Quality detection and sorting system using deep learning technology

Abstract

The invention relates to the technical field of detection, in particular to a quality detection and sorting system utilizing a deep learning technology, which comprises a shooting instrument, a computer, an image processing and analyzing module, a spectrum analyzer, a sorting module, a feedback module and an evaluation module, and the sorting module is used for sorting the detected result data, and feeding back the sorted result data to the computer in real time through the feedback module for a user to check. The image processing and analyzing module is used for processing the shot image to discriminate whether the appearance of the fishing net is broken or not, the material composition of the fishing net is determined through the characteristic wavelength, detection data are sorted, the water area adapted to the current fishing net is obtained through the evaluation module, and feedback is performed through the feedback module. The method can effectively distinguish the flaws of the fishing net, understand the main materials of the fishing net, analyze to obtain an adaptive water area, provide suggestions for fishermen, and bring convenience to the fishermen.

Description

Technical Field

[0001] This invention belongs to the field of fishing net inspection technology, specifically a quality inspection and sorting system utilizing deep learning technology. Background Technology

[0002] Fishing nets are nets used for catching fish. In ancient times, they were made from coarse cloth and hemp through a rolling process. Although these nets were prone to rotting and lacked durability, their fishing efficiency was greatly improved. With the development of the fishing industry, fishing net materials have become increasingly modernized, mainly using polyethylene and nylon, resulting in longer service life and higher catch efficiency. Fishing nets can be classified in various ways. Based on the thickness of the line, the size of the mesh, and the number of layers, they can be divided into different types. For example, gillnets are further divided into floating and sinking nets; floating nets have only floats and no sinkers, while sinking nets have both floats and sinkers. Casting nets are made of nylon thread, rubber thread, or tire cord; the net shape is circular, and skill is required to cast them. Based on the target fish, nets can be divided into crucian carp nets, silverfish nets, catfish nets, etc. Based on the working water depth, they can be divided into floating nets and bottom nets. Based on the opening method, they can be divided into ordinary nets and casting nets, etc.

[0003] Fishermen often purchase brand-new fishing nets, but defects inevitably appear during the manufacturing process. Fishermen rely solely on visual inspection and manual pulling to assess quality, making it difficult to effectively identify these defects or understand the net's main materials. Furthermore, the varying corrosiveness of different water bodies prevents fishermen from assessing the net's corrosion resistance, hindering their uniform use and reducing the net's lifespan – a significant inconvenience. To address this, we propose a quality inspection and ranking system utilizing deep learning technology. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a quality detection and ranking system utilizing deep learning technology to solve the aforementioned technical problems.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a quality detection and ranking system utilizing deep learning technology, comprising: an image capture device, a computer, an image processing and analysis module, a spectral analyzer, a ranking module, a feedback module, and an evaluation module; the ranking module is used to rank the detection results data, and the feedback module provides real-time feedback to the computer for user viewing;

[0006] The detection and sorting steps are as follows:

[0007] S1. Staff place the fishing net to be inspected under the camera and take pictures. The pictures are then uploaded to the computer and analyzed by the image processing and analysis module.

[0008] S2. Place the fishing net under a spectrometer for experimental analysis and determine the material composition of the fishing net by using characteristic wavelengths;

[0009] S3. The image processing and analysis module preprocesses the uploaded images and extracts features from them. It analyzes and processes the collected fishing net images to identify the appearance of the fishing net and whether there is any damage to the net knots.

[0010] S4. Upload the determined fishing net materials to the computer to evaluate the corrosion resistance of the fishing net materials;

[0011] S5. Upload the image recognition results and corrosion resistance assessment results together to the sorting module to sort the currently detected fishing nets.

[0012] S6. Based on the test results, the evaluation module intelligently assesses the applicable areas of the current fishing net and provides relevant suggestions to people;

[0013] S7. Finally, the suggestions and test results are fed back to the computer screen through the feedback module for people to view and analyze.

