning models for training, thereby achieving the assessment of pineapple maturity. However, these methods still have certain limitations: on the one hand, traditional machine learning methods often require manual design of features when faced with high-dimensional image features, and the effect of feature selection is limited by the experience of experts.

[0004] In addition, the maturity of pineapple not only involves its appearance, but is also closely related to its internal biochemical characteristics and physiological changes. During the ripening process, pineapple releases ethylene gas, which is an important ripening signal. At the same time, changes in the hardness of the pineapple surface, changes in gas concentration, and the size and weight of the fruit will affect the maturity. In order to accurately judge the maturity of pineapples, it is necessary to comprehensively consider multiple factors such as appearance, gas characteristics and physical parameters, which is still a difficult problem in the prior art. Traditional maturity determination methods have failed to achieve a comprehensive assessment of these complex factors, so there is a lot of room for improvement in the accuracy and efficiency of pineapple maturity determination.

[0005] In the prior art, the publication number CN110736709A discloses a non-destructive detection method for blueberry maturity based on a deep convolutional neural network, specifically, first picking blueberry samples of different maturity periods, collecting color image information of blueberry fruits before picking in each period, measuring chlorophyll for blueberries in each period, constructing a blueberry chlorophyll prediction content network BCPN, and inputting blueberry images, then performing frame marking and labeling to obtain a chlorophyll content prediction data set, training a chlorophyll content model and mapping the output chlorophyll content with the maturity of the fruit, and after mapping between values, judging the maturity of the current mapping value by the final prediction result mapping value. However, the relationship between chlorophyll content and the maturity of blueberries in this scheme may be affected by other factors (such as environmental conditions, variety differences, etc.). Simply relying on chlorophyll content to predict maturity may lead to a decrease in the accuracy and reliability of the model. Therefore, relying only on a set of detection systems and a set of judgment logic algorithms for general detection not only wastes resources, but also reduces the accuracy and effectiveness of the detection system.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0007] The object of the present invention is to provide a non-destructive detection method and system for pineapple maturity based on deep convolutional neural network to solve the problems raised in the above background technology.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A non-destructive detection method for pineapple maturity based on deep convolutional neural network, the specific steps include:

[0010] Acquire several pineapple appearance images with known surface texture characteristic parameters and color characteristic parameters, preprocess the acquired pineapple appearance images, and use the preprocessed images as sample images, wherein the surface texture characteristic parameters include peel concave depth and texture contrast, and the color characteristic parameters include peel green area and peel yellow area;

[0011] A convolutional neural network prediction model is established based on the sample images, the sample images are used as inputs of the convolutional neural network prediction model, and the surface texture feature parameters and color feature parameters corresponding to each sample image are used as labels to train the convolutional neural network prediction model, so as to obtain a feature analysis prediction model whose input is the pineapple appearance image and whose output is the surface texture feature parameters and color feature parameters corresponding to the input image;

[0012] Collecting the appearance image of the pineapple to be detected, pre-processing it and inputting it into the trained feature analysis prediction model, outputting the surface texture characteristic parameters and color characteristic parameters of the pineapple to be detected, and calculating and generating the pineapple maturity judgment coefficient based on the obtained surface texture characteristic parameters and color characteristic parameters of the pineapple to be detected;

[0013] Putting the pineapple to be tested into a closed environment, detecting the gas characteristics of the pineapple to be tested within a time period K, and correcting the pineapple maturity judgment coefficient based on the obtained gas characteristics to obtain an accurate pineapple maturity judgment coefficient, wherein the gas characteristics include ethylene and carbon dioxide concentrations in the closed environment within the time period K;

[0014] The physical parameters of the pineapple to be tested are collected to dynamically correct the initial maturity judgment threshold to obtain a dynamic maturity judgment threshold, and the precise pineapple maturity judgment coefficient is compared with the dynamic maturity judgment threshold. Different maturity judgment results are issued according to different comparison results. The physical parameters include height, average diameter and skin hardness.

[0015] Further, a plurality of pineapple appearance images with known surface texture characteristic parameters and color characteristic parameters are obtained, and the obtained pineapple appearance images are preprocessed, wherein the preprocessing includes noise reduction and enhancement processing and detail enhancement processing, and the method for performing noise reduction and enhancement processing on the pineapple appearance images is: performing noise reduction processing on the images by using a wavelet transform noise reduction method, and the specific steps of the wavelet transform noise reduction method include: decomposing the image by wavelet transform to obtain wavelet coefficients of the image at different scales and directions; performing threshold processing on the wavelet coefficients to set the low-amplitude wavelet coefficients to zero and retain the high-amplitude wavelet coefficients; performing inverse transformation on the wavelet coefficients after threshold processing, reconstructing the processed coefficients into an image, and completing image noise reduction processing;

[0016] Bilateral filtering is used to enhance the details of the pineapple appearance image. The specific filter transformation is based on the formula:

[0017]

[0018] Where y is the coordinate vector in the image coordinate system, I y is the gray value at the coordinate vector y, B y is the gray value I y The gray value after bilateral filtering transformation, G d and G r are all Gaussian functions, where G d and G r The formula is:

[0019]

[0020] Where x is the coordinate vector in the image coordinate system, I x is the gray value at the coordinate vector x, σ d and σ r G d and G r The standard deviation of .

