Unmanned laboratory sample detection system and method based on image recognition

Abstract

The invention discloses an unmanned laboratory sample detection system and method based on image recognition. According to the system, sample identification information is quickly and accurately recognized through the image recognition module, the pressure testing machine module applies vertical loading force and measures stress and deformation of a sample in real time, and the force feedback control module dynamically adjusts the loading rate and strength according to real-time data to ensure that the loading process meets preset standards. The data acquisition and processing module is used for cleaning and preprocessing acquired data, the mechanical property analysis module is used for calculating mechanical property indexes such as compressive strength and tensile strength of a sample, and the report generation and optimization module is used for automatically generating a detection report meeting industrial standards and marking abnormal data. The detection scheme optimization module generates an optimization scheme based on historical data and real-time data, adjusts detection parameters, and improves detection efficiency and accuracy. According to the invention, automation, intelligentization and high efficiency of unmanned laboratory sample detection are realized, and the method is suitable for mechanical property detection of various materials.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned laboratory detection, in particular to an unmanned laboratory sample detection system and method based on image recognition. BACKGROUND

[0002] In the traditional laboratory sample detection process, sample identification, loading and data collection usually rely on manual operation. This method not only has low efficiency, but also is prone to errors due to human factors. For example, manual identification of sample identification information (such as two-dimensional code, bar code) is prone to misreading or missing reading, which affects the accuracy of subsequent detection. In addition, manual operation is difficult to ensure the stability and accuracy of the loading process when loading, especially in the case of long-term continuous loading, which is prone to problems such as inconsistent loading rate or loading strength deviation.

[0003] Although the existing automated detection system improves the detection efficiency to some extent, it still has some shortcomings. First, most systems lack the ability to quickly and accurately identify sample identification information, resulting in low efficiency in sample information entry. Second, the existing automated system usually uses a fixed loading scheme when loading, which cannot dynamically adjust the loading rate and strength according to real-time feedback data. This is prone to cause the loading process to not meet the standard when detecting samples with different mechanical properties, affecting the accuracy and reliability of the detection results. In addition, the existing system also has shortcomings in data processing and analysis, which cannot effectively remove noise and abnormal data, limiting the accuracy of mechanical property analysis results.

[0004] In summary, the main problems existing in the prior art include:

[0005] The identification efficiency of sample identification information is low, and misreading or missing reading is prone to occur.

[0006] The loading process lacks dynamic adjustment ability, and it is difficult to ensure that the loading force meets the predetermined standard.

[0007] The data processing and analysis capability is insufficient, which cannot effectively remove noise and abnormal data, affecting the accuracy of the mechanical property analysis results.

[0008] There is a lack of optimization mechanism based on historical data and real-time data, which cannot realize the continuous optimization of the detection scheme.

[0009] In view of the above problems, the present application proposes an unmanned laboratory sample detection system and method based on image recognition. SUMMARY

[0010] The unmanned laboratory sample detection system and method based on image recognition have the advantages of high automation degree, high detection precision, high efficiency and optimization, and solve the problems of low efficiency, large error of manual operation in traditional laboratory detection and lack of dynamic adjustment capability of existing automatic systems.

[0011] To achieve the above object, the present application provides the following technical scheme: an unmanned laboratory sample detection system based on image recognition, comprising the following modules: an image recognition module; a pressure testing machine module; a force feedback control module; a data acquisition and processing module; a mechanical property analysis module; a report generation and optimization module;

[0012] An optimization detection scheme module.

[0013] Preferably, the image recognition module is configured to identify and locate sample identification information through image recognition technology, and the identification information includes a two-dimensional code and a bar code on the sample.

[0014] The pressure testing machine module is configured to apply a vertical load and measure the stress and deformation of the sample in real time, and the pressure testing machine module comprises a loading unit and a sensor system, wherein the sensor system is configured to measure the deformation and strain data of the sample in real time and transmit the feedback data to the force feedback control module in real time.

[0015] The force feedback control module is configured to receive real-time data from the pressure testing machine module, adjust the loading rate and loading strength according to the stress and strain data provided by the sensor system through multiple control algorithms, ensure that the force in the loading process meets the predetermined loading scheme, and realize dynamic adjustment in the loading process.

[0016] The multiple control algorithms comprise:

[0017] A PID control algorithm for adjusting the loading rate: v(t) = K p e(t) + K i ∫0 t e(τ)dτ + K d d / tde(t), wherein v(t) is the loading rate, e(t) is the difference between the current loading force and the set value, K p ,K i ,K d are proportional, integral and differential gains.

