Coking test robot automation test method and test system based on coking test robot

Abstract

The application relates to a coking test robot automatic test method and a test system based on a coking test robot, and belongs to the technical field of coking test. The coking test robot comprises a mechanical arm and a camera installed on the mechanical arm. The method comprises the following steps: acquiring a test instruction for a sample to be tested; calling a test step according to the test instruction; determining a target object, an operation action, a starting point and an ending point of the operation action of the mechanical arm in each test step; executing the test step, the mechanical arm reaches the starting point, an image containing the target object is acquired through the camera, the target object is positioned by analyzing and identifying the image, the mechanical arm executes the corresponding operation action on the target object and moves to the ending point, and all test steps are completed until a test result of the sample to be tested is obtained. The application has the effect of improving the coking test efficiency.

Description

Technical Field

[0001] This application relates to the technical field of coking testing, and in particular to an automated testing method for coking testing using a robot and a testing system based on the coking testing robot. Background Technology

[0002] In the coking industry, coal sample testing is an essential and crucial step. Testing determines the type and quality of the coal, providing a basis for coal blending plans. It also allows for monitoring coal quality fluctuations, timely adjustments to process parameters, ensuring stable production, and guaranteeing that the final product meets the quality standards of downstream industries. Among these processes, testing is the fundamental operation; all analytical items involving composition, content, and proportion require precise data obtained through testing.

[0003] In traditional processes, samples are analyzed using laboratory instruments, and the data is then manually read and recorded in paper forms or computer documents. However, manual testing suffers from low efficiency, susceptibility to contamination, and poor traceability. Given the increasingly stringent requirements for testing data in the coking industry, this has become a bottleneck restricting production efficiency and quality control. Summary of the Invention

[0004] To improve the efficiency of coking testing, this application provides an automated testing method using a coking robot and a testing system based on the coking robot.

[0005] In a first aspect, this application provides an automated testing method using a coking testing robot, wherein the coking testing robot includes a robotic arm and a camera mounted on the robotic arm, and adopts the following technical solution:

[0006] Obtain testing instructions for the sample to be tested;

[0007] Retrieve the test steps according to the test instructions;

[0008] Determine the target object, operation action, start point, and end point of the operation action for each of the aforementioned testing steps;

[0009] During the testing process, the robotic arm reaches the starting point, acquires an image containing the target object through the camera, analyzes and identifies the image to locate the target object, and then performs corresponding operations on the target object and moves to the endpoint until all testing steps are completed, thus obtaining the test results of the sample to be tested.

[0010] By adopting the above technical solution, this method uses the robotic arm of the coking testing robot in collaboration with a camera to automatically retrieve steps according to testing instructions, accurately identify and locate target objects, and complete the operation from start to finish. The entire testing process is automated, reducing human intervention, improving operational accuracy and efficiency, avoiding human error, ensuring accurate and reliable test results, and improving the standardization and safety of the coking testing process. It is suitable for batch sample testing scenarios and optimizes the testing process.

[0011] Furthermore, the testing steps include obtaining the identification number of the sample to be tested, and determining the target object, operation action, start point, and end point of the operation action for each testing step by the robotic arm includes:

[0012] The target object is the sample to be tested, the operation is to move the sample to be tested, the starting point is above the sample rack, and the ending point is above the barcode scanning device.

[0013] Furthermore, the analysis and identification of the image to locate the target object includes:

[0014] Identify the information code in the image and obtain the target object features contained in the information code;

[0015] Retrieve the CNN neural network model used to identify the target object corresponding to the features of the target object;

[0016] The image is identified based on the CNN neural network model to determine the target object in the image;

[0017] Establish a planar coordinate system based on the image and determine the first coordinate set corresponding to the target object;

[0018] Determine the first area of ​​the target object in the image based on the first coordinate set;

[0019] Calculate the ratio of the first area to the actual area of ​​the target object;

[0020] The distance between the robotic arm and the target object is determined based on the ratio.

[0021] A three-dimensional coordinate system is established based on the planar coordinate system with the location of the robotic arm as the origin. Based on the first coordinate set and the distance, the second coordinate set of the target object in the three-dimensional coordinate system is determined.

[0022] By employing the above technical solution, features are obtained through identification codes, and the target object is located using a CNN model. Distance is determined through coordinate calculation and area ratio analysis, ultimately establishing three-dimensional coordinates. This achieves precise target object positioning, providing accurate spatial information for robotic arm operations and improving the accuracy and efficiency of automated testing.

[0023] Further, before determining the first area of ​​the target object in the image based on the first coordinate set, the method includes:

[0024] Determine whether the planar shape of the target object in the image has perspective distortion based on the first coordinate set;

[0025] If so, the lens distortion parameters are obtained through camera calibration, and the image is corrected for distortion to obtain an ideal projection image without distortion.

[0026] Identify key feature points of a target object in an image and determine the coordinates of the key feature points;

[0027] By applying the perspective transformation matrix and combining the coordinates of key feature points with the actual size of the target object, the correspondence between image coordinates and three-dimensional coordinates is established, and the pitch and yaw angles of the camera are calculated.

[0028] The camera angle is adjusted according to the pitch angle and the yaw angle so that the camera can shoot the target object in parallel, thereby obtaining a new image with the target object, and the first coordinate set corresponding to the target object is determined repeatedly.

[0029] By employing the above technical solution, perspective distortion and lens distortion of the target object image are judged and corrected, the camera angle is adjusted to parallel shooting, and the coordinate set is redefined. This effectively eliminates imaging deviations, improves the accuracy of target object area calculation, and lays a precise foundation for subsequent distance and 3D coordinate determination.

[0030] Furthermore, before determining the distance between the robotic arm and the target object based on the ratio, the method further includes:

[0031] Obtain multiple sets of experimental data, including ratios and corresponding distances;

[0032] By fitting the experimental data with the calculation formula, a simulation model for predicting distances is obtained.

[0033] The fitted simulation model is used to predict new experimental data, and error indices are calculated.

[0034] If the error index is greater than the preset value, the model is corrected and refitted to obtain a new simulation model.

[0035] By adopting the above technical solution, a distance prediction model was fitted using experimental data and then validated and optimized with new data to ensure that the model error meets the standards. This improves the accuracy of distance calculation between the robotic arm and the target object, provides reliable data support for 3D positioning, and further guarantees the precision of automated testing operations.

[0036] Furthermore, if the image is identified based on the CNN neural network model, and it is determined that at least two target objects exist in the image, then the method further includes:

[0037] Identify the target object in the image;

[0038] Determine whether the current testing step involves retrieving a sample from the sample rack;

[0039] If so, then assign a number to each of the aforementioned target objects to be determined;

[0040] If a target object is determined from the undetermined target objects for the first time, then the undetermined target object with the smallest number is determined as the target object;

[0041] If the target object is not determined from the list of undetermined target objects for the first time, then the undetermined target object corresponding to the number with the smallest difference from the previous number is determined as the target object in ascending order of number.

[0042] Otherwise, a planar coordinate system is established based on the image to determine the third coordinate set of each of the proposed target objects;

[0043] Obtain the historical coordinates of the target object in each historical image, and determine the coordinate range of the target object based on each historical coordinate;

[0044] The similarity between the target object to be determined and the target object is determined based on the degree of overlap between each of the third coordinate sets and the coordinate range.

[0045] Identify the background objects in the image that are adjacent to each of the target objects to be determined;

[0046] Determine the correlation between the target object to be determined and each of the background objects;

[0047] An evaluation score is calculated based on the similarity and relevance according to a preset weight, and the candidate object with the highest evaluation score is determined as the target object.