[0014] Prioritize that, in step S1, when photographing the fishing net, lay the fishing net flat, remove wrinkles, and allow the camera to photograph the entire net; in step S2, the spectrometer uses a small section of excess wire pre-cut from the fishing net to make a sample, places the sample inside the spectrometer, checks whether the software inside the spectrometer maintains a normal connection with the instrument, sets the corresponding wavelength, intensity, and scanning speed parameters, clicks the scan button, scans, and records the obtained spectral data values, processes the data, and uses peak position and peak area calculations to analyze the material of the fishing net;

[0015] The peak position calculation formula includes empirical formulas and absorbance and optical path formulas;

[0016] The empirical formula is:

[0017] P = 5274 / (a-1) + 4731

[0018] Where P is the wave number and a is the gas refractive index;

[0019] The formulas for absorbance and optical path length are:

[0020] K = log(q0 / q) = εcj

[0021] Where K is absorbance, q0 is emitted light intensity, q is transmitted light intensity, c is fishing net mass, and j is optical path length;

[0022] The formula for calculating peak area is:

[0023] K = a * C

[0024] Where K is the peak area, a is the peak height, C is the peak width, a refers to the distance from the peak top to the baseline, C is the width of the peak bottom, and K represents the content of the analyte. The area is proportional to the content.

[0025] Different materials of fishing nets correspond to different characteristic data. The material of the fishing net is determined by calculating the extracted data.

[0026] Prior to this, the preprocessing in step S3 includes image grayscale conversion, image magnification, and image enhancement. Specifically, the image grayscale conversion step first reads and loads the uploaded image, extracts the red, green, and blue component pixels in the image, which range from 0 to 255, and then determines the weights of the three component pixels, with weights of 0.299 for red, 0.587 for green, and 0.114 for blue. Then, a weighted average is performed on each component pixel to obtain the grayscale value of the image. The grayscale values ​​of each pixel are combined to form a new grayscale image, which is then saved.

[0027] The formula for calculating the weighted average of gray values ​​is as follows:

[0028] K = Ra + Gb + Bc

[0029] Where K represents the grayscale value, a, b, and c are the weights of the components, R is red, G is green, and B is blue;

[0030] The image magnification process involves enlarging the grayscale image and then dividing the magnified image into equal parts to reveal the details in the image. The average division size is determined based on the image size, including 3*3, 4*4, and 5*5. The more divisions, the more details are revealed in the image.

[0031] The image is enlarged and segmented, and then enhanced to make it clearer. By increasing the image sharpness, the edges of the image are made clearer, and the details in the image are displayed and exposed.

[0032] Prioritizes the fishing net features extracted in step S3, which include the integrity of the fishing net and the knots in the net. The image feature extraction step uses Canny to measure image edges and obtain edge information. The Canny measurement step uses the Sobel operator to calculate the gradient magnitude and direction of pixels in the image, scanning the gradient image to identify image edges. The Sobel operator includes both horizontal and vertical directions; the horizontal operator matrix is:

[0033]

[0034] Where Gx is the gradient operator in the x-direction;

[0035] The vertical operator matrix is:

[0036]

[0037] Where Gy is the gradient operator in the y direction, and the gradients of x and y are calculated using the above matrix;

[0038] The formulas for calculating the gradient magnitude and gradient direction are as follows:

[0039] G = sqrt(Gx∧2 + Gy∧2)

[0040] θ = atan2(Gy, Gx)

[0041] Where G is the magnitude of the gradient and θ is the direction of the gradient.

[0042] Prioritize the following steps: In step S4, the material identified in step S2 is uploaded. The computer uses big data to obtain the corrosion resistance of the material. In the big data network, web crawling technology is used to obtain the properties of the material from websites, forums, and literature. In step S5, the sorting module organizes and sorts the uploaded data in ascending order, marking the data as 1, 2, 3, 4, 5, and so on. The labeled data is then stored and saved.

[0043] Prior to this, in step S6, the sorted data is uploaded to the evaluation module for evaluation. The steps for establishing the evaluation module are as follows:

[0044] A1. Determine the purpose of the assessment, which is the applicable geographical area for the fishing nets;

[0045] A2. Collect water quality data for different regions through online platforms, field visits, and market research.

[0046] A3. Establish an evaluation model using deep learning technology;

[0047] A4. Use the dataset to input into the evaluation model to check the correctness of the evaluation model, and adjust and optimize the data of the evaluation model to ensure the accuracy of the evaluation model.