[0021] Furthermore, a convolutional neural network prediction model is established based on the sample image, and a feature analysis prediction model is established based on the convolutional neural network, wherein the convolutional neural network is composed of an input layer, a convolutional layer, a pooling layer, a fully connected layer and an output layer, wherein the activation function in the convolutional layer is a ReLU function, and the specific expression of the ReLU function is:

[0022] ReLU(s l(p,q) )=max(0,s l(p,q) )

[0023] Among them, l represents the corresponding convolutional layer, s l(p,q) It represents the qth eigenvalue of the pth feature map in the lth corresponding convolutional layer;

[0024] For the fully connected layer, the number of neurons in the fully connected layer is set to 32, the initial neural network learning rate is set to 0.001, and the number of training rounds is 100.

[0025] Further, based on the obtained surface texture characteristic parameters and color characteristic parameters of the pineapple to be detected, a pineapple maturity judgment coefficient is calculated and generated, wherein the specific formula for calculating the pineapple maturity judgment coefficient is:

[0026]

[0027] Where CS is the pineapple maturity judgment coefficient, A ye is the yellow area of ​​the pineapple peel to be tested, CH represents the texture contrast of the pineapple to be tested, D is the depth of the pineapple peel depression, A ge is the green area of ​​the pineapple peel to be tested, A total To detect the total area of ​​pineapple peel.

[0028] Further, the pineapple to be detected is placed in a closed environment, and the gas characteristics of the pineapple to be detected in a time period K are detected, wherein the time period with a length of K is calibrated as [T1-K, T1], wherein T1 represents the current time, K represents the length of the time period, the unit is minute, and 30≤K≤75, K is a positive integer;

[0029] The pineapple maturity judgment coefficient is corrected by detecting the gas characteristics in the closed environment in the time period K to obtain the accurate pineapple maturity judgment coefficient, wherein the formula for calculating the accurate pineapple maturity judgment coefficient is as follows:

[0030]

[0031] Wherein, CS′ is the accurate pineapple maturity judgment coefficient, C2H4 is the difference in ethylene concentration between the initial and end times of the monitoring period in a closed environment, CO2 is the difference in carbon dioxide concentration between the initial and end times of the monitoring period in a closed environment, ω1 and ω2 are the weight coefficients of ethylene concentration and carbon dioxide concentration in a closed environment, respectively, where ω1>ω2 and ω1 and ω2 are both greater than 0.

[0032] Furthermore, the physical parameters of the pineapple to be detected are collected to dynamically correct the initial maturity judgment threshold to obtain a dynamic maturity judgment threshold, wherein a plurality of sampling areas are randomly selected within the entire height range of the pineapple to be detected, the diameter data of the sampling areas are obtained, and the average diameter of the pineapple to be detected is calculated based on the diameter data of the sampling areas. The specific calculation is based on the formula:

[0033]

[0034] Where, L o represents the sampling length in the oth sampling area, o is the index of the sampling area, and o∈[1,O], O is the number of sampling areas, D(q) o is the diameter of the pineapple at the qth position within the sampling length of the oth sampling area, q is the coordinate of the sampling point within the sampling length, Ra o is the diameter data of the oth sampling area, Ra is the average diameter of the pineapple to be tested;

[0035] The formula for calculating the dynamic maturity threshold is as follows:

[0036]

[0037] Where yz is the dynamic maturity judgment threshold, yz0 is the initial maturity judgment threshold, GH is the height of the pineapple to be tested, and S is the skin hardness of the pineapple to be tested.

[0038] Furthermore, the precise pineapple maturity judgment coefficient is compared with the dynamic maturity judgment threshold, and different maturity judgment results are issued according to different comparison results. The logic for issuing different maturity judgment results is as follows:

[0039] When CS′≥1.0*yz, the maturity of the pineapple to be tested is judged to be fully mature, indicating that the pineapple to be tested can be eaten directly;

[0040] When 0.5*yz≤CS′<1.0*yz, it is determined that the maturity of the pineapple to be tested is not fully mature, indicating that the pineapple to be tested is suitable for storage and transportation;

[0041] When 0≤CS′<0.5*yz, it is judged that the maturity of the pineapple to be tested is immature, which means that the pineapple to be tested is not suitable for picking and should continue to grow.