[0048] By adopting the above technical solution, for multi-target recognition scenarios, target objects are accurately screened through multi-dimensional judgment based on factors such as numbering order, historical coordinate range, and relevance to the background. This improves the accuracy of robotic arm operations in complex environments and ensures the orderly conduct of testing procedures.

[0049] Further, determining the correlation between the target object and the background object includes:

[0050] If there are multiple background objects, a classification model is applied to analyze the background objects and determine whether they are related to the current testing step.

[0051] If they are not related, the correlation between the target object to be determined and the background object will not be calculated.

[0052] If relevant, determine the positional relationship and distance between each target object to be determined and the background object;

[0053] The correlation index between each of the target objects to be determined and the background objects is determined according to the positional relationship;

[0054] The objects to be determined are sorted according to their distance from closest to furthest to obtain a first sequence, and a multiplier is assigned to each object to be determined according to the order of the first sequence.

[0055] By multiplying each of the aforementioned correlation indices by its corresponding multiplier, the primary correlation between the target object and the background object is calculated.

[0056] The mean of the primary correlation between the target object and each background object is calculated to obtain the correlation score.

[0057] By employing the above technical solution, relevant background objects are screened using a classification model, a correlation index is determined based on positional relationships, and a primary correlation degree is calculated by multiplying by distance and taking the average. This accurately quantifies the association between the target and the background, improving the accuracy of target recognition in multi-target scenarios.

[0058] Furthermore, prior to performing the aforementioned testing steps, the method further includes:

[0059] If the current testing step is not the first testing step, then obtain the first image information of the target object including the previous testing step;

[0060] The target detection model is applied to identify the actual position of the target object in the first image information, and the distance between the actual position and the preset standard position is determined;

[0061] The degree of completion of the previous test step is determined based on the distance.

[0062] If the completion rate of the operation is lower than the preset value, correction operation information is generated based on the actual position and standard position of the target object;

[0063] The robotic arm executes the corrective operation information to bring the target object to the standard position.

[0064] By employing the above technical solution, the completion rate of the operation is determined by detecting the distance between the actual position of the target object in the previous step and the standard position. If the standard is not met, correction information is generated to drive the robotic arm to adjust. This ensures that each step of the operation meets the standard, improving the consistency and accuracy of the testing process.

[0065] Secondly, this application provides an automated testing device for coking analysis using a robot, which adopts the following technical solution:

[0066] The test instruction acquisition module is used to acquire test instructions for the samples to be tested.

[0067] The test step retrieval module is used to retrieve the test steps according to the test instructions.

[0068] The testing step analysis module is used to determine the target object, operation action, start point and end point of the operation action of the robotic arm in each testing step;

[0069] The testing step execution module is used to execute the testing steps. The robotic arm reaches the starting point, acquires an image containing the target object through a camera, analyzes and identifies the image to locate the target object, and causes the robotic arm to perform corresponding operation actions on the target object and move to the endpoint until all testing steps are completed, and the test results of the sample to be tested are obtained.

[0070] By adopting the above technical solution, the robotic arm of the coking testing robot works in conjunction with the camera. The testing step retrieval module automatically retrieves the steps according to the testing instructions, the testing step analysis module accurately identifies and locates the target object, and completes the operation from the start to the end. The testing step execution module automatically executes the testing steps throughout the entire process, reducing manual intervention, improving operational accuracy and efficiency, avoiding human error, ensuring accurate and reliable test results, and improving the standardization and safety of the coking testing process. It is suitable for batch sample testing scenarios and optimizes the testing process.

[0071] Thirdly, this application provides a coking testing robot, which adopts the following technical solution:

[0072] A coking testing robot includes a body, a walking mechanism installed at the bottom of the body, a sample rack on the body, and samples to be tested and samples after testing placed in the sample rack; a robotic arm is also installed on the body, and a camera is installed above the gripping part of the robotic arm.

[0073] It also includes: at least one processor;

[0074] At least one memory;

[0075] At least one computer program, wherein the at least one computer program is stored in the memory and configured to be executed by the at least one processor, the at least one computer program being configured to: perform an automated coking testing robot method as described in any one of the first aspects.

[0076] By adopting the above technical solution, the processor executes the computer program in the memory, and through the collaboration of the robotic arm and camera of the coking testing robot, it automatically retrieves the steps according to the testing instructions, accurately identifies and locates the target object, and completes the operation from the start to the end. The entire testing process is automated, reducing human intervention, improving operational accuracy and efficiency, avoiding human error, ensuring accurate and reliable test results, and improving the standardization and safety of the coking testing process. It is suitable for batch sample testing scenarios and optimizes the testing process.

[0077] Fourthly, this application provides a testing system based on a coking testing robot, employing the following technical solution:

[0078] Coal Sample Management Module: Monitors the coal sample preparation process and enters the numbering information generated on the finished coal sample packaging according to preset rules;

[0079] Coal sample analysis module: Enables the analysis robot to perform the method described in any one of claims 1-7 to complete the coal sample analysis;

[0080] Test result acquisition module: Receives test results from various testing devices via the communication interface;

[0081] Laboratory result analysis module: compares the laboratory results with standardized test values, identifies and marks abnormal data;

[0082] Laboratory Result Review Module: Sends laboratory results and abnormal data to the administrator's account and obtains the administrator's review results for the laboratory results and abnormal data;

[0083] Laboratory review module: Responds to the review instructions issued by the administrator, generates the review process, obtains the review results, and sends the review results to the administrator's account for review;

[0084] Data statistics module: Generates statistical tables based on test results;

[0085] The coal sample management module oversees the coal sample preparation process, including:

[0086] Obtain video information of the coal sample preparation site;

[0087] The video is input into the target detection model to identify and mark core items, which include one or more of workers, equipment, tools and installation equipment.

[0088] By applying a behavior recognition model to identify tagged videos, the continuous actions of workers and the status of equipment can be obtained.

[0089] The continuous actions and equipment status are compared with preset rules to determine whether the worker's operation is in accordance with regulations; if not, an alarm message is generated.

[0090] By adopting the above technical solutions, the testing robot strictly follows the preset program to perform weighing, heating and other operations, avoiding human visual errors and endpoint judgment deviations. It also automatically reads data in conjunction with high-precision equipment, eliminating transcription errors and ensuring that the errors of key indicators such as ash and sulfur content are controllable. It shortens the single-sample testing cycle, ensures testing safety, and reduces the risk of workplace injuries. The data is automatically uploaded to the system and linked with the coal sample code to form a traceability chain, which facilitates quality traceability and data analysis and reduces the loopholes of manual recording.

[0091] In summary, this application includes at least one of the following beneficial technical effects:

[0092] 1. Through the collaboration of the robotic arm and camera of the coking testing robot, the steps are automatically retrieved according to the testing instructions, the target object is accurately identified and located, and the operation is completed from the start to the end. The testing steps are executed automatically throughout the process, reducing human intervention.

[0093] 2. Improve operational precision and efficiency, avoid human error, ensure accurate and reliable test results, and enhance the standardization and safety of the coking testing process. Attached Figure Description

[0094] Figure 1 This is a schematic diagram of the structure of the coking testing robot in this application embodiment when performing testing operations.

[0095] Figure 2 This is a schematic diagram of the electrical control structure of the coking testing robot in the embodiments of this application.

[0096] Figure 3 This is a flowchart illustrating the automated testing method for coking analysis using a robot, as described in this application.

[0097] Figure 4 This is a structural block diagram of the automated testing device for coking testing robots in the embodiments of this application.

[0098] Figure 5 This is a structural block diagram of the testing system based on a coking testing robot in the embodiments of this application.