[0048] Prioritize the A3 step, which involves collecting and preparing sample data, designing the model, using a convolutional neural network (CNN) as the model framework, defining the network structure, defining the forward computation function, and finally defining the loss function to complete the model construction. The CNN includes convolutional layers, input layers, pooling layers, connection layers, and an output layer. Convolutional layers extract features by capturing different local information in the input image; pooling layers downsample the feature map to reduce the number of parameters; the input layer is the neural network's entry point; the output layer outputs the numerical values ​​input to the model; and the connection layer integrates the extracted features and converts them into the final output.

[0049] Prior to this, the forward computation function is propagated in the forward direction, calculating the values ​​of the variables sequentially from input to output. By recording the forward computation process of the variables, the gradient of the parameters is calculated. The forward computation function is expressed as follows:

[0050] z = A*c + n

[0051] p = f(z)

[0052] Where c is the input vector, A is the weight matrix, n is the bias term, z is the linear transformation result, p is the output vector, and f is the activation function; the linear transformation result z is obtained by multiplying the weight matrix A and the input vector c, and adding the bias term n. The output vector p of this layer is obtained by performing a nonlinear transformation on z through the activation function f.

[0053] The loss function is logarithmic loss.

[0054] Prior to this, in step S7, the feedback module converts the given suggestions into feedback signals, transmits them to the display screen, and generates a data report on the result of the current label for analysis.

[0055] Prioritizes the feedback signal conversion by converting analog signal data into digital signals; the data report generation step involves analyzing the data using regression analysis to extract key information and representing it in the form of charts, images, or text; the regression analysis expression is as follows:

[0056] p = ax + b

[0057] Where p is the dependent variable, x is the independent variable, a is the intercept of the regression line, and b is the slope of the regression line.

[0058] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0059] This application involves pre-photographing fishing nets, processing the images using an image processing and analysis module to identify any damage to the nets' appearance, conducting experimental analysis using a spectral analyzer to determine the net's material composition through characteristic wavelengths, sorting the test data, determining the suitable waters for the nets using an evaluation module, and providing feedback through a feedback module. This approach effectively identifies defects in the fishing nets, provides an understanding of the main materials used in the nets, and analyzes the suitable waters, offering suggestions and convenience to fishermen. Attached Figure Description

[0060] Figure 1 This is a framework diagram of the detection and sorting steps of the present invention. Detailed Implementation

[0061] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0062] This invention provides a technical solution: a quality detection and ranking system utilizing deep learning technology. The quality detection and ranking system includes: an image capture device, a computer, an image processing and analysis module, a spectral analyzer, a ranking module, a feedback module, and an evaluation module. The ranking module is used to rank the detection results data, and the feedback module provides real-time feedback to the computer for user viewing.

[0063] The detection and sorting steps are as follows:

[0064] S1. Staff place the fishing net to be inspected under the camera and take pictures. The pictures are then uploaded to the computer and analyzed by the image processing and analysis module.

[0065] S2. Place the fishing net under a spectrometer for experimental analysis and determine the material composition of the fishing net by using characteristic wavelengths;

[0066] S3. The image processing and analysis module preprocesses the uploaded images and extracts features from them. It analyzes and processes the collected fishing net images to identify the appearance of the fishing net and whether there is any damage to the net knots.

[0067] S4. Upload the determined fishing net materials to the computer to evaluate the corrosion resistance of the fishing net materials;

[0068] S5. Upload the image recognition results and corrosion resistance assessment results together to the sorting module to sort the currently detected fishing nets.

[0069] S6. Based on the test results, the evaluation module intelligently assesses the applicable areas of the current fishing net and provides relevant suggestions to people;

[0070] S7. Finally, the suggestions and test results are fed back to the computer screen through the feedback module for people to view and analyze.

[0071] Furthermore, in step S1, when photographing the fishing net, the net is laid flat and wrinkles are removed so that the camera can capture the net completely. In step S2, a small section of excess wire is pre-cut from the net to create a sample. The sample is placed inside the spectrometer, and the software is checked to ensure a proper connection with the instrument. The corresponding wavelength, intensity, and scanning speed parameters are then set. The scan button is clicked to perform the scan and record the obtained spectral data values. The data is then processed, and peak position and peak area calculations are used to analyze the material of the fishing net.