[0018] An incremental control algorithm for adjusting the loading strength: F load (t) = F initial + ΔF(t), wherein F load (t) is the loading strength at time t, and F initialis the initial loading intensity, and ΔF(t) is the increment calculated according to the current stress and deformation data;

[0019] The proportional feedback control algorithm for dynamically adjusting the loading process in real time based on feedback data is F adjusted =F current +K f ·(F desired -F current ), wherein F adjusted is the adjusted loading intensity, F current is the current loading intensity, F desired is the predetermined target loading intensity, and K f is the feedback control gain;

[0020] The data acquisition and processing module is configured to acquire data from the pressure testing machine module and the sensor system, remove noise and abnormal data after preprocessing, generate a data set suitable for mechanical property analysis, and store and transmit the data in real time;

[0021] The mechanical property analysis module is configured to calculate the mechanical property indicators of the sample, including compressive strength and tensile strength, based on the preprocessed data using a stress-strain analysis function, and the calculation result is obtained by the stress-strain analysis function;

[0022] The report generation and optimization module is configured to automatically generate a detection report meeting the requirements of industry standards based on the output result of the mechanical property analysis module, and the report includes standard value comparison, compliance analysis and abnormal data labeling of the detection result;

[0023] The optimized detection scheme module is configured to predict the future mechanical properties of the sample and generate an optimized detection scheme based on historical detection data and real-time sample data through machine learning algorithms and data analysis techniques, and the optimized scheme adjusts the detection parameters according to the specific performance requirements of the sample, including loading time, loading rate and measurement accuracy.

[0024] The unmanned laboratory sample detection method based on image recognition includes the following steps:

[0025] S1, sample preparation and positioning: scanning the sample through image recognition technology, extracting the identification information of the sample, and determining the subsequent detection parameters according to the information;

[0026] S2, loading control and data acquisition: starting the pressure testing machine and collecting stress and deformation data of the sample during the loading process;

[0027] S3, force feedback control and adjustment: dynamically adjusting the loading rate and loading intensity through the feedback loop according to the real-time collected data to ensure that the loading process meets the set standards;

[0028] S4, data processing and analysis: clean and denoise the collected raw data, and apply the processed data to stress-strain analysis function to calculate the compressive strength, tensile strength and other mechanical performance indicators of the sample;

[0029] S5, report generation and optimization: automatically generate a standard detection report according to the analysis results, and mark the abnormal data;

[0030] S6, optimization of detection scheme: based on historical data and real-time data analysis, generate an optimized detection scheme, and adjust the loading time and loading rate parameters.

[0031] Preferably, the sample preparation and positioning comprises:

[0032] S11: Place the sample on the detection platform, start the image recognition module, collect the sample image and process it;

[0033] S12: The image recognition module extracts the identification information from the collected image and performs data analysis; by calling the image recognition function, the identification information in the image will be extracted and analyzed as ID sample , which provides basic data for subsequent detection operations; the specific operation form of the image recognition function is as follows:

[0034] I gray = Grayscale(I input );

[0035] I binary = Threshold(I gray , T);

[0036] P identifier = EdgeDetection(I binary );

[0037] ID sample = Decode(P identifier );

[0038] Wherein, I gray : grayscale image, I binary : binary image, P identifier : edge detection result of identifier, ID sample : sample identification information, including two-dimensional code, barcode, I input : input image collected by camera, T: binary threshold value;

[0039] S13: The identified identification information is transmitted to the pressure testing machine module to determine the subsequent loading conditions.

[0040] Preferably, the loading control and data acquisition include:

[0041] S21: Based on the identified sample information, set the initial loading conditions, including loading force and loading rate; this process involves calling the loading control and data acquisition functions, based on the sample's identification information (ID). sample And the loading time T in the preset loading conditions load Calculate the initial loading force F load And set the loading conditions; the specific operation form of the loading control and data acquisition functions is as follows: F load =LoadCalculation(I sample ,T load ,Δx);

[0042] Among them, F load Loading force, I sample Sample identification information is obtained through an image recognition module. load Loading time, set according to sample characteristics; Δx: Deformation amount, measured by sensor.

[0043] S22: Start the pressure testing machine and begin the loading process. At the same time, the force and deformation data of the sample are collected in real time through sensors.

[0044] S23: Sensor data is transmitted to the data acquisition and processing module in real time for preliminary data processing and cleaning.

[0045] Preferably, the force feedback control and adjustment includes:

[0046] S31: Real-time monitoring of sensor data, dynamic adjustment of loading rate and loading intensity to ensure that the loading process meets predetermined standards. By calling the force feedback control function, the loading force is adjusted according to real-time data feedback to ensure that the force during the loading process always meets the standard. The specific calculation form of the force feedback control function is as follows:

[0047] F adjusted =FeedbackControl(F load ,Δerror) where, F adjusted Adjusted loading force, F load Current loading force, real-time data; Δerror: loading error, calculated based on real-time feedback data.

[0048] S32: The force feedback control module confirms the precise application of stress during the loading process through real-time adjustments.

[0049] Preferably, the data processing and analysis includes:

[0050] S41: The collected raw data is transmitted to the data processing module for denoising and outlier detection; the raw data is cleaned by a data cleaning function to remove noise and outliers, ensuring the accuracy of the data;

[0051] S42: The cleaned data enters the mechanical property analysis module, and the stress-strain analysis function is used to calculate the compressive strength, tensile strength and other mechanical property indexes of the sample; the stress-strain analysis function is called to calculate the stress σ and strain of the sample based on the cleaned data, and the specific operation form of the stress-strain analysis function is as follows: σ = F load / A, = Δx / L, where σ: stress, mechanical property analysis result, : strain, calculated by deformation and original length, F load : load, A: cross-sectional area of the sample, Δx: deformation, L: original length of the sample;

[0052] S43: Generate a mechanical property analysis report of the sample based on the calculation results to provide a basis for subsequent report generation.