[0099] Reference numerals: 1. Body; 2. Walking mechanism; 3. Sample rack; 4. Sample to be tested; 5. Robotic arm; 6. Camera; 7. Worktable; 8. Weighing balance; 81. Balance door; 82. Display panel; 9. Placement platform. Detailed Implementation

[0100] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0101] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0102] This application provides a coking testing robot, referring to... Figure 1 and Figure 2 The coking testing robot includes a body 1, a walking mechanism 2 installed at the bottom of the body 1 to enable the coking testing robot to move, a sample rack 3 on the body 1 for placing the sample 4 to be tested and the sample after testing, a robotic arm 5 is also installed on the body 1, and a camera 6 is installed above the gripping part of the robotic arm 5.

[0103] Camera 6 captures an image of the position to be operated by the gripper, and the position is located through image analysis. The robotic arm 5 can operate flexibly to complete various operations.

[0104] Coking testing is a crucial step in the quality analysis of raw materials, intermediate products, and final products during the coking production process. The procedures vary depending on the specific product being tested. For example, coke, one of the main coking products, is tested for ash content, volatile matter, fixed carbon, sulfur content, and mechanical strength. Coking coal is primarily tested for moisture, ash content, volatile matter, caking index, and maximum cuticle thickness. Coal tar is primarily tested for density, moisture, ash content, and distillation range. Coal gas is primarily tested for calorific value and composition (CO, CO2, H2, CH4, etc.).

[0105] Furthermore, depending on the test objects and test items, corresponding testing equipment is set on the workbench 7 in each laboratory. Taking the weighing item, which is essential in the testing process for each test object, as an example, the testing equipment in the weighing laboratory includes a weighing balance 8 and an auxiliary weighing platform 9, etc. The coking testing robot performs the testing operation in the laboratory and automatically completes the weighing item.

[0106] Specifically, the weighing balance 8 and the placement platform 9 are placed on the workbench 7. One side of the weighing balance 8 has an openable balance door 81. The robotic arm 5 can open the balance door 81 to place the sample 4 to be tested into the weighing balance 8 for weighing. The weighing balance 8 has a display panel 82 that displays the weight of the sample 4. The coking testing robot can read the weighing data by photographing the display panel 82 with the camera 6. The placement platform 9 is used to temporarily place the sample 4 when the robotic arm 5 opens the balance door 81, and is located on the side where the balance door 81 is located.

[0107] To facilitate rapid location of the target object, an information code containing the object's characteristics can be affixed near it. Therefore, after camera 6 captures an image, the coagulation analysis robot identifies the information code in the image to obtain the target object's characteristics. These characteristics include the object's name and object number.

[0108] Reference Figure 4 The coking and testing robot 100 also includes a processor 101 and a memory 103. The robotic arm 5, camera 6, and memory 103 are all connected to the processor 101, such as via a bus 102. Optionally, the electronic device 100 may also include a transceiver 104. It should be noted that in practical applications, the transceiver 104 is not limited to one type, and the structure of the electronic device 100 does not constitute a limitation on the embodiments of this application.

[0109] Processor 301 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 101 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0110] Bus 102 may include a pathway for transmitting information between the aforementioned components. Bus 102 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 102 may be divided into address bus, data bus, control bus, etc.

[0111] The memory 103 may be a ROM (Read Only Memory) or other type of static storage device capable of storing static information and instructions, RAM (Random Access Memory) or other type of dynamic storage device capable of storing information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0112] The memory 103 is used to store application code that executes the solution of this application, and its execution is controlled by the processor 301. The processor 101 is used to execute the application code stored in the memory 103 to implement the content shown in the foregoing method embodiments.

[0113] Figure 2 The coking and testing robot 100 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0114] Furthermore, this application discloses an automated testing method using a coking robot. (Refer to...) Figure 3 This method is applied to the testing operations of a coking laboratory robot, and is executed by the coking laboratory robot. It includes (steps S201 to S204):

[0115] Step S201: Obtain the testing instructions for the sample to be tested.

[0116] Specifically, the coking testing robot can perform a number of testing tasks other than weighing. For example, it can significantly improve efficiency, enhance monitoring accuracy and consistency, and ensure personnel safety in processes involving manual operation, such as basic component analysis, intermediate product testing, and finished product testing.

[0117] Step S202: Retrieve the test steps according to the test instructions.

[0118] Specifically, after receiving the testing instructions, the coking testing robot retrieves the various testing steps. The testing instructions include a series of testing steps that the coking testing robot must complete sequentially.

[0119] Taking the testing instruction as an example of a weighing item, the testing steps include a series of operations such as resetting the balance, taking the sample to be tested from the sample rack, obtaining the sample number information, placing the sample to be tested on the placement platform, opening the balance door, transferring the sample to be tested from the placement platform into the balance, closing the balance door, reading the weighing data, opening the balance door, taking the sample out of the balance and placing it on the placement platform, and putting the sample back into the sample rack.

[0120] Step S203: Determine the target object, operation action, start point and end point of the operation action for each testing step.

[0121] Specifically, each operation step requires the robotic arm to perform corresponding actions, starting from the starting point, performing operations on the target object and reaching the endpoint.

[0122] For example, in the test procedure of resetting the balance, the target object is the balance reset button, the operation is to press the button, the starting point is the current position, and the ending point is the balance reset button.

[0123] Taking the sample to be tested: The target object is actually the cup containing the coal to be tested. Here, the cup containing the coal to be tested will be referred to as the sample to be tested. The operation is to pick up the sample to be tested, starting from the sample rack and ending at the top of the sample rack.

[0124] Obtain the sample number information: The target object is the sample to be tested, the operation is to move the sample to be tested, the starting point is above the sample rack, and the ending point is above the barcode scanner.

[0125] Similarly, the target object, operation action, start point and end point of the operation action are set for each testing step according to actual needs.

[0126] Step S204: Perform the testing steps. The robotic arm reaches the starting point, acquires an image containing the target object through the camera, analyzes and identifies the image to locate the target object, and causes the robotic arm to perform corresponding operations on the target object and move to the endpoint until all testing steps are completed, and the test results of the sample to be tested are obtained.

[0127] Specifically, the coking testing robot generates operation instructions based on the testing steps, and generates corresponding stroke codes for the robotic arm for each step, enabling the robotic arm to accurately complete the operation. The coking testing robot executes the testing steps sequentially, completing each step individually, and after completing all testing steps, obtains the test results for the sample.

[0128] The test results of the sample to be tested include the sample number and the corresponding test data. The sample number comes from the code set on the bottom of the cup. The scanning device obtains the number information after scanning the code.

[0129] Taking the weighing project as an example, after the sample to be tested is operated by the robotic arm, it is placed in the weighing balance. After the weighing balance weighs the sample, the weighing data is displayed on the display panel. When the robotic arm performs the step of reading the weighing data, it takes an image of the weighing balance display panel through a camera, obtains the corresponding weighing data through data recognition, and matches and saves it one by one with the number information.

[0130] The coking testing robot also locates the target object by analyzing and recognizing images, including (steps S11 to S18):

[0131] Step S11: Identify the information code in the image and obtain the target object features contained in the information code.

[0132] Specifically, for example, information codes can be set on the attachments of frequently operated target objects such as placement platforms, weighing balance doors, and reset buttons to facilitate quick location of the target object.

[0133] Step S12: Retrieve the CNN neural network model used to identify the target object's features.

[0134] Specifically, the coking and testing robot pre-defines the target object features and the corresponding CNN neural network model for each type of target object, thus facilitating the direct acquisition of the CNN neural network model used to identify the target object after obtaining the target object features.