[0042] The present invention also provides a non-destructive detection system for pineapple maturity based on a deep convolutional neural network, and the non-destructive detection system for pineapple maturity based on a deep convolutional neural network is used to perform the non-destructive detection method for pineapple maturity based on a deep convolutional neural network, comprising:

[0043] The sample image acquisition and processing module is used to obtain a number of pineapple appearance images with known surface texture characteristic parameters and color characteristic parameters, preprocess the obtained pineapple appearance images, and use the preprocessed images as sample images, wherein the surface texture characteristic parameters include the pit depth and texture contrast of the peel, and the color characteristic parameters include the green area of ​​the peel and the yellow area of ​​the peel;

[0044] A prediction model training module is used to establish a convolutional neural network prediction model based on sample images, take the sample images as input of the convolutional neural network prediction model, and use the surface texture feature parameters and color feature parameters corresponding to each sample image as labels to train the convolutional neural network prediction model, so as to obtain a feature analysis prediction model whose input is a pineapple appearance image and whose output is the surface texture feature parameters and color feature parameters corresponding to the input image;

[0045] A maturity analysis module is used to collect the appearance image of the pineapple to be detected, pre-process it and input it into the trained feature analysis prediction model, output the surface texture characteristic parameters and color characteristic parameters of the pineapple to be detected, and calculate and generate the pineapple maturity judgment coefficient based on the obtained surface texture characteristic parameters and color characteristic parameters of the pineapple to be detected;

[0046] A gas characteristic correction module is used to place the pineapple to be detected in a closed environment, detect the gas characteristics of the pineapple to be detected in a time period K, and correct the pineapple maturity judgment coefficient based on the obtained gas characteristics to obtain an accurate pineapple maturity judgment coefficient, wherein the gas characteristics include ethylene and carbon dioxide concentrations in the closed environment in the time period K;

[0047] The maturity judgment module is used to collect the physical parameters of the pineapple to be tested to dynamically correct the initial maturity judgment threshold, obtain the dynamic maturity judgment threshold, compare the precise pineapple maturity judgment coefficient with the dynamic maturity judgment threshold, and issue different maturity judgment results according to different comparison results. The physical parameters include height, average diameter and skin hardness.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] Firstly, a prediction model was established by collecting a large number of pineapple appearance images with known maturity and extracting features such as skin depression depth, texture contrast, green area and yellow area. Through precise image processing technology and machine learning models, the appearance features of pineapples, including surface texture and color, were fully extracted to accurately predict maturity. The tedious operation and subjective errors in the traditional manual feature extraction process were avoided, greatly improving the automation and intelligence level of the system. Secondly, a method of real-time monitoring of gas characteristics was introduced, especially the concentration changes of ethylene and carbon dioxide in a closed environment, to dynamically correct the maturity judgment coefficient of pineapples. At the same time, the physical properties of the pineapples to be tested, such as height, average diameter and skin hardness, were dynamically adjusted to make the maturity judgment threshold more adaptable and accurate. Finally, the dynamic maturity judgment threshold was used for correction, so that the model could be adaptively adjusted according to the actual situation of the pineapple. This scheme significantly improves the adaptability and accuracy of the system under different growth environments. It can provide more accurate and comprehensive maturity assessment, and provide more efficient and accurate decision support for pineapple picking, transportation and storage. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a schematic diagram of the overall method flow of the present invention;

[0051] Figure 2 It is a schematic diagram of the overall system structure of the present invention. DETAILED DESCRIPTION

[0052] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0053] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0054] Example:

[0055] See also Figure 1 , the present invention provides a technical solution:

[0056] A non-destructive detection method for pineapple maturity based on deep convolutional neural network, the specific steps include:

[0057] Step 1: Acquire several pineapple appearance images with known surface texture characteristic parameters and color characteristic parameters, preprocess the acquired pineapple appearance images, and use the preprocessed images as sample images, wherein the surface texture characteristic parameters include the pit depth and texture contrast of the peel, and the color characteristic parameters include the green area of ​​the peel and the yellow area of ​​the peel.

[0058] Among them, a pineapple appearance image with several known surface texture characteristic parameters and color characteristic parameters is obtained, and the obtained pineapple appearance image is preprocessed, wherein the preprocessing includes noise reduction and enhancement processing and detail enhancement processing. The method for performing noise reduction and enhancement processing on the pineapple appearance image is: using a wavelet transform denoising method to denoise the image, and the wavelet transform denoising method specifically includes the following steps: decomposing the image through wavelet transform to obtain wavelet coefficients of the image at different scales and directions; performing threshold processing on the wavelet coefficients to set low-amplitude wavelet coefficients to zero and retain high-amplitude wavelet coefficients; performing inverse transform on the wavelet coefficients after threshold processing, reconstructing the processed coefficients into an image, and completing image denoising processing;

[0059] Bilateral filtering is used to enhance the details of the pineapple appearance image. The specific filter transformation is based on the formula:

[0060]

[0061] Where y is the coordinate vector in the image coordinate system, I y is the gray value at the coordinate vector y, B y is the gray value I y The gray value after bilateral filtering transformation, G d and G r are all Gaussian functions, where G d and G r The formula is:

[0062]

[0063] Where x is the coordinate vector in the image coordinate system, I x is the gray value at the coordinate vector x, σ d and σ r G d and G r The standard deviation of .