[0072] The peak position calculation formula includes empirical formulas and absorbance and optical path formulas;

[0073] The empirical formula is:

[0074] P = 5274 / (a-1) + 4731

[0075] Where P is the wave number and a is the gas refractive index;

[0076] The formulas for absorbance and optical path length are:

[0077] K = log(q0 / q) = εcj

[0078] Where K is absorbance, q0 is emitted light intensity, q is transmitted light intensity, c is fishing net mass, and j is optical path length;

[0079] The formula for calculating peak area is:

[0080] K = a * C

[0081] Where K is the peak area, a is the peak height, C is the peak width, a refers to the distance from the peak top to the baseline, C is the width of the peak bottom, and K represents the content of the analyte. The area is proportional to the content.

[0082] Peak area is usually calculated using the integral method, which involves integrating the area under the peak curve. In the specific calculation, a peak area calculation table is first established, and then the parameters such as vertical peak height, half peak width, and peak area are input. Finally, the peak area is calculated.

[0083] Different materials of fishing nets correspond to different characteristic data. The material of the fishing net is determined by calculating the extracted data.

[0084] Furthermore, the preprocessing in step S3 includes image grayscale conversion, image magnification, and image enhancement. Specifically, the image grayscale conversion step first reads and loads the uploaded image, extracts the red, green, and blue component pixels (ranging from 0 to 255), determines the weights of the three components (red 0.299, green 0.587, and blue 0.114), performs a weighted average on each component pixel to obtain the image's grayscale value, and combines the grayscale values ​​of each pixel to form a new grayscale image before saving it.

[0085] The formula for calculating the weighted average of gray values ​​is as follows:

[0086] K = Ra + Gb + Bc

[0087] Where K represents the grayscale value, a, b, and c are the weights of the components, R is red, G is green, and B is blue;

[0088] The formula for the weighted average method is used to calculate the weighted average, which is an average calculation method that takes into account the relative importance of each value. In the context of image grayscale processing, the weighted average method is used to convert the RGB components of a color image into grayscale values, where the importance of each color component is determined according to its contribution to human vision.

[0089] Grayscale images have no color, and all RGB color components are equal, making the image more suitable for subsequent image processing and analysis. Grayscale processing can eliminate the influence of color information on image processing results, thereby more accurately extracting and analyzing features in the image.

[0090] The image magnification process involves enlarging the grayscale image and then dividing the magnified image into equal parts to reveal the details in the image. The average division size is determined based on the image size, including 3*3, 4*4, and 5*5. The more divisions, the more details are revealed in the image.

[0091] The image is enlarged and segmented, and then enhanced to make it clearer. By increasing the image sharpness, the edges of the image are made clearer, and the details in the image are displayed and exposed.

[0092] Furthermore, the fishing net features extracted in step S3 include the integrity of the fishing net and the knots in the net. The image feature extraction step uses Canny to measure image edges and obtain image edge information. The Canny measurement step uses the Sobel operator to calculate the gradient magnitude and direction of pixels in the image, and scans the gradient image to identify the image edges. The Sobel operator includes two directions, horizontal and vertical, and the horizontal operator matrix is:

[0093]

[0094] Where Gx is the gradient operator in the x-direction;

[0095] The vertical operator matrix is:

[0096]

[0097] Where Gy is the gradient operator in the y direction, and the gradients of x and y are calculated using the above matrix;

[0098] The formulas for calculating the gradient magnitude and gradient direction are as follows:

[0099] G = sqrt(Gx∧2 + Gy∧2)

[0100] θ = atan2(Gy, Gx)

[0101] Where G is the magnitude of the gradient, and θ is the direction of the gradient;

[0102] Typically, only the gradient magnitude is of concern, because the gradient magnitude reaches its maximum value at the edges. Therefore, edge detection can be achieved by setting a threshold and marking pixels with gradient magnitudes greater than the threshold as edge pixels.

[0103] The Canny edge detection algorithm can accurately identify edges in an image and has high noise resistance.

[0104] Furthermore, in step S4, the material determined in step S2 is uploaded. The computer uses big data to obtain the corrosion resistance of the current material. In the big data network, web crawling technology is used to obtain the properties of the current material from websites, forums and literature. In step S5, the sorting module organizes and sorts the currently uploaded data in ascending order, and marks the current data as 1, 2, 3, 4, 5... and so on. The labeled data is then stored and saved.