[0053] Preferably, the report generation and optimization sub-steps are:

[0054] S51: The report generation module automatically generates a detection report based on the mechanical property analysis results, including compliance analysis of the detection results; this process is implemented by a report generation function, and the specific operation form is as follows:

[0055] R final = GenerateReport(D analysis ,A threshold ), where,

[0056] R final : final detection report, D analysis : mechanical property analysis data, A threshold : abnormal data threshold, used to mark outliers;

[0057] S52: Mark the abnormal data in the report;

[0058] Preferably, the optimization detection scheme includes:

[0059] S61: Optimize the detection scheme by analyzing historical data and real-time data; generate optimized experimental conditions, including adjusting the loading time and loading rate; this process is implemented by a detection scheme optimization function, and the specific operation form is as follows:

[0060] P opt = OptimizeDetection(P prev ,D analysis ) where:

[0061] Popt : optimized detection scheme

[0062] P prev : previous experiment scheme

[0063] D analysis : mechanical property analysis data

[0064] S62: automatically generating an optimized detection process and detection parameters according to the optimization results.

[0065] Compared with the prior art, the present application has the following advantages:

[0066] 1. The image recognition module is set to quickly and accurately identify sample identification information, improving sample identification efficiency and accuracy.

[0067] 2. The force feedback control module is set to dynamically adjust the loading process, ensuring that the loading force meets the predetermined standard and improving the accuracy and stability of the detection.

[0068] 3. The data acquisition and processing module cleans and preprocesses the collected data, removing noise and abnormal data to ensure the accuracy and reliability of the data.

[0069] 4. The optimized detection scheme module generates an optimized scheme based on historical data and real-time data, adjusts detection parameters, improves detection efficiency and accuracy, and realizes continuous optimization of the detection process. BRIEF DESCRIPTION OF DRAWINGS

[0070] Fig. 1 is the system module architecture diagram of the present application;

[0071] Fig. 2 is the sample detection flowchart of the present application;

[0072] Fig. 3 is the force feedback control flowchart of the present application;

[0073] Fig. 4 is the data processing and analysis flowchart of the present application. DETAILED DESCRIPTION

[0074] The technical solutions in the embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0075] Please refer to Figs. 1 to 4The present application relates to an unmanned laboratory sample detection system based on image recognition, which realizes efficient, accurate and reliable sample detection through automated and intelligent technical means.

[0076] The system includes the following modules: image recognition module; pressure testing machine module; force feedback control module; data acquisition and processing module; mechanical property analysis module; report generation and optimization module; optimized detection scheme module;

[0077] These modules work together to form a complete detection process, from sample identification and positioning, to loading control, data acquisition and processing, to mechanical property analysis and report generation, and finally through the optimized detection scheme module to continuously optimize the detection process.

[0078] 1. Image recognition module

[0079] The image recognition module is one of the core components of the system, and its main function is to identify and locate sample identification information through image recognition technology. These identification information usually includes two-dimensional codes, barcodes, etc. on the sample. The working principle and execution process of this module are as follows:

[0080] Sample placement: Place the sample on the detection platform, ensuring that the sample's identification information (such as two-dimensional code, barcode) faces the camera.

[0081] Image acquisition: Start the image recognition module, and the camera collects the image of the sample.

[0082] Image processing: The collected image is first subjected to grayscale processing, converting the color image into a grayscale image to reduce data volume and simplify subsequent processing steps.

[0083] Binary processing: The grayscale image is further subjected to binary processing, dividing the pixel values in the image into two categories to facilitate subsequent edge detection and decoding operations.

[0084] Edge detection: Perform edge detection on the binary image to locate the position of the identifier (such as two-dimensional code, barcode).

[0085] Decoding: Decode the detected identifier to extract the sample's identification information (including two-dimensional code ID, barcode information).

[0086] Through the above steps, the image recognition module can quickly and accurately identify the sample's identification information, providing basic data for the subsequent detection process.

[0087] 2. Pressure testing machine module

[0088] The pressure testing machine module is the key module for realizing the mechanical property detection of samples. Its main function is to apply vertical loading force and measure the stress and deformation of samples in real time. The module includes a loading unit and a sensor system. The sensor system is responsible for measuring the deformation and strain data of the sample in real time and transmitting these feedback data to the force feedback control module in real time. The specific working principle and execution process are as follows:

[0089] Loading unit: According to the identification information of the sample

[0090] and the preset loading conditions, the loading unit applies vertical loading force.

[0091] Sensor system: The sensor system measures the deformation and strain data of the sample in real time and transmits these data to the force feedback control module.

[0092] Through the pressure testing machine module, the system can accurately apply load and monitor the stress and deformation of the sample in real time, providing data support for subsequent mechanical property analysis.

[0093] 3. Force feedback control module

[0094] The force feedback control module is the key module to ensure that the loading process meets the predetermined standards. This module receives real-time data from the pressure testing machine module and adjusts the loading rate and loading strength based on the stress and strain data provided by the sensor system through multiple control algorithms. The specific control algorithms include: p PID control algorithm for adjusting loading rate: v(t) = K i e(t) + K t ∫0 d e(τ)dτ + K p d / tde(t), where v(t) is the loading rate, e(t) is the difference between the current loading force and the set value, K i ,K d ,K load are the proportional, integral and differential gains; these parameters are determined through experiments and debugging to ensure the stability and accuracy of the loading rate.

