Test tube sorting intelligent identification method and device based on multi-source sensors

Abstract

The present invention discloses a test tube sorting intelligent identification method and device based on multi-source sensors, and relates to the field of test tube sorting technology. The method includes: performing multi-source data collection on the test tube to be sorted through a multi-source sensor module to generate test tube data; performing test tube sample type matching, locating the first sorting point, and reading the first test tube sorting attribute information; performing a dual-path sample abnormality inspection in combination with the test tube data and the first test tube sorting attribute information, reading the first target position of the first sorting point, and generating a preset sorting scheme when the test tube to be sorted reaches the first target position; sorting the test tube to be sorted to the first test tube storage rack according to the preset sorting scheme. The present invention solves the technical problem of inaccurate sorting and low sorting efficiency in the test tube sorting process in the prior art, thereby affecting the quality of the sample, and achieves the technical effect of improving sorting accuracy and efficiency and ensuring the quality of the sample.

Description

Technical Field

[0001] The present invention relates to the technical field of test tube sorting, and in particular to a test tube sorting intelligent identification method and device based on multi-source sensors. Background Art

[0002] In many fields such as medical treatment and scientific research, accurate and efficient sorting of a large number of test tubes is a crucial work link. The test tube sorting technology used in the past has many limitations. Traditional sorting methods mostly rely on manual visual recognition and manual classification, which is not only labor-intensive, but also easily affected by human subjective factors, resulting in a high sorting error rate. At the same time, manual sorting is inefficient and difficult to cope with large-scale test tube sorting tasks. In terms of technical means, some previous automated sorting systems had relatively simple functions and could only be judged based on a few simple features, such as relying only on the appearance and shape of the test tube or some information on the label. But in fact, many factors such as the sample type, weight, and state of the test tube have an important impact on the accuracy and rationality of sorting.

[0003] The prior art has technical problems such as inaccurate sorting and low sorting efficiency during the test tube sorting process, thereby affecting the quality of the samples. Summary of the invention

[0004] The present application provides a test tube sorting intelligent identification method and device based on multi-source sensors, which is used to solve the technical problems in the prior art of inaccurate sorting and low sorting efficiency during test tube sorting, thereby affecting the quality of samples.

[0005] In view of the above problems, the present application provides a test tube sorting intelligent identification method and device based on multi-source sensors.

[0006] The first aspect of the present application provides a test tube sorting intelligent identification method based on a multi-source sensor, the method comprising:

[0007] Multi-source data is collected for the test tubes to be sorted through the multi-source sensor module on the test tube sorting line to generate test tube label analysis data, test tube image data and test tube weight data; test tube sample type matching is performed based on the test tube label analysis data, the first sorting point is located, and the first test tube sorting attribute information is read, wherein the first test tube sorting attribute information includes sample type and sample weight; a two-way sample abnormality check is performed in combination with the test tube image data, the test tube weight data and the first test tube sorting attribute information; when the two-way sample abnormality check passes, the first target position of the first sorting point is read, and when the test tube to be sorted reaches the first target position, a test tube sorting scheme decision is made based on the test tube image data to generate a preset sorting scheme; and the first sorting robot is controlled to sort the test tube to be sorted to the first test tube storage rack according to the preset sorting scheme.

[0008] The second aspect of the present application provides a test tube sorting intelligent identification device based on a multi-source sensor, the device comprising:

[0009] A multi-source data acquisition module, the multi-source data acquisition module is used to perform multi-source data acquisition on the test tubes to be sorted through the multi-source sensor module on the test tube sorting line, and generate test tube label analysis data, test tube image data and test tube weight data; a first test tube sorting attribute information reading module, the first test tube sorting attribute information reading module performs test tube sample type matching based on the test tube label analysis data, locates the first sorting point, and reads the first test tube sorting attribute information, wherein the first test tube sorting attribute information includes the sample type and the sample weight; a dual-path sample abnormality inspection module, the dual-path sample abnormality inspection module is used The invention relates to a method for performing a two-way sample abnormality inspection in combination with the test tube image data, the test tube weight data and the first test tube sorting attribute information; a preset sorting scheme generating module, wherein the preset sorting scheme generating module is used to read the first target position of the first sorting point when the two-way sample abnormality inspection passes, and when the test tube to be sorted reaches the first target position, make a test tube sorting scheme decision based on the test tube image data to generate a preset sorting scheme; and a test tube sorting module, wherein the test tube sorting module is used to control the first sorting robot arm to sort the test tube to be sorted to the first test tube storage rack according to the preset sorting scheme.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] Through the multi-source sensor module on the test tube sorting line, multi-source data is collected for the test tube to be sorted, and test tube label analysis data, test tube image data and test tube weight data are generated; based on the test tube label analysis data, the test tube sample type is matched, the first sorting point is located, and the first test tube sorting attribute information is read; the double-path sample abnormality inspection is performed in combination with the test tube image data, the test tube weight data and the first test tube sorting attribute information; when the double-path sample abnormality inspection passes, the first target position of the first sorting point is read, and when the test tube to be sorted reaches the first target position, a preset sorting scheme is generated; and the first sorting mechanical arm is controlled to sort the test tube to be sorted to the first test tube storage rack according to the preset sorting scheme. The technical effect of improving sorting accuracy and efficiency and ensuring the quality of samples is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0013] Figure 1 A schematic diagram of the process flow of the test tube sorting intelligent identification method based on multi-source sensors provided in an embodiment of the present application;