[0064] The specific method for obtaining the depth of the fruit peel depression is as follows: using image analysis software or programming tools (such as OpenCV) to perform edge detection and extract the outline of the pineapple peel; by calculating the shape characteristics of the outline, the depressed area is identified and quantified. Morphological operations (such as corrosion and expansion) can be used to enhance the extraction of depressed features; the depth of the extracted depressed area is measured, which can usually be calculated by comparing the height difference between the depressed area and the adjacent normal area.

[0065] The texture features of the pineapple peel image are calculated through methods such as gray level co-occurrence matrix (GLCM); the texture contrast is calculated, usually quantified through specific statistics of GLCM (such as contrast value), reflecting the fineness and changes of the texture in the image.

[0066] The green area is extracted from the preprocessed image using the color threshold segmentation method. Green pixels can be screened according to the specified HSV or Lab color range, and the area of ​​the green area is calculated by pixel counting. Usually, the number of green pixels in the green area is counted after binarization processing to obtain the green area of ​​the peel.

[0067] The method for obtaining the area of ​​the yellow region of the peel is the same as the method for obtaining the area of ​​the green region of the peel.

[0068] Step 2: Establish a convolutional neural network prediction model based on the sample image, take the sample image as the input of the convolutional neural network prediction model, and use the surface texture feature parameters and color feature parameters corresponding to each sample image as labels to train the convolutional neural network prediction model, and obtain a feature analysis prediction model with the input as the pineapple appearance image and the output as the surface texture feature parameters and color feature parameters corresponding to the input image.

[0069] A convolutional neural network prediction model is established based on the sample image, and a feature analysis prediction model is established based on the convolutional neural network, wherein the convolutional neural network consists of an input layer, a convolutional layer, a pooling layer, a fully connected layer and an output layer, wherein the activation function in the convolutional layer is a ReLU function, and the specific expression of the ReLU function is:

[0070] ReLU(s l(p,q) )=max(0,s l(p,q) )

[0071] Among them, l represents the corresponding convolutional layer, s l(p,q) It represents the qth eigenvalue of the pth feature map in the lth corresponding convolutional layer;

[0072] For the fully connected layer, the number of neurons in the fully connected layer is set to 32, the initial neural network learning rate is set to 0.001, and the number of training rounds is 100.

[0073] By using convolutional neural networks, large amounts of image data can be processed efficiently and features can be automatically extracted without human intervention. This process greatly improves the speed and accuracy of feature extraction and reduces labor costs and human resource investment. Convolutional neural networks have powerful feature extraction capabilities and can automatically learn and recognize complex patterns in images through multiple convolutional layers. Compared to traditional feature extraction methods (such as manually designed features), this can more comprehensively capture subtle changes in the surface texture and color characteristics of pineapples, thereby improving the accuracy of predictions.

[0074] Step 3: Collect the appearance image of the pineapple to be detected, pre-process it and input it into the trained feature analysis prediction model, output the surface texture feature parameters and color feature parameters of the pineapple to be detected, and calculate and generate the pineapple maturity judgment coefficient based on the obtained surface texture feature parameters and color feature parameters of the pineapple to be detected.

[0075] After the above-mentioned preprocessing, the appearance image of the pineapple to be detected is collected and input into the trained feature analysis prediction model, and the model outputs the surface texture feature parameters and color feature parameters of the pineapple to be detected.

[0076] Based on the obtained surface texture characteristic parameters and color characteristic parameters of the pineapple to be tested, the pineapple maturity judgment coefficient is calculated and generated, wherein the specific formula for calculating the pineapple maturity judgment coefficient is as follows:

[0077]

[0078] Where CS is the pineapple maturity judgment coefficient, A ye is the yellow area of ​​the pineapple peel to be tested, CH represents the texture contrast of the pineapple to be tested, D is the depth of the pineapple peel depression, A ge is the green area of ​​the pineapple peel to be tested, A total To detect the total area of ​​pineapple peel.

[0079] Among them, the pineapple maturity judgment coefficient CS combines the surface texture characteristic parameters and color characteristic parameters of the pineapple and is used to characterize the maturity of the pineapple. The larger the value of the pineapple maturity judgment coefficient CS, the higher the maturity of the pineapple.

[0080] It should be noted that the yellow area A of the pineapple peel to be tested ye The yellow area usually reflects the maturity of the pineapple. The more mature the pineapple, the larger the yellow area. Therefore, the area of ​​the yellow area on the peel is A. ye It is proportional to the pineapple maturity judgment coefficient CS. When the yellow area increases, the pineapple maturity judgment coefficient CS increases accordingly, indicating that the pineapple is more mature.

[0081] CH represents the texture contrast of the pineapple to be tested, which is used to measure the degree of change in the grayscale value of pixels in the image. The higher the contrast value, the more drastic the grayscale change in the image, that is, the rougher the image texture; the lower the contrast value, the smoother the texture. The surface texture of an immature pineapple is rough, and as it gradually matures, the surface texture gradually becomes smoother. Therefore, the texture contrast is inversely proportional to the pineapple maturity judgment coefficient CS. The larger its value, the lower the maturity of the pineapple. At the same time, through the exponential form e CH These results suggest that under high texture contrast, fruit reflecting immaturity may visually exhibit more complex textures.