[0095] Incremental control algorithm for adjusting loading strength: F initial (t) = F load + ΔF(t), where F initial (t) is the loading strength at time t, F adjusted is the initial loading strength, and ΔF(t) is the increment calculated based on the current stress and deformation data; this algorithm adjusts the loading strength in real time to ensure that the loading process meets the predetermined standards.

[0096] Proportional feedback control algorithm for real-time dynamic adjustment of the loading process: F current+K f ·(F desired -F current ), where F adjusted is the adjusted loading intensity, F current is the current loading intensity, F desired is the predetermined target loading intensity, K f is the feedback control gain. This algorithm dynamically adjusts the loading intensity in real-time based on feedback data, ensuring the accuracy and stability of the loading process.

[0097] Through the force feedback control module, the system can monitor and adjust the loading process in real-time, ensuring that the loading force always meets the predetermined standards, thereby improving the accuracy and reliability of the detection.

[0098] 4. Data Acquisition and Processing Module: The data acquisition and processing module is responsible for collecting data from the pressure testing machine module and the sensor system, and preprocessing these data to generate a dataset suitable for mechanical property analysis. The specific working principle and execution process are as follows:

[0099] Data acquisition: The sensor system collects real-time stress and deformation data of the sample and transmits these data to the data acquisition and processing module.

[0100] Data preprocessing: The original data collected is cleaned and denoised to remove noise and abnormal data, ensuring the accuracy and reliability of the data. Through the data acquisition and processing module, the system can generate high-quality datasets to provide reliable data support for subsequent mechanical property analysis.

[0101] 5. Mechanical Property Analysis Module

[0102] The mechanical property analysis module calculates the compressive strength, tensile strength and other mechanical performance indicators of the sample based on the preprocessed data using stress-strain analysis functions. The specific working principle and execution process are as follows:

[0103] Stress-strain analysis: According to the cleaned data, the stress σ and strain of the sample are calculated, and through the mechanical property analysis module, the system can accurately calculate the mechanical performance indicators of the sample, providing scientific basis for subsequent report generation.

[0104] 6. Report Generation and Optimization Module

[0105] The report generation and optimization module automatically generates a detection report that meets the requirements of industry standards based on the output results of the mechanical property analysis module. The report includes a comparison of standard values of detection results, compliance analysis and labeling of abnormal data. The specific working principle and execution process are as follows:

[0106] Report generation: Based on the mechanical property analysis results, a detection report is generated.

[0107] Anomaly data labeling: Labeling of abnormal data in the detection results to ensure transparency and accuracy of the report.

[0108] Through the report generation and optimization module, the system can generate high-quality detection reports, providing users with detailed detection results and analysis.

[0109] 7. Optimization of detection scheme module

[0110] The optimization of detection scheme module is based on historical detection data and real-time collected sample data, through machine learning algorithms and data analysis techniques, to predict the future mechanical properties of the sample and generate an optimized detection scheme. The specific working principle and execution process are as follows:

[0111] Data analysis: Analyze historical detection data and real-time collected sample data to predict the future mechanical properties of the sample.

[0112] Optimized scheme generation: According to the analysis results, generate an optimized detection scheme, adjust the detection parameters (including loading time, loading rate and measurement accuracy).

[0113] Unmanned laboratory sample detection method based on image recognition: Sample preparation and positioning

[0114] S1, sample preparation and positioning

[0115] Sample preparation and positioning is the starting step of the entire unmanned laboratory sample detection process, and its core goal is to quickly and accurately identify the identification information (such as two-dimensional code, bar code) of the sample through image recognition technology, and determine the specific parameters for subsequent detection according to these information. This process not only provides basic data for subsequent detection operations, but also ensures the automation and intelligence of the entire detection process.

[0116] S11: Sample placement and image acquisition

[0117] Sample placement: Place the sample to be detected in the designated position of the detection platform. The detection platform is usually equipped with positioning devices to ensure the stability and consistency of the sample. The identification information (such as two-dimensional code, bar code) of the sample should be directed towards the camera to facilitate the scanning of the image recognition module.

[0118] Start the image recognition module: Start the image recognition module through the operation interface or automatic control system. The core component of the image recognition module is the camera, which is responsible for collecting image information of the sample.

[0119] Image acquisition: The camera scans the sample and collects images containing identification information. The collected images will be transmitted to the image processing unit to provide raw data for subsequent image processing and recognition operations.

[0120] S12: Image processing and identification information extraction

[0121] Image preprocessing: The collected image is first subjected to grayscale processing, converting the color image into a grayscale image. This process reduces data volume and simplifies subsequent processing steps.

[0122] I gray = Grayscale(I input ); where I gray is the grayscale image and I input is the input image.

[0123] Binaryzation: The grayscale image is further subjected to binaryzation, dividing the pixel values in the image into two categories (usually black and white). This process distinguishes between the foreground and background in the image by setting a threshold value T, thereby highlighting the identification information. binary = Threshold(I gray , T); where I binary is the binaryzation image and T is the binaryzation threshold.