[0014] Figure 2 A schematic diagram of the structure of a test tube sorting intelligent identification device based on multi-source sensors provided in an embodiment of the present application.

[0015] Explanation of the accompanying drawings: multi-source data acquisition module 10, first test tube sorting attribute information reading module 20, dual-path sample abnormality inspection module 30, preset sorting scheme generation module 40, sorting test tube module 50. DETAILED DESCRIPTION

[0016] The present application provides a test tube sorting intelligent identification method and device based on multi-source sensors, which is used to solve the technical problems in the prior art of inaccurate sorting and low sorting efficiency during the test tube sorting process, thereby affecting the quality of samples.

[0017] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0018] Embodiment 1

[0019] like Figure 1 As shown, the present application provides a test tube sorting intelligent identification method based on multi-source sensors, the method comprising:

[0020] Step S100: multi-source data collection is performed on the test tubes to be sorted by the multi-source sensor module on the test tube sorting line to generate test tube label analysis data, test tube image data and test tube weight data.

[0021] Specifically, the test tube sorting line is equipped with a multi-source sensor module, which can collect all-round data on the test tubes to be sorted. For the test tube labels, optical recognition technology or scanning technology is used to parse them, obtain relevant information on the labels, and generate test tube label parsing data. This part of the data contains important identifications such as the number, type, and collection time of the samples in the test tube. At the same time, image acquisition equipment, such as high-definition cameras, are used to capture the appearance of the test tubes, obtain the shape, size, color, and damage of the test tubes, and form test tube image data. In addition, a high-precision weighing sensor is used to accurately measure the weight of the test tubes and obtain the test tube weight data. The collection of these multi-source data provides rich, comprehensive, and accurate basic information for subsequent test tube sorting work.

[0022] Step S200: performing test tube sample type matching based on the test tube label parsing data, locating a first sorting point, and reading first test tube sorting attribute information, wherein the first test tube sorting attribute information includes sample type and sample weight.

[0023] Specifically, the parsed data will be matched with the preset sample type database, and the specific type of the current test tube sample will be accurately determined by comparing the key information in the label, such as the sample number, type code, etc. The first sorting point is located according to the sample type. Different sample types may correspond to different sorting positions to achieve classified storage and management. While locating the sorting point, the first test tube sorting attribute information related to the sample will also be read, among which the key attributes include sample type (such as blood, urine, tissue fluid, etc.) and sample weight. Through such processing, the specific ownership and related attributes of each test tube to be sorted can be clarified, so as to be fully prepared for subsequent inspection and sorting decisions.

[0024] Step S300: performing a dual-path sample abnormality inspection in combination with the test tube image data, the test tube weight data and the first test tube sorting attribute information.

[0025] Specifically, from the perspective of the test tube image data, the collected images are analyzed in detail to check whether the appearance of the test tube is damaged, deformed, stained, or other abnormalities. For example, if cracks are found in the test tube, it means that the sample is at risk of leakage; if the color of the test tube shown in the image is abnormal, it suggests that the sample is contaminated. At the same time, for the test tube weight data, it is compared with the expected normal weight range. If the actual weight deviates significantly from the normal range set according to the sample type and the first test tube sorting attribute information, this indicates that there is a problem with the sample size, such as too little is insufficient collection, and too much is mixed with other substances. The results of the image analysis and the results of the weight comparison are combined with each other. If the image shows that the appearance of the test tube is normal, but the weight is abnormal, or vice versa, further in-depth inspection and judgment are required. Through this dual-path combined abnormality detection method, various abnormalities in the test tube samples can be discovered more comprehensively and accurately, ensuring the accuracy and reliability of subsequent processing and analysis.