[0135] Step S13: Identify the target object in the image based on the CNN neural network model.

[0136] Specifically, the coking and testing robot is pre-trained to obtain a CNN neural network model, collects labeled image datasets, and divides them into training / validation / test sets; a CNN network containing convolutional, pooling, and fully connected layers is built, iteratively trained using the training set, backpropagation optimizes the parameters, the validation set is used for tuning to prevent overfitting, and the test set is used to evaluate the model performance.

[0137] Step S14: Establish a planar coordinate system based on the image and determine the first coordinate set corresponding to the target object.

[0138] Specifically, if only one target object is identified in the image, the first coordinate set corresponding to the target object is directly obtained. However, in actual work, if other items are placed in the work area, it may affect the image recognition and location of the target object. For example, if other test samples are placed on the table, the image information acquired when placing the sample to be tested into the weighing balance will also include the other test samples. To further analyze and determine the target object, the method further includes (steps S1401 to S1411):

[0139] Step S1401: Determine the target object in the image.

[0140] Specifically, the coking analysis robot will identify each target object from the image as a target object to be determined.

[0141] Step S1402: Determine whether the current testing step is to retrieve the sample to be tested from the sample rack.

[0142] If so, proceed to step S1403: assign a number to each target object to be determined.

[0143] Specifically, in the testing step of retrieving the sample to be tested from the sample rack, there may be multiple samples to be tested from the same batch on the sample rack. In this step, the target object is the sample to be tested, and multiple samples to be tested can be identified from the image information. At this time, each target object is assigned a number.

[0144] Step S1404: If a target object is determined from the undetermined target objects for the first time, then the undetermined target object with the smallest number is determined as the target object.

[0145] Step S1405: If this is not the first time a target object has been determined from the list of undetermined target objects, then the undetermined target object corresponding to the number with the smallest difference from the previous number is determined as the target object in ascending order of number.

[0146] Specifically, when a target object is first identified from the pool of potential target objects, it means that the sample to be tested is first taken from the sample rack, and the sample with the smallest number is taken as the first target object. If it is not the first time a target object is identified from the pool of potential target objects, then the sample to be tested has already been taken from the sample rack, that is, the samples to be tested are taken in order of their numbers.

[0147] If the current testing step does not involve retrieving the sample from the sample rack, then multiple target objects will not necessarily exist in other testing steps. If multiple target objects do exist, further analysis is needed to determine which one should be used as the target object.

[0148] Otherwise, proceed to step S1406: establish a planar coordinate system based on the image and determine the third coordinate set of each target object to be determined.

[0149] Step S1407: Obtain the historical coordinates of the target object in each historical image, and determine the coordinate range of the target object based on each historical coordinate.

[0150] Specifically, the coking and testing robot acquires historical images of the same target object taken from the same starting point, and then determines the historical coordinates of the target object in the historical images. The union of the historical coordinates is obtained, which is the coordinate range of the target object. In other words, during normal testing operations, the target object can only appear within the coordinate range.

[0151] Step S1408: Determine the similarity between the target object to be determined and the target object based on the overlap between each third coordinate set and the coordinate range.

[0152] Furthermore, the third coordinate set represents the coordinate range of the target object. The ratio of the area of ​​the overlapping range to the area of ​​the coordinate range is used to determine the similarity. That is, when the third coordinate set completely falls within the coordinate range, the similarity between the third coordinate set and the target object is 100%. The smaller the overlapping range, the smaller the similarity.

[0153] Step S1409: Determine the background objects in the image that are adjacent to each target object to be determined.

[0154] Furthermore, besides excluding potential target objects from their coordinate range, the relationship between potential target objects and background objects can be used to further determine the target object. Taking a weighing process as an example, when transferring the sample to be tested from the platform to the weighing balance, the target object is the sample itself. First, an image of the platform is captured, showing both the sample and the platform, with the sample currently placed on it. However, if other samples are also placed on the platform, multiple potential target objects will be identified in the image. The platform is then considered the background object, and the target object is determined based on the correlation analysis between each potential target object and the background object. For example, a sample located on the platform is more likely to be the target object than a sample located next to the platform.

[0155] The coking and testing robot uses a CNN neural network to identify various objects in an image, and then analyzes and determines the background objects adjacent to each target object.

[0156] Step S1410: Determine the correlation between the target object and each background object.

[0157] Specifically, when there is a background object in the image, the correlation can be determined directly based on the positional relationship and distance between the target object and the background object. However, there may be multiple background objects in the image. Therefore, when determining the correlation, it is also necessary to consider whether the background object itself is related to the current test step. Therefore, step S1410 specifically includes (steps S21 to S27):

[0158] Step S21: If there are multiple background objects, apply a classification model to analyze the background objects and determine whether they are related to the current test step.

[0159] Specifically, the coking and testing robot is pre-trained with a classification model to identify relevant or irrelevant object categories and collect labeled data containing the objects. The classification model can be a CNN neural network model. The labeled data is divided into training sets to train the classification model until the classification model can stably classify whether the background object is related to the testing steps.

[0160] If they are not related, proceed to step S22: do not calculate the correlation between the target object and the background object.

[0161] If relevant, proceed to step S23: determine the positional relationship and distance between each target object to be determined and the background object.

[0162] Specifically, the distance can be selected as the straight-line distance between the center points of the target object and the background object in the coordinate system. Positional relationships include far apart, adjacent, and overlapping. Far apart means that the distance between the target object and the background object is greater than a threshold, adjacent means that the distance between the target object and the background object is less than or equal to the threshold, and overlapping means that the coordinate ranges of the target object and the background object overlap.

[0163] Step S24: Determine the correlation index between each target object and the background object according to their positional relationship.

[0164] Specifically, the coking analysis robot presets a corresponding correlation index for each positional relationship and assigns a gradually decreasing correlation index in the order of overlap, adjacency, and distance. For example, the correlation indices assigned in sequence are 3, 2, and 1.

[0165] Step S25: Sort each target object in order of distance from nearest to farthest to obtain the first sequence, and assign multipliers to each target object in the order of the first sequence.

[0166] Specifically, each target object to be determined is assigned a gradually decreasing multiplier according to the order of the first sequence. For example, if there are 3 target objects to be determined in the first sequence, the corresponding multipliers are 3, 2, and 1.

[0167] Step S26: Multiply each correlation index by its corresponding multiplier to calculate the primary correlation between the target object and the background object.

[0168] Step S27: Calculate the mean of the primary correlation between the target object and each background object to obtain the correlation.

[0169] For example, if the primary correlation between target object A and background object 1 is 6, and the primary correlation between target object A and background object 2 is 9, then the calculated average correlation is 7.5. Similarly, if the correlation between target object B and target object C is 6, then target object A has the highest correlation with the background objects.

[0170] Step S1411: Calculate and determine the evaluation score according to the similarity and relevance according to the preset weight, and determine the pending target object with the highest evaluation score as the target object.

[0171] Specifically, the coking analysis robot is configured with preset weights as needed. Similarity and relevance are weighted and calculated according to these preset weights to obtain an evaluation score. Higher similarity and relevance result in a higher evaluation score. The candidate object with the highest evaluation score is essentially identified as the target object.

[0172] Step S15: Determine the first area of ​​the target object in the image based on the first coordinate set.

[0173] Furthermore, before executing step S15, the image needs to be reviewed to check whether there is perspective distortion in the image of the target object due to the camera not being directly facing the target object. If so, it needs to be corrected in time to avoid affecting the accuracy of calculating the distance between the robotic arm and the target object, including (steps Sa to Se).

[0174] Step Sa: Determine whether the planar shape of the target object in the image has perspective distortion based on the first coordinate set.