[0124] Edge detection: The binaryzation image is subjected to edge detection, locating the position of the identifier (such as a two-dimensional code, barcode). The edge detection algorithm is usually based on the gradient information of the image, which can effectively identify the outline of the identifier. identifier = EdgeDetection(I binary ); P identifier is the edge detection result of the identifier.

[0125] Identification information extraction and decoding: The detected identifier is decoded to extract the identification information of the sample (such as two-dimensional code ID, barcode information). The decoding process usually involves analyzing the geometric shape and encoding rules of the identifier to recover the original identification information. sample = Decode(P identifier ); ID sample is the sample identification information, including two-dimensional codes, barcodes, etc. These identification information will serve as the basis data for subsequent detection operations.

[0126] S13: Identification information transmission and loading condition determination

[0127] Identification information transmission: The extracted identification information ID sample will be transmitted to the pressure testing machine module. This process is usually implemented through an internal communication network to ensure fast and accurate data transmission.

[0128] Loading condition determination: The pressure testing machine module determines the loading conditions according to the received identification information ID sample, combined with the preset detection scheme, to determine the subsequent loading conditions. These loading conditions include loading force, loading rate, loading time and other parameters, which will directly affect the mechanical property detection results of the sample.

[0129] Through the above steps, the sample preparation and positioning process realizes the complete process from sample placement to identification information extraction, and then to loading condition determination. This process not only ensures the rapid and accurate identification of sample information, but also provides accurate parameter settings for subsequent mechanical property detection. The efficiency of the image recognition module, the reliability of data transmission, and the accuracy of the loading conditions together form a complete technical closed loop, ensuring the efficient operation of the entire unmanned laboratory sample detection system;

[0130] S2, loading control and data acquisition

[0131] Loading control and data acquisition is a key step in the sample detection process of the unmanned laboratory, its main goal is to start the pressure testing machine according to the sample identification information and the preset loading conditions, and to collect the force and deformation data of the sample in the loading process in real time. This process not only ensures the accuracy and stability of the loading process, but also provides high-quality raw data for subsequent data processing and mechanical property analysis.

[0132] S21: Initial loading condition setting

[0133] Sample information analysis: According to the sample identification ID extracted by the image recognition module sample , the system calls the preset detection scheme to obtain the loading conditions related to the sample. These loading conditions usually include loading force, loading rate, loading time and other parameters.

[0134] Loading condition calculation: Through the call of the loading control and data acquisition function, according to the sample identification information ID sample and the preset loading condition (loading time T load ), the initial loading force F load is calculated. This process ensures the parameterization and automation of the loading process.

[0135] F load = LoadCalculation(I sample , T load , Δx), where F load : initial loading force, I sample : sample identification information obtained through the image recognition module, T load : is the preset loading time, set according to the sample characteristics, Δx: deformation measured by the sensor, used for real-time adjustment of the loading force, sensor measurement;

[0136] S22: Loading process start and data acquisition

[0137] Pressure testing machine start: initial loading force F calculated load The pressure testing machine module starts the loading process. The loading unit gradually applies vertical loading force according to the preset loading rate and loading time.

[0138] Real-time data acquisition: During the loading process, the sensor system measures the force and deformation data of the sample in real time. These data include:

[0139] Force data: Real-time measurement of the force F on the sample current .

[0140] Deformation data: Real-time measurement of the deformation Δx of the sample.

[0141] Strain data: Calculate the strain by the deformation and the original size of the sample.

[0142] Data transmission: The collected real-time data is transmitted to the data acquisition and processing module through the internal communication network, providing a basis for subsequent data processing and analysis.

[0143] S23: Preliminary data processing and cleaning

[0144] Data reception: The data acquisition and processing module receives real-time data from the sensor system, including force data, deformation data and strain data.

[0145] Preliminary data processing: Preliminary processing of the collected raw data, including data format conversion, timestamp alignment and other operations to ensure data consistency and readability.

[0146] Data cleaning: Remove noise and abnormal data through data cleaning function. This process can effectively improve the quality of data and reduce errors in subsequent analysis.

[0147] Cleaned data = DataCleaning(original data), where DataCleaning is the data cleaning function used to remove noise and outliers;

[0148] The goal of the data cleaning function DataCleaning is to remove noise and outliers to ensure the accuracy and reliability of the data. The specific steps are as follows:

[0149] Remove noise: Smooth the data through a low-pass filter (such as a moving average filter) to remove high-frequency noise.

[0150] Detect and remove outliers: Detect and remove outliers through statistical methods (such as Z-score or IQR method).

[0151] 1. Remove noise

[0152] Moving average filter: Moving average filter is a common low-pass filter that smooths data by calculating the average value of data within a window. Assuming x(t) is the original data, and the window size is N, the output y(t) of moving average filter is:

[0153] y(t) = (1 / N) * sum(x(i) for i = t-N+1 to t), where y(t) is the smoothed data, and N is the window size.

[0154] 2. Detect and remove outliers

[0155] Z-score method: Z-score method is a common outlier detection method. Assuming x(t) is the smoothed data, its mean is μ, and its standard deviation is σ, the Z-score of each data point is:

[0156] z(t) = (x(t) - μ) / σ, usually, data points with absolute value of Z-score greater than a certain threshold (such as 3) are considered outliers. A threshold θ can be set to mark and remove data points with absolute value of Z-score greater than θ as outliers.