[0082] The depth of the pit D of the peel is relatively tight and less pitted for pineapples with lower maturity, while the color of the peel of more mature pineapples will gradually change from green to yellow or golden yellow as the pineapples mature, and the peel may become looser, with some small pits or slight shrinkage. Therefore, the pit depth D of the peel is proportional to the pineapple maturity judgment coefficient CS. At the same time, the square term indicates that as the pit depth D of the peel increases, the speed at which the pineapple maturity judgment coefficient CS increases further.

[0083] The green area of ​​the pineapple peel to be tested is A ge Pineapples with lower maturity usually have more green areas, which can be measured by comparing the green area with the total skin area of ​​the tested pineapple. Reflects maturity, A ge The bigger, The larger the value, the lower the maturity. It is inversely proportional to the pineapple maturity judgment coefficient CS.

[0084] Step 4: Place the pineapple to be tested in a closed environment, detect the gas characteristics of the pineapple to be tested within a time period K, and correct the pineapple maturity judgment coefficient based on the obtained gas characteristics to obtain an accurate pineapple maturity judgment coefficient, wherein the gas characteristics include the ethylene and carbon dioxide concentrations in the closed environment within the time period K.

[0085] Put the pineapple to be detected into a closed environment, and detect the gas characteristics of the pineapple to be detected within a time period K, wherein the time period with a length of K is calibrated as [T1-K, T1], wherein T1 represents the current time, K represents the length of the time period, and the unit is minute, and 30≤K≤75, K is a positive integer;

[0086] The methods for detecting ethylene and carbon dioxide concentrations include: Electrochemical sensor: suitable for detecting carbon dioxide (CO2) concentration, with high sensitivity and selectivity. Generally used for low concentration detection. Infrared sensor: used for the detection of ethylene (C2H4) and carbon dioxide, using the absorption characteristics of gas to infrared light, it can achieve high-precision measurement. Gas chromatography is suitable for quantitative analysis of ethylene and carbon dioxide. Through sample extraction and injection into the gas chromatograph, the separation ability of different gases in the chromatographic column is used for quantification. This is a highly sensitive and selective laboratory method.

[0087] The pineapple maturity judgment coefficient is corrected by detecting the gas characteristics in the closed environment in the time period K to obtain the accurate pineapple maturity judgment coefficient, wherein the formula for calculating the accurate pineapple maturity judgment coefficient is as follows:

[0088]

[0089] Wherein, CS′ is the accurate pineapple maturity judgment coefficient, C2H4 is the difference in ethylene concentration between the initial and end times of the monitoring period in a closed environment, CO2 is the difference in carbon dioxide concentration between the initial and end times of the monitoring period in a closed environment, ω1 and ω2 are the weight coefficients of ethylene concentration and carbon dioxide concentration in a closed environment, respectively, where ω1>ω2 and ω1 and ω2 are both greater than 0.

[0090] Among them, the precise pineapple maturity judgment coefficient CS′ is used to further accurately judge the maturity of the pineapple through the released ethylene concentration and carbon dioxide concentration. The larger the value of the precise pineapple maturity judgment coefficient CS′, the higher the maturity of the pineapple.

[0091] Ethylene is an important plant hormone that plays a key role in the ripening process of fruits. It is involved in regulating the growth, development and ripening process of plants. During the ripening process of pineapples, the fruit will naturally release ethylene. The production of ethylene is usually related to biochemical changes in the fruit, such as the conversion of starch to sugar, reduced acidity and enhanced aroma. During the ripening stage, the concentration of ethylene released by pineapples will increase significantly. Therefore, an increase in ethylene concentration usually indicates an increase in the maturity of the pineapple. The ethylene concentration is proportional to CS′ and is expressed in an exponential form. It means that when the ethylene concentration exceeds a certain value, CS′ increases significantly, indicating a higher degree of maturity.

[0092] During the ripening process, pineapples will respire and release carbon dioxide. As the fruit matures, its respiration rate usually increases, and the concentration of carbon dioxide produced also rises. Changes in carbon dioxide concentration reflect the metabolic activity of the fruit. Ripe fruits usually have more active metabolism and release more carbon dioxide. During the ripening process, the increase in carbon dioxide concentration is usually positively correlated with the maturity of the fruit. Higher carbon dioxide concentrations indicate that the fruit is undergoing rapid physiological changes, indicating that it is more mature. The logarithmic form ln(1+CO2) shows that when the carbon dioxide concentration exceeds a certain range, the effect on maturity gradually weakens, indicating that when the pineapple is too ripe, the respiration rate will decrease.

[0093] Ethylene is widely considered to be a key ripening hormone in plants, which is directly involved in regulating the ripening process of fruits. Its increased concentration usually triggers a series of biochemical changes, such as increased sugar content, reduced acidity, and aroma release, which directly affect the maturity and sensory quality of fruits. In many fruits, especially fruits like pineapple, the release of ethylene increases significantly during the ripening period, and the relationship between its concentration and maturity is closer. Therefore, the influence of ethylene is usually more significant in the judgment of maturity, so ω1>ω2 is set and ω1 and ω2 are both greater than 0.