[0175] Specifically, perspective distortion causes the geometric relationships of a planar graphic to deviate from their true properties. For example, a rectangle may appear as a trapezoid in perspective, opposite sides may no longer be parallel, and the angle between adjacent sides may not be 90°. By analyzing the geometric relationships of the first coordinate convergence point, such as side length ratios, angles, and whether parallel lines intersect, it can be determined whether the graphic is subject to perspective distortion due to the camera's shooting angle. If the planar image has no perspective distortion, then the planar graphic of the target object in the image should have the same shape as the front view of the target object. The electronic device presets a front view of each target object and compares the planar graphic with the front view shape. If the shapes match, there is no perspective distortion; otherwise, perspective distortion is confirmed to exist.

[0176] If so, proceed to step Sb: obtain lens distortion parameters through camera calibration, correct image distortion, and obtain an ideal projection image without distortion.

[0177] Specifically, the lens's own optical characteristics, such as radial and tangential distortion, can cause image distortion. This distortion is different from perspective distortion and needs to be eliminated first. Camera calibration calculates the lens's distortion parameters by photographing a calibration plate of known size; then, these parameters are used to correct the image distortion, resulting in an "ideal projected image" that eliminates lens optical errors.

[0178] Step Sc: Identify the key feature points of the target object in the image and determine the coordinates of the key feature points.

[0179] Specifically, key feature points are representative points on a target object, such as vertices and edge intersections, whose true spatial locations are known or can be determined by the object's dimensions. Feature detection algorithms are used to identify these key feature points of the target object from the image, and then their specific coordinates in a planar coordinate system are extracted.

[0180] Step Sd: Apply the perspective transformation matrix, combine the coordinates of key feature points and the actual size of the target object, establish the correspondence between image coordinates and 3D coordinates, and calculate the camera's pitch and yaw angles.

[0181] Specifically, the perspective transformation matrix describes the projection relationship from a 3D object to a 2D image. Given the image coordinates and true 3D dimensions of key feature points, matrix operations can establish the correspondence between image coordinates and 3D world coordinates. The camera's pitch and yaw angles are the core angular parameters affecting perspective projection; by solving this correspondence, these two angles can be deduced.

[0182] Step Se: Adjust the camera angle according to the pitch and yaw angles so that the camera is parallel to the target object to obtain a new image with the target object, and repeat to determine the first coordinate set corresponding to the target object.

[0183] Specifically, adjusting the camera angle aims to eliminate perspective distortion: by correcting the pitch and yaw angles, the camera's optical axis is kept parallel to the target object's plane. In the new image captured at this point, the planar shape of the target object is closer to its true shape. Subsequently, the target object in the new image is re-identified to determine a new first coordinate set, forming an iterative optimization process until the image shows no significant perspective distortion.

[0184] Step S16: Calculate the ratio of the first area to the actual area of ​​the target object.

[0185] Step S17: Determine the distance between the robotic arm and the target object based on the ratio.

[0186] Specifically, in order for the robotic arm to accurately grasp the target object, it is necessary to accurately determine the distance between the robotic arm and the target object, so as to generate operation actions based on the distance.

[0187] The robotic arm takes an image of the target object. The size of the target object in the image is proportional to its actual size. As the camera moves closer, the area of ​​the target object in the image gradually increases.

[0188] The first coordinate set includes the outline of the target object and the coordinates of each point inside the outline. This first coordinate set facilitates the calculation of the first area of ​​the target object. The coking and testing robot pre-determines the distance corresponding to each ratio, and then, after determining the ratio of the first area to the actual area of ​​the target object from the image, obtains the distance between the robotic arm and the target object.

[0189] Furthermore, during step S17, a simulation model for estimating distance can be applied to determine the distance between the robotic arm and the target object.

[0190] Multiple sets of experimental data, including ratios and corresponding distances, are obtained; a simulation model for predicting distances is obtained by fitting the experimental data with the calculation formula; the fitted simulation model is used to predict new experimental data and the error index is calculated; if the error index is greater than the preset value, the model is corrected and refitted to obtain a new simulation model.

[0191] First, the coking analysis robot experimentally measures different ratios and their corresponding actual distances, recording multiple sets of data pairs. Based on the collected experimental data and a pre-set calculation formula (such as a linear formula), a fitting algorithm is used to solve for the parameters in the formula, resulting in an initial simulation model that can predict distances based on the ratios. The ratios from new experimental data are then input into the fitted model to obtain predicted distance values. Error indices such as mean absolute error and root mean square error are used to calculate the deviation between the predicted value and the actual distance, evaluating the model's accuracy. If the error index exceeds a preset threshold, the source of the error is analyzed, such as an inappropriate formula form or inaccurate parameters. The calculation formula is adjusted or the parameter solving method is optimized, and the model is refitted using the original experimental data to obtain a new simulation model until the error meets the standard. This allows for the direct determination of the distance between the robotic arm and the target object based on the simulation model.

[0192] Step S18: Establish a three-dimensional coordinate system with the location of the robotic arm as the origin based on the planar coordinate system. Determine the second coordinate set of the target object in the three-dimensional coordinate system based on the first coordinate set and the distance.

[0193] Specifically, since the robotic arm moves in three-dimensional space, in order to accurately plan the arm's movements based on the position of the target object, a three-dimensional coordinate system is established based on a planar coordinate set. The x and y values ​​are determined according to the first coordinate system. The position of the gripping part of the robotic arm is taken as the starting point of the z-axis. The z value of the target object in the three-dimensional coordinate system is determined according to the distance, and then the second coordinate set of the target object in the three-dimensional coordinate system is determined.

[0194] Furthermore, the robotic arm accurately locates the target object based on the second coordinate set, and assigns a stroke to the operation based on the location of the target object, such as the orientation and distance when performing operations like picking up, putting down, and pulling.

[0195] Furthermore, to ensure that each testing step is completed correctly and to facilitate the smooth progress of the testing process, the testing robot performs the following steps (steps S31 to S36) before executing each testing step:

[0196] Step S31: Determine whether the current test step is the first test step; if the current test step is not the first test step, then execute steps S32 to S34.

[0197] Specifically, if the current testing step is the first testing step, no correction is required. If it is not the first testing step, the results of the previous testing step need to be analyzed to determine whether they meet the standards. If they do not meet the standards, adjustments need to be made in a timely manner to avoid affecting the operation of subsequent testing steps.

[0198] Step S32: Obtain the first image information of the target object, including the information from the previous testing step.

[0199] Specifically, after completing the previous testing step, the coking testing robot takes a first image of the target object as it leaves the endpoint.

[0200] Step S33: Apply the target detection model to identify the actual position of the target object in the first image information and determine the distance between the actual position and the preset standard position.

[0201] Specifically, object detection models, such as YOLO and Faster R-CNN, can identify the target object and its corresponding bounding box coordinates in the first image information, i.e., the actual position of the target object. Further, the coking and testing robot takes a second image of the target object in a standard position from the same endpoint. Then, based on the object detection model, it identifies the target object and its corresponding bounding box coordinates in the second image information, i.e., the standard position of the target object. The actual position is then compared with the standard position, and the distance is calculated using the same corner point or center point as a reference.

[0202] Step S34: Determine the completion rate of the previous test step based on the distance.

[0203] Specifically, the electronic device presets the operation completion rate corresponding to each distance. As the distance increases, the operation completion rate decreases. For example, the operation completion rate is 100% at a distance of 0 cm, 90% at a distance of 1 cm, and 80% at a distance of 2 cm.

[0204] Step S35: If the operation completion rate is lower than the preset value, generate correction operation information based on the actual position and standard position of the target object.