[0105] Furthermore, in step S6, the sorted data is uploaded to the evaluation module for evaluation. The steps for establishing the evaluation module are as follows:

[0106] A1. Determine the purpose of the assessment, which is the applicable geographical area for the fishing nets;

[0107] A2. Collect water quality data for different regions through online platforms, field visits, and market research.

[0108] A3. Establish an evaluation model using deep learning technology;

[0109] A4. Input the dataset into the evaluation model to check the correctness of the evaluation model, and adjust and optimize the data of the evaluation model to ensure the accuracy of the evaluation model.

[0110] In step A4, the prepared data is input into the model, and the model parameters are optimized through an iterative process to minimize the gap between the predicted and actual values. The training process typically uses gradient descent optimization algorithms, including backpropagation, to calculate the gradient of the parameters. During training, the model's performance is evaluated by calculating the model's loss function and accuracy in order to make corresponding adjustments and optimizations. The solution-finding method adopted by the model, i.e., the optimizer, is set, and computing resources are specified.

[0111] Furthermore, in step A3, the model is constructed by collecting and preparing sample data, designing the model, using a convolutional neural network as the model framework, defining the network structure, defining the forward computation function, and finally defining the loss function. The convolutional neural network includes convolutional layers, input layers, pooling layers, connection layers, and output layers. Convolutional layers extract features by capturing different local information in the input image; pooling layers downsample the feature map to reduce the number of parameters; the input layer is the neural network's entry point; the output layer outputs the numerical values ​​input to the model; and the connection layer integrates the extracted features and converts them into the final output.

[0112] Convolutional neural networks first use small filters to slide across different regions of an image to extract features from different regions. This process involves dimensionality reduction and parameter sharing. The output of the convolutional layers is downsampled to reduce parameters and control overfitting, forming pooling layers. The convolutional features are then used for classification through fully connected layers. Finally, the backpropagation algorithm is used to train the parameters, and the Dropout strategy is used to prevent overfitting.

[0113] Furthermore, the forward computation function is propagated in the forward direction, calculating the values ​​of variables sequentially from input to output. By recording the forward computation process of the variables, the gradient of the parameters is calculated. The forward computation function formula is expressed as:

[0114] z = A*c + n

[0115] p = f(z)

[0116] Where c is the input vector, A is the weight matrix, n is the bias term, z is the linear transformation result, p is the output vector, and f is the activation function; the linear transformation result z is obtained by multiplying the weight matrix A and the input vector c, and adding the bias term n. The output vector p of this layer is obtained by performing a nonlinear transformation on z through the activation function f.

[0117] The loss function is logarithmic loss.

[0118] Furthermore, in step S7, the feedback module converts the given suggestions into feedback signals, which are then transmitted to the display screen to generate a data report on the result of the current label for analysis.

[0119] Furthermore, the feedback signal conversion transforms analog signal data into digital signals; the data report generation step involves analyzing the data using regression analysis to extract key information, which is then represented using charts, images, or text; the regression analysis expression is as follows:

[0120] p = ax + b

[0121] Where p is the dependent variable, x is the independent variable, a is the intercept of the regression line, and b is the slope of the regression line.

[0122] Example 1

[0123] A quality detection and ranking system utilizing deep learning technology, with the following detection and ranking steps:

[0124] S1. Staff place the fishing net to be inspected under the camera and take pictures. The pictures are then uploaded to the computer and analyzed by the image processing and analysis module.

[0125] S2. Place the fishing net under a spectrometer for experimental analysis and determine the material composition of the fishing net by using characteristic wavelengths;

[0126] S3. The image processing and analysis module preprocesses the uploaded images and extracts features from them. It analyzes and processes the collected fishing net images to identify the appearance of the fishing net and whether there is any damage to the net knots.

[0127] S4. Upload the determined fishing net materials to the computer to evaluate the corrosion resistance of the fishing net materials;

[0128] S5. Upload the image recognition results and corrosion resistance assessment results together to the sorting module to sort the currently detected fishing nets.

[0129] S6. Based on the test results, the evaluation module intelligently assesses the applicable areas of the current fishing net and provides relevant suggestions to people;

[0130] S7. Finally, the suggestions and test results are fed back to the computer screen through the feedback module for people to view and analyze.