[0157] IQR method: IQR (Interquartile Range) method is a quantile-based outlier detection method. Assuming Q1 is the 25th percentile of data, and Q3 is the 75th percentile, then IQR is: IQR = Q3 - Q1, usually, data points less than Q1 - 1.5 × IQR or greater than Q3 + 1.5 × IQR are considered outliers. A multiplier k (such as 1.5 or 3) can be set to mark and remove data points outside Q1 - k × IQR and Q3 + k × IQR as outliers.

[0158] Specific operation form of data cleaning function

[0159] Combining the above methods, the specific operation form of the data cleaning function DataCleaning can be represented as:

[0160] Moving average filter: y(t) = (1 / N) * sum(x(i) for i = t-N+1 to t);

[0161] Z-score detection and removal of outliers: z(t) = (y(t) - μ) / σ, outliers = {y(t) | |z(t)| > θ}

[0162] IQR detection and removal of outliers:

[0163] Q1 = 25th percentile (y(t))

[0164] Q3 = 75th percentile (y(t))

[0165] IQR = Q3 - Q1

[0166] Outliers = {y(t) | y(t) < Q1 - k x IQR or y(t) > Q3 + k x IQR}

[0167] Remove outliers:

[0168] Cleaned data = {y(t) | y(t) is not an outlier};

[0169] Through the above steps, the loading control and data acquisition process realizes the complete process from the setting of loading conditions to the starting of the loading process, and then to the data acquisition and preliminary processing. This process not only ensures the accuracy and stability of the loading process, but also provides high-quality raw data for subsequent data processing and mechanical property analysis. The accuracy of the loading control module, the real-time nature of data acquisition, and the reliability of data processing together form a complete technical closed loop, ensuring the efficient operation of the entire unmanned laboratory sample detection system.

[0170] S3, Force Feedback Control and Adjustment

[0171] Force feedback control and adjustment is a key step in the sample detection process of the unmanned laboratory. Its main goal is to dynamically adjust the loading rate and loading strength according to the real-time data collected, to ensure that the loading process meets the predetermined standards. This process not only ensures the accuracy and stability of the loading process, but also provides high-quality raw data for subsequent data processing and mechanical property analysis.

[0172] S31: Real-time Monitoring and Dynamic Adjustment

[0173] Real-time data monitoring: The force feedback control module receives real-time sensor data from the pressure testing machine module, including the current loading force F current and the deformation amount Δx.

[0174] Error calculation: According to the predetermined loading scheme, calculate the error Δerror between the current loading force and the target loading force,

[0175] Δerror = F desired - F current , where F desired is the predetermined target loading force, and F current is the current loading force;

[0176] Dynamic adjustment: By calling the force feedback control function, adjust the loading force according to real-time data feedback to ensure that the force during the loading process always meets the standard. F adjusted = FeedbackControl(F load , Δerror) where F adjusted : adjusted loading force, Fload : Current loading force, real-time data, Δerror: Loading error, calculated through real-time feedback data.

[0177] S32: Stress precise application

[0178] Stress adjustment: The force feedback control module adjusts the loading force F adjusted based on the adjusted loading force, and updates the control parameters of the loading unit in real time to ensure the precise application of stress during the loading process.

[0179] Stability confirmation: Through continuous monitoring and adjustment, the stability and accuracy of the loading process are ensured, and detection errors caused by loading force fluctuations are avoided.

[0180] S4, Data processing and analysis

[0181] Data processing and analysis is a key step in the sample detection process of the unmanned laboratory, its main goal is to clean, denoise and analyze the collected raw data, and calculate the mechanical performance indicators of the sample. This process not only ensures the accuracy and reliability of the data, but also provides scientific basis for subsequent report generation.

[0182] S41: Data cleaning and denoising

[0183] Data reception: The data acquisition and processing module receives real-time data from the sensor system, including force data, deformation data and strain data.

[0184] Data cleaning: Through the data cleaning function, remove noise and outliers to ensure the accuracy and reliability of the data. The cleaned data = DataCleaning (original data) Where, DataCleaning is the data cleaning function, used to remove noise and outliers.

[0185] S42: Mechanical property analysis

[0186] Stress and strain calculation: Based on the cleaned data, calculate the stress σ and strain of the sample, and the specific operation form of the stress and strain analysis function is as follows: σ = F load / A, = Δx / L, where σ: stress, mechanical property analysis result, : strain, calculated by deformation and original length, F load : loading force, A: cross-sectional area of the sample, Δx: deformation, L: original length of the sample. Performance index calculation: According to the stress and strain analysis function, calculate the compressive strength, tensile strength and other mechanical performance indicators of the sample.

[0187] S43: Analysis report generation

[0188] Analysis Result Summary: The calculated mechanical property indicators are summarized to generate a mechanical property analysis report for the sample. Report Storage: The analysis report is stored in the system to provide basic data for the subsequent report generation and optimization module.

[0189] S5, Report Generation and Optimization

[0190] Report generation and optimization is a key step in the unmanned laboratory sample detection process, its main goal is to generate a detection report according to the mechanical property analysis results, and mark abnormal data. This process not only ensures the accuracy and compliance of the report, but also provides detailed analysis of the detection results for users.

[0191] S51: Report Generation

[0192] Report Template Call: The report generation module calls the preset report template and fills in the report content according to the mechanical property analysis results.