[0205] Specifically, electronic devices have preset values ​​for the lowest possible completion rate. When the completion rate is greater than or equal to the preset value, it has virtually no impact on subsequent testing steps. For example, the preset value can be set to 80%, meaning the actual position of the target object can deviate from the standard position by a maximum of 2cm.

[0206] When the coking analysis robot determines that the completion rate of the operation is lower than the preset value, it generates correction operation information based on the actual and standard positions of the target object. The correction operation information includes the target object, the operation action, and the start and end points of the action. Generally, the operation action is consistent with the operation action in the previous analysis step, with the start point being the actual position of the target object and the end point being the standard position of the target object. The coking analysis robot generates travel instructions for its robotic arm based on the correction operation information and then completes the correction operation according to these instructions.

[0207] Step S36: The robotic arm executes the correction operation information to bring the target object to the standard position.

[0208] To better implement the above method, this application also provides an automated testing device for coking analysis robots, referring to... Figure 4 The 300-unit automated testing robot for coking processes includes:

[0209] The test instruction acquisition module 301 is used to acquire test instructions for the sample to be tested;

[0210] The test procedure retrieval module 302 is used to retrieve test procedures according to test instructions;

[0211] The laboratory step analysis module 303 is used to determine the target object, operation action, start point and end point of the operation action of the robotic arm in each laboratory step.

[0212] The testing step execution module 304 is used to execute the testing steps. When the robotic arm reaches the starting point, it acquires an image containing the target object through the camera, analyzes and identifies the image to locate the target object, and enables the robotic arm to perform corresponding operation actions on the target object and move to the endpoint until all testing steps are completed and the test results of the sample to be tested are obtained.

[0213] When analyzing and identifying images to locate target objects, the laboratory procedure execution module 304 is specifically used for:

[0214] Identify the information codes in an image and obtain the target object features contained in the information codes;

[0215] Retrieve the CNN neural network model used to identify the target object's features;

[0216] Image recognition based on CNN neural network model to identify target objects in images;

[0217] Establish a planar coordinate system based on the image and determine the first coordinate set corresponding to the target object;

[0218] Determine the first area of ​​the target object in the image based on the first coordinate set;

[0219] Calculate the ratio of the first area to the actual area of ​​the target object;

[0220] The distance between the robotic arm and the target object is determined based on the ratio;

[0221] A three-dimensional coordinate system is established based on the planar coordinate system, with the location of the robotic arm as the origin. The second coordinate set of the target object in the three-dimensional coordinate system is determined based on the first coordinate set and the distance.

[0222] Furthermore, the coking laboratory testing robot automated testing device 300 also includes:

[0223] The perspective distortion judgment module is used to determine whether there is perspective distortion in the planar graphic of the target object in the image based on the first coordinate set;

[0224] The distortion correction module is used to obtain lens distortion parameters through camera calibration when the condition is met, and then perform distortion correction on the image to obtain an ideal projection image without distortion.

[0225] The key feature point recognition module is used to identify key feature points of target objects in an image and determine the coordinates of the key feature points;

[0226] The angle calculation module is used to apply the perspective transformation matrix, combine the coordinates of key feature points and the actual size of the target object, establish the correspondence between image coordinates and three-dimensional coordinates, and calculate the camera's pitch angle and yaw angle.

[0227] The adjustment module is used to adjust the camera angle according to the pitch and yaw angles so that the camera can shoot the target object in parallel, obtain a new image with the target object, and repeatedly determine the first coordinate set corresponding to the target object.

[0228] Furthermore, the coking laboratory testing robot automated testing device 300 also includes:

[0229] The experimental data acquisition module is used to acquire multiple sets of experimental data, including ratios and corresponding distances.

[0230] The simulation model fitting module is used to fit experimental data and calculation formulas to obtain a simulation model for predicting distances.

[0231] The error calculation module is used to predict new experimental data by applying the fitted simulation model and calculate error indicators.

[0232] The correction module is used to correct the model and refit it to obtain a new simulation model when the error index is greater than the preset value.

[0233] Furthermore, the coking laboratory testing robot automated testing device 300 also includes:

[0234] The pending target object determination module is used to determine the pending target objects in the image;

[0235] The step determination module is used to determine whether the current testing step is to retrieve the sample to be tested from the sample rack.

[0236] If so, then assign a number to each undetermined target object;

[0237] The first target object determination module is used to determine the target object with the smallest number as the target object if the target object is determined from the undetermined target objects for the first time.

[0238] The second target object determination module is used to determine the target object corresponding to the number with the smallest difference from the previous number in ascending order of the numbers if the target object is not determined from the undetermined target objects for the first time.

[0239] The third coordinate set determination module is used to establish a planar coordinate system based on the image and determine the third coordinate set of each target object to be determined when the determination is negative.

[0240] The coordinate range determination module is used to obtain the historical coordinates of the target object in each historical image and determine the coordinate range of the target object based on the historical coordinates.

[0241] The similarity determination module is used to determine the similarity between the target object to be determined and the target object based on the degree of overlap between each third coordinate set and the coordinate range.

[0242] The background object determination module is used to determine the background objects in the image that are adjacent to each target object to be determined.

[0243] The correlation determination module is used to determine the correlation between the target object and each background object.

[0244] The third target object determination module is used to calculate and determine the evaluation score according to the similarity and relevance according to the preset weight, and determine the pending target object with the highest evaluation score as the target object.

[0245] The relevance determination module is specifically used for:

[0246] If there are multiple background objects, a classification model is applied to analyze the background objects and determine whether they are related to the current testing step.

[0247] If they are not related, the correlation between the target object and the background object will not be calculated.

[0248] If relevant, determine the positional relationship and distance between each target object and the background object;

[0249] Determine the correlation index between each target object and the background object based on their positional relationship;

[0250] Sort the target objects in order of distance from nearest to farthest to obtain the first sequence, and assign multipliers to each target object in the order of the first sequence.

[0251] The primary correlation between the target object and the background object is calculated by multiplying each correlation index by its corresponding multiplier.

[0252] The mean of the primary correlation between the target object and each background object is calculated to obtain the correlation score.

[0253] Furthermore, the automated testing robot 200 for coking tests also includes:

[0254] The first image information acquisition module is used to acquire first image information of the target object including the previous test step if the current test step is not the first test step.

[0255] The distance determination module is used to apply the target detection model to identify the actual position of the target object in the first image information and determine the distance between the actual position and the preset standard position.

[0256] The operation completion determination module is used to determine the operation completion rate of the previous test step based on the distance.

[0257] The correction operation generation module is used to generate correction operation information based on the actual position and standard position of the target object if the operation completion rate is lower than the preset value.

[0258] The correction operation execution module is used to enable the robotic arm to execute correction operation information so that the target object reaches the standard position.

[0259] The various variations and specific examples of the methods in the foregoing embodiments are also applicable to the automated coking robot testing device of this embodiment. Through the foregoing detailed description of the automated coking robot testing method, those skilled in the art can clearly understand the implementation method of the automated coking robot testing device of this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0260] This application also provides a testing system based on a coking testing robot, referring to... Figure 5 The testing system 400 based on the coking testing robot includes:

[0261] Coal Sample Management Module 401: Supervises the coal sample preparation process and enters the numbering information generated on the finished coal sample packaging according to preset rules.

[0262] Before testing, workers need to prepare the raw coal sample. The preparation process includes steps such as crushing, mixing, reducing and drying, gradually preparing the final analytical coal sample that can represent the characteristics of the original coal sample and meet the testing requirements. This process maintains the representativeness of the coal sample and avoids component segregation or loss.