[0131] Comparative Example 1

[0132] The steps for inspecting fishing nets are as follows:

[0133] A1. Fishermen check the sturdiness of the fishing net by pulling it with their hands;

[0134] A2. Fishermen visually inspect the fishing nets to check for any damage.

[0135] Comparative Example 1 struggles to effectively identify flaws in fishing nets and fails to provide an understanding of their main materials. Furthermore, the corrosiveness of different water bodies varies significantly, preventing fishermen from assessing the corrosion resistance of nets and leading to their indiscriminate use and reduced lifespan. Example 1, however, pre-photographs the fishing nets and uses an image processing and analysis module to identify any breaks in the net's appearance. A spectral analyzer is used to analyze the nets, determining their material composition through characteristic wavelengths. The test data is then sorted, and an evaluation module determines the suitable waters for the current net. A feedback module provides further feedback, effectively identifying net flaws, understanding the main materials, and determining the suitable waters, offering valuable advice to fishermen – a highly convenient approach.

[0136] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0137] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A quality detection and ranking system utilizing deep learning technology, characterized in that, The quality inspection and sorting system includes: an image camera, a computer, an image processing and analysis module, a spectrum analyzer, a sorting module, a feedback module, and an evaluation module. The sorting module sorts the inspection results data and feeds them back to the computer in real time through the feedback module for users to view. The detection and sorting steps are as follows: S1. Staff place the fishing net to be inspected under the camera and take pictures. The pictures are then uploaded to the computer and analyzed by the image processing and analysis module. S2. Place the fishing net under a spectrometer for experimental analysis and determine the material composition of the fishing net by using characteristic wavelengths; S3. The image processing and analysis module preprocesses the uploaded images and extracts features from them. It analyzes and processes the collected fishing net images to identify the appearance of the fishing net and whether there is any damage to the net knots. S4. Upload the determined fishing net materials to the computer to evaluate the corrosion resistance of the fishing net materials; S5. Upload the image recognition results and corrosion resistance assessment results together to the sorting module to sort the currently detected fishing nets. S6. Based on the test results, the evaluation module intelligently assesses the applicable areas of the current fishing net and provides relevant suggestions to people; S7. Finally, the suggestions and test results are fed back to the computer screen through the feedback module for people to view and analyze.

2. The quality detection and ranking system utilizing deep learning technology according to claim 1, characterized in that: In step S1, when photographing the fishing net, lay the net flat, remove any wrinkles, and allow the camera to capture the net completely. In step S2, a small section of excess material is pre-cut from the fishing net to create a sample. The sample is placed inside the spectrometer, and the software is checked to ensure a proper connection with the instrument. The corresponding wavelength, intensity, and scanning speed parameters are then set. The scan button is clicked to perform the scan and record the obtained spectral data values. The data is then processed, and peak position and peak area calculations are used to analyze the material of the fishing net. The peak position calculation formula includes empirical formulas and absorbance and optical path formulas; The empirical formula is: P = 5274 / (a-1) + 4731 Where P is the wave number and a is the gas refractive index; The formulas for absorbance and optical path length are: K = log(q0 / q) = εcj Where K is absorbance, q0 is emitted light intensity, q is transmitted light intensity, c is fishing net mass, and j is optical path length; The formula for calculating peak area is: K = a * C Where K is the peak area, a is the peak height, C is the peak width, a refers to the distance from the peak top to the baseline, C is the width of the peak bottom, and K represents the content of the analyte. The area is proportional to the content. Different materials of fishing nets correspond to different characteristic data. The material of the fishing net is determined by calculating the extracted data.

3. The quality detection and ranking system utilizing deep learning technology according to claim 1, characterized in that: The preprocessing in step S3 includes image grayscale conversion, image magnification, and image enhancement. Specifically, the image grayscale conversion step first reads and loads the uploaded image, extracts the red, green, and blue component pixels (ranging from 0 to 255), determines the weights for each component (red 0.299, green 0.587, and blue 0.114), performs a weighted average on each component pixel to obtain the image's grayscale value, and combines these grayscale values ​​to form a new grayscale image before saving it. The formula for calculating the weighted average of gray values ​​is as follows: K = Ra + Gb + Bc Where K represents the grayscale value, a, b, and c are the weights of the components, R is red, G is green, and B is blue; The image magnification process involves enlarging the grayscale image and then dividing the magnified image into equal parts to reveal the details in the image. The average division size is determined based on the image size, including 3*3, 4*4, and 5*5. The more divisions, the more details are revealed in the image. The image is enlarged and segmented, and then enhanced to make it clearer. By increasing the image sharpness, the edges of the image are made clearer, and the details in the image are displayed and exposed.