[0193] Compliance Analysis: Compliance analysis of test results, compare standard values, mark whether it meets industry standards.

[0194] R final = GenerateReport(D analysis , A threshold ), where,

[0195] R final : final test report, D analysis : mechanical property analysis data, A threshold : abnormal data threshold, used to mark abnormal values;

[0196] S52: Abnormal Data Marking

[0197] Abnormal data detection: Detect abnormal data through statistical methods and mark in the report.

[0198] Report Optimization: Format optimization of the report to ensure readability and professionalism of the report.

[0199] S6, Optimization of Detection Scheme

[0200] Optimization of detection scheme is a key step in the unmanned laboratory sample detection process, its main goal is to predict the future mechanical properties of the sample based on historical detection data and real-time collected sample data, through machine learning algorithms and data analysis techniques, and generate an optimized detection scheme. This process not only improves detection efficiency, but also ensures the accuracy and reliability of the detection results.

[0201] S61: Data Analysis and Optimization Scheme Generation

[0202] Historical data call: the optimization detection scheme module calls historical detection data, combines with real-time collected sample data, and performs comprehensive analysis; performance prediction: the future mechanical performance of the sample is predicted through a machine learning algorithm. Optimization scheme generation: according to the analysis result, the optimized detection scheme is generated, and the detection parameters (loading time, loading rate and measurement accuracy) are adjusted. opt = OptimizeDetection(P prev , D analysis ) where:

[0203] P opt : optimized detection scheme; P prev : previous experimental scheme; D analysis : mechanical performance analysis data;

[0204] S62: optimization scheme application

[0205] Parameter adjustment: according to the optimized detection scheme, the loading parameters of the pressure testing machine module are adjusted; process optimization: the optimized detection scheme is applied to the subsequent detection process, forming a continuous optimization technical closed loop.

[0206] Through the above steps, the force feedback control and adjustment, data processing and analysis, report generation and optimization, and the optimization detection scheme module work together to ensure the efficient operation of the unmanned laboratory sample detection system. The functions of each module are closely connected to form a complete technical closed loop, ensuring the accuracy and reliability of the detection process.

[0207] Although embodiments of the present application have been shown and described, it will be understood by those having ordinary skill in the art that various changes, modifications, substitutions and alterations can be made without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. An image recognition based unmanned laboratory sample detection system, characterized in that, Comprise the following modules: Image recognition module; Pressure testing machine module; Force feedback control module; Data acquisition and processing module; Mechanical property analysis module; Report generation and optimization module; Optimized detection scheme module.

2. The image recognition-based unmanned laboratory sample detection system according to claim 1, wherein: The image recognition module is configured to identify and locate sample identification information through image recognition technology, and the identification information includes a two-dimensional code and a bar code on the sample; The pressure testing machine module is configured to apply vertical loading force and measure the stress and deformation of the sample in real time, and comprises a loading unit and a sensor system configured to measure the deformation and strain data of the sample in real time and transmit feedback data to the force feedback control module in real time; The force feedback control module is configured to receive real-time data from the pressure testing machine module, adjust the loading rate and loading strength based on the stress and strain data provided by the sensor system through multiple control algorithms, ensure that the force during the loading process meets the predetermined loading scheme, and realize dynamic adjustment during the loading process; The multiple control algorithms include: PID control algorithm to regulate the loading rate: v(t) = K p • e(t) + K i ∫0te(τ)dτ + K d d / td e(t), where v(t) is the loading rate, e(t) is the difference between the current loading force and the set value, K p ,K i ,K d are the proportional, integral and derivative gains; Incremental control algorithm for adjusting the loading intensity: F load (t) = F initial + ΔF(t), where F load (t) is the loading intensity at time t, F initial is the initial loading intensity, and ΔF(t) is the increment calculated from the current force and deformation data. A proportional feedback control algorithm dynamically adjusts the loading process in real time through feedback data: F adjusted = F current + K f · (F desired - F current ), wherein F adjusted is the adjusted loading intensity, F current is the current loading intensity, F desired is the predetermined target loading intensity, and K f is the feedback control gain; The data acquisition and processing module is configured to collect data from the pressure testing machine module and the sensor system, remove noise and abnormal data after preprocessing, generate a data set suitable for mechanical property analysis, and store and transmit the data in real time; The mechanical property analysis module is configured to calculate the mechanical property indicators of the sample including compressive strength and tensile strength based on the preprocessed data using a stress-strain analysis function, and the calculation result is obtained by the stress-strain analysis function; The report generation and optimization module is configured to automatically generate a detection report meeting the requirements of industry standards based on the output results of the mechanical property analysis module, and the report includes standard value comparison, compliance analysis and abnormal data labeling of the detection results; The optimized detection scheme module is configured to predict the future mechanical properties of the sample and generate an optimized detection scheme based on historical detection data and real-time sample data through machine learning algorithms and data analysis techniques, and the optimized scheme adjusts the detection parameters according to the specific performance requirements of the sample, including loading time, loading rate and measurement accuracy.