[0263] To ensure the impartiality and accuracy of the testing, all coal samples were packaged in the same way, and each coal sample was assigned a unique code according to the preset coding rules.

[0264] The common coding structure includes five core information categories, and the order or number of fields can be adjusted according to enterprise needs. These include coal sample type code, sampling date code, sampling location or source code, coal type code, and serial number. Each field can be displayed consecutively or separated by special symbols. After workers enter the relevant data into the system, the coal sample management module 401 automatically generates codes according to preset rules.

[0265] When laboratory personnel receive coal samples, they scan the code and enter it into the testing system. The system automatically retrieves the basic information of the coal sample, and the test results are directly linked to the code and stored. Through the code, the entire process information of the coal sample, such as sampling records, sample preparation process, test results, and review opinions, can be queried in the system, realizing "one code to check everything".

[0266] Furthermore, the workers' operations during sample preparation are crucial to the accuracy of subsequent testing and are related to the fairness and accuracy of the testing. Therefore, the coal sample management module 401 supervises the coal sample preparation process by: acquiring video information of the coal sample preparation site; inputting the video into a target detection model to identify and mark core items, which include one or more of workers, equipment, tools, and installation equipment; applying a behavior recognition model to identify the marked video to obtain the worker's continuous actions and equipment status; comparing the continuous actions and equipment status with preset rules to determine whether the worker's operation is standardized; if not, generating an alarm message.

[0267] For example, to determine whether the sample preparation worker is wearing a safety helmet and dust mask, and to identify whether the worker's sample pouring actions comply with regulations, the video stream is input into the target detection model. The target detection model identifies the sample preparation worker, safety helmet, dust mask, equipment such as a separator, and the coal sample shovel. Furthermore, the behavior recognition model performs time-series analysis on the subsequent 10 seconds of video, capturing the continuous actions and equipment status of the sample preparation worker. It determines that the worker is wearing a safety helmet and dust mask. The model tracks the movement trajectory of the coal sample shovel and finds that the worker tilts the shovel to the left side of the separator when pouring the sample, causing the coal sample to mainly flow into the left funnel, with the coal sample volume in the right funnel being less than 40% of the normal proportion. This is judged as "uneven pouring". The model also identifies that the worker touches the inside funnel of the separator with their left hand to correct the coal sample flow (e.g., the contact time between the hand and the coal sample is ≥2 seconds), violating the rule that the worker must not touch the coal sample with their hands. Upon detecting a violation, the system triggers an alarm, such as activating an on-site audible and visual alarm, displaying the violation information on a screen in the sample preparation workshop, and automatically capturing video clips of the violation, associating them with the current coal sample code, and uploading them to the violation record module. The system also pushes alarm information, including the violation type and a link to the on-site video, to the supervisor's mobile app. Supervisors can remotely view the video and issue commands such as "pause sample preparation, clean the separator, and retry."

[0268] Coal sample testing module 402: Enables the testing robot to execute the automated testing method of the coking testing robot to complete the coal sample testing.

[0269] Test Result Acquisition Module 403: Collects test results from various testing devices based on the communication interface.

[0270] Specifically, the testing equipment used in the testing process is automatically connected to the system. After the testing equipment obtains the test results, the system collects and summarizes the scattered equipment data, improving the speed and accuracy of data entry.

[0271] Laboratory result analysis module 404: compares the laboratory results with standardized test values, identifies abnormal data, and marks them.

[0272] Specifically, the test result analysis module 404 compares the collected test results with the standardized test values ​​in the system, such as industry standard values ​​and historical normal range values, to determine whether the data is within a reasonable range; if it exceeds the range, it is automatically marked as abnormal data for subsequent key review.

[0273] Laboratory Result Review Module 405: Sends laboratory results and abnormal data to the administrator's account and obtains the administrator's review results for the laboratory results and abnormal data.

[0274] Specifically, the system initiates a results review process and collects managerial decisions to ensure traceability of results and clear accountability. Normal test results are sent to the manager's account along with marked abnormal data, allowing the manager to review the accuracy and validity of the results. The system simultaneously records the manager's review comments, creating a review record if the results are approved or rejected.

[0275] Laboratory review module 406: Responds to the review instructions issued by the administrator, generates the review process, obtains the review results, and sends the review results to the administrator's account for review.

[0276] Specifically, when a manager issues an instruction requiring review, a standardized review process is automatically generated, such as steps for resampling and retesting. After the review is completed, new review results are collected and sent to the manager's account for a second review until the results are confirmed.

[0277] Data statistics module 407: Generates statistical tables based on test results.

[0278] Specifically, the data statistics module 407 converts the test results into structured statistical tables, facilitating data viewing and analysis. Based on all the test results stored in the system, it automatically generates statistical tables according to preset dimensions (such as time, coal sample type, test items, etc.), such as "Monthly Coal Sample Calorific Value Statistics Table" and "Comparison Table of Sulfur Content of Coal Samples in Each Batch," providing data support for subsequent production decisions and quality analysis.

[0279] Therefore, the testing system based on a coking testing robot in this application improves accuracy through automation and standardization. The testing robot strictly follows preset programs to perform weighing, heating, and other operations, avoiding human visual errors and endpoint judgment deviations. Furthermore, it automatically reads data in conjunction with high-precision equipment, eliminating transcription errors and ensuring controllable errors in key indicators such as ash and sulfur content. Second, the system improves efficiency by eliminating manual preparation and waiting, shortening the testing cycle for single samples. Third, the system ensures safety by automatically completing hazardous operations such as high-temperature heating and reagent addition, avoiding personnel contact with high-temperature equipment or toxic reagents and reducing the risk of workplace injuries. Fourth, the system optimizes management by automatically uploading data to the system and linking it with coal sample codes to form a traceability chain, facilitating quality traceability and data analysis and reducing errors from manual recording.

[0280] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

[0281] Additionally, it should be understood that relational terms such as "first" and "second" are used merely 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. 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.

Claims

1. An automated testing method using a coking robot, characterized in that, The coking testing robot includes a robotic arm and a camera mounted on the robotic arm, and the method includes: Obtain testing instructions for the sample to be tested; Retrieve the test steps according to the test instructions; Determine the target object, operation action, start point and end point of the operation action for each of the aforementioned testing steps; During the testing process, the robotic arm reaches the starting point, acquires an image containing the target object through the camera, analyzes and identifies the image to locate the target object, and then performs corresponding operations on the target object and moves to the endpoint until all testing steps are completed, thus obtaining the test results of the sample to be tested. The analysis and identification of the image to locate the target object includes: Identify the information code in the image and obtain the target object features contained in the information code; Retrieve the CNN neural network model used to identify the target object corresponding to the features of the target object; The image is identified based on the CNN neural network model to determine the target object in the image; Establish a planar coordinate system based on the image and determine the first coordinate set corresponding to the target object; Determine the first area of ​​the target object in the image based on the first coordinate set; Calculate the ratio of the first area to the actual area of ​​the target object; The distance between the robotic arm and the target object is determined based on the ratio. A three-dimensional coordinate system is established based on the planar coordinate system with the location of the robotic arm as the origin. Based on the first coordinate set and the distance, the second coordinate set of the target object in the three-dimensional coordinate system is determined. If the image is identified based on the CNN neural network model, and it is determined that at least two target objects exist in the image, then the method further includes: Identify the target object in the image; determine if the current testing step involves retrieving the sample from the sample rack. If so, then assign a number to each of the aforementioned target objects to be determined; If a target object is determined from the undetermined target objects for the first time, then the undetermined target object with the smallest number is determined as the target object; If the target object is not determined from the list of undetermined target objects for the first time, then the undetermined target object corresponding to the number with the smallest difference from the previous number is determined as the target object in ascending order of number. Otherwise, a planar coordinate system is established based on the image to determine the third coordinate set of each of the target objects to be determined; Obtain the historical coordinates of the target object in each historical image, and determine the coordinate range of the target object based on each historical coordinate; The similarity between the target object to be determined and the target object is determined based on the degree of overlap between each of the third coordinate sets and the coordinate range. Identify the background objects in the image that are adjacent to each of the target objects to be determined; Determine the correlation between the target object to be determined and each of the background objects; An evaluation score is calculated based on the similarity and relevance according to a preset weight, and the candidate object with the highest evaluation score is determined as the target object.