[0094] Step 5: Collect the physical parameters of the pineapple to be tested and dynamically correct the initial maturity judgment threshold to obtain a dynamic maturity judgment threshold. Compare the precise pineapple maturity judgment coefficient with the dynamic maturity judgment threshold, and issue different maturity judgment results according to different comparison results. The physical parameters include height, average diameter and skin hardness.

[0095] The method for obtaining the skin hardness is as follows: Ultrasonic technology can be used to measure the hardness of the fruit skin. The speed and attenuation of sound waves propagating inside the fruit are evaluated. An ultrasonic sensor is placed on the surface of the skin. The instrument sends sound waves and measures their return time and intensity. The hardness value is calculated based on the data transmitted back by the ultrasound.

[0096] The physical parameters of the pineapple to be detected are collected to dynamically correct the initial maturity judgment threshold to obtain a dynamic maturity judgment threshold, wherein a plurality of sampling areas are randomly selected within the entire height range of the pineapple to be detected, the diameter data of the sampling areas are obtained, and the average diameter of the pineapple to be detected is calculated based on the diameter data of the sampling areas. The specific calculation is based on the formula:

[0097]

[0098] Where, L o represents the sampling length in the oth sampling area, o is the index of the sampling area, and o∈[1,O], O is the number of sampling areas, D(q) ois the diameter of the pineapple at the qth position within the sampling length of the oth sampling area, q is the coordinate of the sampling point within the sampling length, Ra o is the diameter data of the oth sampling area, Ra is the average diameter of the pineapple to be tested;

[0099] The formula for calculating the dynamic maturity threshold is as follows:

[0100]

[0101] Where yz is the dynamic maturity judgment threshold, yz0 is the initial maturity judgment threshold, GH is the height of the pineapple to be tested, and S is the skin hardness of the pineapple to be tested.

[0102] Among them, the larger the average diameter Ra value of the tested pineapple is, the fuller the pineapple fruit is. The immature fruit is not fully developed and generally has a smaller diameter. Therefore, the ratio of the diameter to the height is used to determine the average diameter of the pineapple. The larger the average diameter Ra value of the pineapple to be tested is, the smaller the dynamic maturity judgment threshold should be, indicating that the fruit is ripe.

[0103] The skin hardness S of the pineapple to be tested is directly related to its maturity. The skin of a mature pineapple is softer, while the skin of an unripe pineapple is harder. In the formula, Represents the square root of the skin hardness. When the skin hardness S increases, the dynamic maturity judgment threshold will also increase accordingly. A high skin hardness means that the pineapple is not fully mature and the maturity judgment threshold needs to be increased. Pineapples with high skin hardness are usually still in the immature stage. Therefore, the greater the hardness, the higher the maturity judgment threshold, which means that the fruit may not be very mature.

[0104] The precise pineapple maturity judgment coefficient is compared with the dynamic maturity judgment threshold, and different maturity judgment results are issued according to different comparison results. The logic for issuing different maturity judgment results is as follows:

[0105] When CS′≥1.0*yz, the maturity of the pineapple to be tested is judged to be fully mature, indicating that the pineapple to be tested can be eaten directly;

[0106] When 0.5*yz≤CS′<1.0*yz, it is determined that the maturity of the pineapple to be tested is not fully mature, indicating that the pineapple to be tested is suitable for storage and transportation;

[0107] When 0≤CS′<0.5*yz, it is judged that the maturity of the pineapple to be tested is immature, which means that the pineapple to be tested is not suitable for picking and should continue to grow.

[0108] See also Figure 2The present invention also provides a non-destructive detection system for pineapple maturity based on a deep convolutional neural network, and the non-destructive detection system for pineapple maturity based on a deep convolutional neural network is used to perform the above-mentioned non-destructive detection method for pineapple maturity based on a deep convolutional neural network, comprising:

[0109] The sample image acquisition and processing module is used to obtain a number of pineapple appearance images with known surface texture characteristic parameters and color characteristic parameters, preprocess the obtained pineapple appearance images, and use the preprocessed images as sample images, wherein the surface texture characteristic parameters include the pit depth and texture contrast of the peel, and the color characteristic parameters include the green area of ​​the peel and the yellow area of ​​the peel;

[0110] A prediction model training module is used to establish a convolutional neural network prediction model based on sample images, take the sample images as input of the convolutional neural network prediction model, and use the surface texture feature parameters and color feature parameters corresponding to each sample image as labels to train the convolutional neural network prediction model, so as to obtain a feature analysis prediction model whose input is a pineapple appearance image and whose output is the surface texture feature parameters and color feature parameters corresponding to the input image;

[0111] A maturity analysis module is used to collect the appearance image of the pineapple to be detected, pre-process it and input it into the trained feature analysis prediction model, output the surface texture characteristic parameters and color characteristic parameters of the pineapple to be detected, and calculate and generate the pineapple maturity judgment coefficient based on the obtained surface texture characteristic parameters and color characteristic parameters of the pineapple to be detected;