3. The image recognition based unmanned laboratory sample detection method according to any one of claims 1-2, characterized in that, Comprise the following steps: S1, sample preparation and positioning: scan the sample through image recognition technology, extract the identification information of the sample, and determine the subsequent detection parameters according to the information; S2, loading control and data acquisition: start the pressure testing machine and collect the stress and deformation data of the sample during the loading process; S3, force feedback control and adjustment: dynamically adjust the loading rate and loading strength through the feedback loop according to the real-time collected data to ensure that the loading process meets the set standard; S4, data processing and analysis: clean and denoise the collected raw data, and apply the processed data to the stress-strain analysis function to calculate the compressive strength and tensile strength mechanical property indicators of the sample; S5, report generation and optimization: automatically generate a standard detection report according to the analysis results, and mark abnormal data. S6, optimize the detection scheme: based on historical data and real-time data analysis, generate an optimized detection scheme, adjust the loading time, loading rate parameters.

4. The image recognition based unmanned laboratory sample detection method according to claim 3, characterized in that: The sample preparation and positioning includes: S11: The sample is placed on the detection platform, the image recognition module is started, the sample image is collected and processed; S12: The image recognition module extracts identification information from the collected image and performs data analysis; by calling the image recognition function, the identification information in the image will be extracted and analyzed as ID sample , providing basic data for subsequent detection operations; the specific operation form of the image recognition function is as follows: I gray = Grayscale(I input ); I binary = Threshold(I gray ,T); P identifier = EdgeDetection(I binary ); ID sample = Decode(P identifier ); wherein I gray : a grayscale image, I binary : a binary image, P identifier : an edge detection result of the identifier, ID sample : sample identification information, including a two-dimensional code, a barcode, I input : an input image, captured by a camera, T: a binary threshold value; S13: The identified identification information is transmitted to the pressure testing machine module to determine the subsequent loading conditions.

5. The image recognition based unmanned laboratory sample detection method of claim 3, wherein: The loading control and data acquisition includes: S21: Based on the identified sample information, set the initial loading conditions, including loading force and loading rate; this process is achieved by calling the loading control and data acquisition functions, based on the sample's identification information (ID). sample And the loading time T in the preset loading conditions load Calculate the initial loading force F load And set the loading conditions; the specific operation form of the loading control and data acquisition functions is as follows: F load =LoadCalculation(I sample ,T load ,Δx); wherein F load : loading force, I sample : sample identification information, obtained by an image recognition module, T load : loading time, set according to sample characteristics, Δx: deformation amount, measured by a sensor; S22: Start the pressure testing machine and begin the loading process, while collecting real-time force and deformation data of the sample through the sensor; S23: The sensor data is transmitted to the data acquisition and processing module in real time for preliminary data processing and cleaning.

6. The image recognition based unmanned laboratory sample detection method of claim 3, wherein: The force feedback control and adjustment includes: S31: Real-time monitoring of sensor data, dynamic adjustment of loading rate and loading intensity to ensure that the loading process meets the predetermined standards, and through the call of the force feedback control function, the loading force is adjusted according to the real-time data feedback to ensure that the force in the loading process always meets the standards, and the specific operation form of the force feedback control function is as follows: F adjusted = FeedbackControl(F load , Δerror) where F adjusted : adjusted load force, F load : current load force, real-time data, Δerror: load error, calculated from real-time feedback data; S32: The force feedback control module confirms the accurate application of stress in the loading process through real-time adjustment.

7. The image recognition based unmanned laboratory sample detection method of claim 3, wherein: The data processing and analysis includes: S41: The collected raw data is transmitted to the data processing module for noise removal and outlier detection; the raw data is cleaned through the data cleaning function to remove noise and outliers, ensuring the accuracy of the data; S42: The cleaned data enters a mechanical property analysis module, and the stress-strain analysis function is used to calculate the compressive strength, tensile strength and other mechanical property indexes of the sample; the stress-strain analysis function is called, and the stress σ and strain of the sample are calculated based on the cleaned data, and the specific operation form of the stress-strain analysis function is as follows: σ = F load / A, = Δx / L, wherein σ: stress, mechanical property analysis result, : strain, calculated by deformation amount and original length, F load : loading force, A: cross-sectional area of the sample, Δx: deformation amount, L: original length of the sample; S43: Generate a mechanical property analysis report for the sample based on the calculation results to provide a basis for subsequent report generation.

8. The image recognition based unmanned laboratory sample detection method of claim 3, wherein: The report generation and optimization sub-steps include: S51: The report generation module automatically generates a detection report based on the mechanical property analysis results, including compliance analysis of the detection results; this process is implemented through the report generation function, and its specific operation form is as follows: R final = GenerateReport(D analysis , A threshold ), where, R final : final test report, D analysis : mechanical property analysis data, A threshold : threshold for abnormal data, for marking outliers; S52: Abnormal data in the report is marked.

9. The image recognition based unmanned laboratory sample detection method of claim 3, wherein: The optimized detection scheme includes: S61: Optimize the detection scheme by analyzing historical data and real-time data; generate optimized experimental conditions, including adjusting the loading time and loading rate; this process is implemented through the detection scheme optimization function, and its specific operation form is as follows: P opt = OptimizeDetection(P prev , D analysis ) where: P opt : optimized detection scheme; P prev : previous experimental protocol; D analysis : Mechanical property analysis data; S62: Automatically generate an optimized detection process and detection parameters based on the optimization results.

Images