2. The method according to claim 1, characterized in that, Before determining the first area of ​​the target object in the image based on the first coordinate set, the method includes: Determine whether the planar shape of the target object in the image has perspective distortion based on the first coordinate set; If so, the lens distortion parameters are obtained through camera calibration, and the image is corrected for distortion to obtain an ideal projection image without distortion. Identify key feature points of a target object in an image and determine the coordinates of the key feature points; By applying the perspective transformation matrix and combining the coordinates of key feature points with the actual size of the target object, the correspondence between image coordinates and three-dimensional coordinates is established, and the pitch and yaw angles of the camera are calculated. The camera angle is adjusted according to the pitch angle and the yaw angle so that the camera can shoot the target object in parallel, thereby obtaining a new image with the target object, and the first coordinate set corresponding to the target object is determined repeatedly.

3. The method according to claim 1, characterized in that, Before determining the distance between the robotic arm and the target object based on the ratio, the method further includes: Obtain multiple sets of experimental data, including ratios and corresponding distances; By fitting the experimental data with the calculation formula, a simulation model for predicting distances is obtained. The fitted simulation model is used to predict new experimental data, and error indices are calculated. If the error index is greater than the preset value, the model is corrected and refitted to obtain a new simulation model.

4. The method according to claim 1, characterized in that, Determining the correlation between the target object and the background object includes: If there are multiple background objects, a classification model is applied to analyze the background objects and determine whether they are related to the current testing step. If they are not related, the correlation between the target object to be determined and the background object will not be calculated. If relevant, determine the positional relationship and distance between each target object to be determined and the background object; The correlation index between each of the target objects to be determined and the background objects is determined according to the positional relationship; The objects to be determined are sorted according to their distance from closest to furthest to obtain a first sequence, and a multiplier is assigned to each object to be determined according to the order of the first sequence. By multiplying each of the aforementioned correlation indices by its corresponding multiplier, the primary correlation between the target object and the background object is calculated. The mean of the primary correlation between the target object and each background object is calculated to obtain the correlation score.

5. The method according to claim 1, characterized in that, Before performing the aforementioned testing steps, the method further includes: If the current testing step is not the first testing step, then obtain the first image information of the target object including the previous testing step; The target detection model is applied to identify the actual position of the target object in the first image information, and the distance between the actual position and the preset standard position is determined; The degree of completion of the previous test step is determined based on the distance. If the completion rate of the operation is lower than the preset value, correction operation information is generated based on the actual position and standard position of the target object; The robotic arm executes the corrective operation information to bring the target object to the standard position.

6. An automated testing robot for coking analysis, characterized in that, include: The test instruction acquisition module is used to acquire test instructions for the samples to be tested. The test step retrieval module is used to retrieve the test steps according to the test instructions. The testing step analysis module is used to determine the target object, operation action, start point and end point of the operation action of the robotic arm in each testing step; The testing step execution module is used to execute the testing steps. The robotic arm reaches the starting point, acquires an image containing the target object through a camera, analyzes and identifies the image to locate the target object, and causes the robotic arm to perform corresponding operation actions on the target object and move to the endpoint until all testing steps are completed and the test results of the sample to be tested are obtained. The testing step execution module analyzes and identifies the image to locate the target object, including: Identify the information code in the image and obtain the target object features contained in the information code; Retrieve the CNN neural network model used to identify the target object corresponding to the features of the target object; The image is identified based on the CNN neural network model to determine the target object in the image; Establish a planar coordinate system based on the image and determine the first coordinate set corresponding to the target object; Determine the first area of ​​the target object in the image based on the first coordinate set; Calculate the ratio of the first area to the actual area of ​​the target object; The distance between the robotic arm and the target object is determined based on the ratio. A three-dimensional coordinate system is established based on the planar coordinate system with the location of the robotic arm as the origin. Based on the first coordinate set and the distance, the second coordinate set of the target object in the three-dimensional coordinate system is determined. If the image is identified based on the CNN neural network model, and it is determined that there are at least two target objects in the image, then the method further includes: Identify the target object in the image; determine if the current testing step involves retrieving the sample from the sample rack. If so, then assign a number to each of the aforementioned target objects to be determined; If a target object is determined from the undetermined target objects for the first time, then the undetermined target object with the smallest number is determined as the target object; If the target object is not determined from the list of undetermined target objects for the first time, then the undetermined target object corresponding to the number with the smallest difference from the previous number is determined as the target object in ascending order of number. Otherwise, a planar coordinate system is established based on the image to determine the third coordinate set of each of the proposed target objects; Obtain the historical coordinates of the target object in each historical image, and determine the coordinate range of the target object based on each historical coordinate; The similarity between the target object to be determined and the target object is determined based on the degree of overlap between each of the third coordinate sets and the coordinate range. Identify the background objects in the image that are adjacent to each of the target objects to be determined; Determine the correlation between the target object to be determined and each of the background objects; An evaluation score is calculated based on the similarity and relevance according to a preset weight, and the candidate object with the highest evaluation score is determined as the target object.

7. A coking testing robot, characterized in that, Includes a body (1), a walking mechanism (2) is installed at the bottom of the body (1), a sample rack (3) is provided on the body (1), and the sample rack (3) contains the sample to be tested (4) and the sample after testing; a robotic arm (5) is also installed on the body (1), and a camera (6) is installed above the gripping part of the robotic arm (5). It also includes: at least one processor (101); the robotic arm (5) and the camera (6) are both connected to the processor (101); At least one memory (103); At least one computer program, wherein the at least one computer program is stored in the memory (103) and configured to be executed by the at least one processor (101), the at least one computer program being configured to: perform an automated testing method for coking analysis robots as described in any one of claims 1 to 5.

8. A testing system based on a coking testing robot, characterized in that, include: Coal Sample Management Module: Monitors the coal sample preparation process and enters the numbering information generated on the finished coal sample packaging according to preset rules; Coal sample analysis module: Enables the analysis robot to perform the method described in any one of claims 1-5 to complete the coal sample analysis; Test Result Acquisition Module: Collects test results from various testing devices based on the communication interface; Laboratory result analysis module: compares the laboratory results with standardized test values, identifies and marks abnormal data; Laboratory Result Review Module: Sends laboratory results and abnormal data to the administrator's account and obtains the administrator's review results for the laboratory results and abnormal data; Laboratory review module: Responds to the review instructions issued by the administrator, generates the review process, obtains the review results, and sends the review results to the administrator's account for review; Data statistics module: Generates statistical tables based on test results; The coal sample management module oversees the coal sample preparation process, including: Obtain video information of the coal sample preparation site; The video is input into the target detection model to identify and mark core items, which include one or more of workers, equipment, tools and installation equipment. By applying a behavior recognition model to identify tagged videos, the continuous actions of workers and the status of equipment can be obtained. The continuous actions and equipment status are compared with preset rules to determine whether the worker's operation is in accordance with regulations; if not, an alarm message is generated.

Images