[0112] A gas characteristic correction module is used to place the pineapple to be detected in a closed environment, detect the gas characteristics of the pineapple to be detected in a time period K, and correct the pineapple maturity judgment coefficient based on the obtained gas characteristics to obtain an accurate pineapple maturity judgment coefficient, wherein the gas characteristics include ethylene and carbon dioxide concentrations in the closed environment in the time period K;

[0113] The maturity judgment module is used to collect the physical parameters of the pineapple to be tested to dynamically correct the initial maturity judgment threshold, obtain the dynamic maturity judgment threshold, compare the precise pineapple maturity judgment coefficient with the dynamic maturity judgment threshold, and issue different maturity judgment results according to different comparison results. The physical parameters include height, average diameter and skin hardness.

[0114] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0115] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0116] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0117] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A nondestructive detection method for pineapple maturity based on deep convolutional neural network, characterized in that: The specific steps include: Acquire several pineapple appearance images with known surface texture characteristic parameters and color characteristic parameters, preprocess the acquired pineapple appearance images, and use the preprocessed images as sample images, wherein the surface texture characteristic parameters include peel concave depth and texture contrast, and the color characteristic parameters include peel green area and peel yellow area; A convolutional neural network prediction model is established based on the sample images, the sample images are used as inputs of the convolutional neural network prediction model, and the surface texture feature parameters and color feature parameters corresponding to each sample image are used as labels to train the convolutional neural network prediction model, so as to obtain a feature analysis prediction model whose input is the pineapple appearance image and whose output is the surface texture feature parameters and color feature parameters corresponding to the input image; Collecting the appearance image of the pineapple to be detected, pre-processing it and inputting it into the trained feature analysis prediction model, outputting the surface texture characteristic parameters and color characteristic parameters of the pineapple to be detected, and calculating and generating the pineapple maturity judgment coefficient based on the obtained surface texture characteristic parameters and color characteristic parameters of the pineapple to be detected; Place the pineapple to be tested in a closed environment and test it during the time period The gas characteristics of the pineapple to be detected in the time period are included in the gas characteristics, and the pineapple maturity judgment coefficient is corrected based on the obtained gas characteristics to obtain an accurate pineapple maturity judgment coefficient. Ethylene and carbon dioxide concentrations in the inner closed environment; The physical parameters of the pineapple to be tested are collected to dynamically correct the initial maturity judgment threshold to obtain a dynamic maturity judgment threshold, and the precise pineapple maturity judgment coefficient is compared with the dynamic maturity judgment threshold. Different maturity judgment results are issued according to different comparison results. The physical parameters include height, average diameter and skin hardness.

2. A non-destructive detection method for pineapple maturity based on a deep convolutional neural network according to claim 1, characterized in that: A pineapple appearance image with several known surface texture characteristic parameters and color characteristic parameters is obtained, and the obtained pineapple appearance image is preprocessed, wherein the preprocessing includes noise reduction and enhancement processing and detail enhancement processing. The method for performing noise reduction and enhancement processing on the pineapple appearance image is: performing noise reduction processing on the image by adopting a wavelet transform noise reduction method, and the specific steps of the wavelet transform noise reduction method include: decomposing the image by wavelet transform to obtain wavelet coefficients of the image at different scales and directions; performing threshold processing on the wavelet coefficients to set low-amplitude wavelet coefficients to zero and retain high-amplitude wavelet coefficients; performing inverse transformation on the wavelet coefficients after threshold processing, reconstructing the processed coefficients into an image, and completing image noise reduction processing; Bilateral filtering is used to enhance the details of the pineapple appearance image. The specific filter transformation is based on the formula: ； In the formula, is the coordinate vector in the image coordinate system, is the coordinate vector The gray value at Gray value The gray value after bilateral filtering transformation is and are all Gaussian functions, where and The formula is: ； ； In the formula, is the coordinate vector in the image coordinate system, is the coordinate vector The gray value at and They are and The standard deviation of .

3. A non-destructive detection method for pineapple maturity based on a deep convolutional neural network according to claim 2, characterized in that: A convolutional neural network prediction model is established based on the sample image, and a feature analysis prediction model is established based on the convolutional neural network, wherein the convolutional neural network consists of an input layer, a convolutional layer, a pooling layer, a fully connected layer and an output layer, wherein the activation function in the convolutional layer is function, The specific expression of the function is: ； in, Indicates The corresponding convolutional layers, It means the The corresponding convolutional layer The feature map eigenvalues; For the fully connected layer, the number of neurons in the fully connected layer is set to 32, the initial neural network learning rate is set to 0.001, and the number of training rounds is 100.