4. The quality detection and ranking system utilizing deep learning technology according to claim 1, characterized in that: The fishing net features extracted in step S3 include the integrity of the fishing net and the knots in the net. The image feature extraction step uses Canny to measure image edges and obtain image edge information. The Canny measurement step uses the Sobel operator to calculate the gradient magnitude and direction of pixels in the image, scans the gradient image, and identifies the image edges. The Sobel operator includes two directions, horizontal and vertical. The horizontal operator matrix is: Where Gx is the gradient operator in the x-direction; The vertical operator matrix is: Where Gy is the gradient operator in the y direction, and the gradients of x and y are calculated using the above matrix; The formulas for calculating the gradient magnitude and gradient direction are as follows: G = sqrt(Gx∧2 + Gy∧2) θ = atan2(Gy, Gx) Where G is the magnitude of the gradient and θ is the direction of the gradient.

5. The quality detection and ranking system utilizing deep learning technology according to claim 1, characterized in that: In step S4, the material determined in step S2 is uploaded. The computer uses big data to obtain the corrosion resistance of the material. In the big data network, web crawling technology is used to obtain the properties of the material from websites, forums and literature. In step S5, the sorting module organizes and sorts the uploaded data in ascending order, and marks the data as 1, 2, 3, 4, 5 and so on. The labeled data is then stored and saved.

6. The quality detection and ranking system utilizing deep learning technology according to claim 1, characterized in that: In step S6, the sorted data is uploaded to the evaluation module for evaluation. The steps to establish the evaluation module are as follows: A1. Determine the purpose of the assessment, which is the applicable geographical area for the fishing nets; A2. Collect water quality data for different regions through online platforms, field visits, and market research. A3. Establish an evaluation model using deep learning technology; A4. Use the dataset to input into the evaluation model to check the correctness of the evaluation model, and adjust and optimize the data of the evaluation model to ensure the accuracy of the evaluation model.

7. The quality detection and ranking system utilizing deep learning technology according to claim 6, characterized in that: The A3 step involves collecting and preparing sample data, designing the model, using a convolutional neural network as the model framework, defining the network structure, defining the forward computation function, and finally defining the loss function to complete the model construction. A convolutional neural network includes convolutional layers, input layers, pooling layers, connection layers, and output layers. Convolutional layers extract features by capturing different local information in the input image, pooling layers are used to downsample the feature map to reduce the number of parameters, the input layer is the entry point of the neural network, and the output layer is used to output the numerical values ​​input to the model. The connection layer is used to integrate the extracted features and transform them into the final output.

8. The quality detection and ranking system utilizing deep learning technology according to claim 7, characterized in that: The forward computation function propagates forward, calculating the values ​​of variables sequentially from input to output. By recording the forward computation process of the variables, the gradient of the parameters is calculated. The forward computation function is expressed as: z = A*c + n p = f(z) Where c is the input vector, A is the weight matrix, n is the bias term, z is the linear transformation result, p is the output vector, and f is the activation function; the linear transformation result z is obtained by multiplying the weight matrix A and the input vector c, and adding the bias term n. The output vector p of this layer is obtained by performing a nonlinear transformation on z through the activation function f. The loss function is logarithmic loss.

9. The quality detection and ranking system utilizing deep learning technology according to claim 1, characterized in that: In step S7, the feedback module converts the given suggestions into feedback signals, transmits them to the display screen, and generates a data report for the current label's result, which is then analyzed.

10. The quality detection and ranking system utilizing deep learning technology according to claim 9, characterized in that: Feedback signal conversion involves converting analog signal data into digital signals; the data report generation step involves analyzing the data using regression analysis to extract key information and representing it using charts, images, or text; the regression analysis expression is as follows: p = ax + b Where p is the dependent variable, x is the independent variable, a is the intercept of the regression line, and b is the slope of the regression line.

Images