4. A non-destructive detection method for pineapple maturity based on a deep convolutional neural network according to claim 1, characterized in that: Based on the obtained surface texture characteristic parameters and color characteristic parameters of the pineapple to be tested, the pineapple maturity judgment coefficient is calculated and generated, wherein the specific formula for calculating the pineapple maturity judgment coefficient is as follows: ； In the formula, is the pineapple maturity judgment coefficient, is the yellow area of ​​the pineapple peel to be tested, represents the texture contrast of the pineapple to be detected, is the pit depth of the pineapple peel to be tested, is the green area of ​​the pineapple peel to be tested, is the total peel area of ​​the pineapple to be tested.

5. A non-destructive detection method for pineapple maturity based on a deep convolutional neural network according to claim 4, characterized in that: Place the pineapple to be tested in a closed environment and test it during the time period Gas characteristics of the pineapple to be tested; By detecting the time period The gas characteristics in the closed environment are used to correct the pineapple maturity judgment coefficient to obtain the accurate pineapple maturity judgment coefficient. The specific calculation formula for the accurate pineapple maturity judgment coefficient is as follows: ； In the formula, is the accurate pineapple maturity judgment coefficient, is the difference in ethylene concentration between the initial and final time of the monitoring period in a closed environment, is the difference between the carbon dioxide concentration at the beginning and end of the monitoring period in the closed environment, and are the weight coefficients of ethylene concentration and carbon dioxide concentration in a closed environment, respectively, where and and Both are greater than 0.

6. A non-destructive detection method for pineapple maturity based on a deep convolutional neural network according to claim 5, characterized in that: The physical parameters of the pineapple to be detected are collected to dynamically correct the initial maturity judgment threshold to obtain a dynamic maturity judgment threshold, wherein a plurality of sampling areas are randomly selected within the entire height range of the pineapple to be detected, the diameter data of the sampling areas are obtained, and the average diameter of the pineapple to be detected is calculated based on the diameter data of the sampling areas. The specific calculation is based on the formula: ； ； In the formula, Indicated in The sampling length of the sampling area, is the index of the sampling region, and , is the number of sampling areas, For the The sampling length of the sampling area The diameter of the pineapple, is the coordinate of the sampling point within the sampling length, For the The diameter data of the sampling area, is the average diameter of the pineapples to be tested; The formula for calculating the dynamic maturity threshold is as follows: ； In the formula, is the dynamic maturity judgment threshold, is the initial maturity judgment threshold, is the height of the pineapple to be tested, The hardness of the skin of the pineapple to be tested.

7. A non-destructive detection method for pineapple maturity based on a deep convolutional neural network according to claim 6, characterized in that: The precise pineapple maturity judgment coefficient is compared with the dynamic maturity judgment threshold, and different maturity judgment results are issued according to different comparison results. The logic for issuing different maturity judgment results is as follows: when When the pineapple to be tested is judged to be fully mature, it means that the pineapple to be tested can be eaten directly; when When the pineapple to be tested is judged to be not fully mature, it means that the pineapple to be tested is suitable for storage and transportation; when When the pineapple to be tested is judged to be immature, it means that the pineapple to be tested is not suitable for picking and should continue to grow.

8. A non-destructive detection system for pineapple maturity based on deep convolutional neural network, characterized in that: The non-destructive detection system for pineapple maturity based on a deep convolutional neural network is used to execute the non-destructive detection method for pineapple maturity based on a deep convolutional neural network according to any one of claims 1 to 7, comprising: The sample image acquisition and processing module is used to obtain a number of pineapple appearance images with known surface texture characteristic parameters and color characteristic parameters, preprocess the obtained pineapple appearance images, and use the preprocessed images as sample images, wherein the surface texture characteristic parameters include the pit depth and texture contrast of the peel, and the color characteristic parameters include the green area of ​​the peel and the yellow area of ​​the peel; A prediction model training module is used to establish a convolutional neural network prediction model based on sample images, take the sample images as input of the convolutional neural network prediction model, and use the surface texture feature parameters and color feature parameters corresponding to each sample image as labels to train the convolutional neural network prediction model, so as to obtain a feature analysis prediction model whose input is a pineapple appearance image and whose output is the surface texture feature parameters and color feature parameters corresponding to the input image; A maturity analysis module is used to collect the appearance image of the pineapple to be detected, pre-process it and input it into the trained feature analysis prediction model, output the surface texture characteristic parameters and color characteristic parameters of the pineapple to be detected, and calculate and generate the pineapple maturity judgment coefficient based on the obtained surface texture characteristic parameters and color characteristic parameters of the pineapple to be detected; The gas characteristic correction module is used to place the pineapple to be tested in a closed environment and detect the The gas characteristics of the pineapple to be detected in the time period are included in the gas characteristics, and the pineapple maturity judgment coefficient is corrected based on the obtained gas characteristics to obtain an accurate pineapple maturity judgment coefficient. Ethylene and carbon dioxide concentrations in the inner closed environment; The maturity judgment module is used to collect the physical parameters of the pineapple to be tested to dynamically correct the initial maturity judgment threshold, obtain the dynamic maturity judgment threshold, compare the precise pineapple maturity judgment coefficient with the dynamic maturity judgment threshold, and issue different maturity judgment results according to different comparison results. The physical parameters include height, average diameter and skin hardness.

Images