[0026] Step S400: When the two-way sample abnormality inspection passes, the first target position of the first sorting point is read, and when the test tube to be sorted reaches the first target position, a test tube sorting scheme decision is made based on the test tube image data to generate a preset sorting scheme.

[0027] Specifically, only when the previous two-way sample abnormality inspection passes smoothly will the subsequent operation be started. First, the first target position corresponding to the first sorting point that has been determined will be read. This target position is pre-set according to the previous sample type matching and sorting rules, and is the position where the test tube should be accurately placed in the end. When the test tube to be sorted arrives at the first target position with the transmission device, the key sorting scheme decision is made based on the collected test tube image data. By re-analyzing the image data, the specific shape, size, placement direction and other factors of the test tube are considered, combined with the spatial layout and restrictions of the target position. For example, if the target position is a narrow grid and the shape of the test tube is long, a specific grasping angle and placement method will be decided to ensure that the test tube can be placed smoothly and safely. Based on the above analysis and considerations, a preset sorting scheme will be generated. This scheme will specify the specific operation details such as the action process, grasping force, and moving path of the sorting robot arm in detail, so as to accurately place the test tube to be sorted at the first target position, improve the sorting accuracy, and ensure the quality of the sample.

[0028] Step S500: controlling the first sorting robot arm to sort the test tubes to be sorted to the first test tube storage rack according to a preset sorting scheme.

[0029] Specifically, the first sorting robot arm will receive precise instructions, including the position, angle, and strength of grabbing the test tube to be sorted, as well as detailed information such as the moving path, speed, and location of the test tube. The robot arm will execute these instructions with a high degree of accuracy and stability to ensure that the test tube to be sorted is placed smoothly, safely, and accurately in the designated position of the first test tube storage rack. For example, the robot arm uses a specific clamp to grab the neck of the test tube with appropriate force, and then moves along a preset collision-free path, and finally gently places the test tube into the designated grid of the storage rack. Through such precise control, efficient and reliable test tube sorting and storage can be achieved, reducing human errors and operational risks, and improving the efficiency and quality of overall work.

[0030] In a possible implementation, step S200 further includes:

[0031] Step S210: Acquire a plurality of sorting points distributed on the test tube sorting line, wherein the plurality of sorting points have sorting sample type identifiers and preset sample weight identifiers.

[0032] Step S220: performing point matching based on the test tube label parsing data and the sorting sample type identifier to generate the first sorting point and the first test tube sorting attribute information.

[0033] Specifically, it actively obtains detailed information about multiple sorting points on the test tube sorting line. These sorting points are given specific identification during design and layout. Among them, the sorting sample type identification clearly indicates the specific type of samples that can be processed at each point. For example, some points are marked for processing blood samples, some are marked for processing urine samples, and some are marked for processing tissue sections, etc. This classification identification allows different types of samples to be accurately guided to the corresponding points. The preset sample weight identification sets the range of sample weights that each sorting point can bear or handle, which helps to ensure that each point will not encounter problems when processing samples due to weight exceeding its carrying capacity, and it is also convenient to reasonably allocate points according to the weight of the samples in the subsequent sorting process.

[0034] The obtained test tube label parsing data is compared and matched with the sorting sample type identification of each sorting point, and the detailed type information of the sample in the current test tube to be sorted is obtained from the test tube label parsing data, and then this type information is compared with the sorting sample type identification of each sorting point one by one. If the type of the sample in the test tube is consistent with the identification of a certain sorting point, then this point will be determined as the first sorting point. At the same time, combined with the preset sample weight identification of the point and the information about the actual weight of the sample in the test tube label parsing data, the first test tube sorting attribute information is generated. Exemplarily, the label parsing data of a test tube indicates that it is a 5 ml plasma sample, and the sorting sample type identification of a certain sorting point is a plasma sample, and the preset sample weight identification is "3-10 ml", then this point will become the first sorting point, and the generated first test tube sorting attribute information will include the sample type of "plasma sample" and the sample weight of "5 ml". Through precise point matching, each test tube is ensured to be assigned to the sorting point that is most suitable for processing its sample type and weight, greatly reducing the possibility of incorrect sorting.

[0035] In a possible implementation, step S300 further includes:

[0036] Step S310: Preset the test tube image data configuration based on the sample type in the first test tube sorting attribute information, and establish a sample type matching abnormality inspection branch.

[0037] Step S320: establishing a sample weight matching abnormality inspection branch based on the sample weight in the first test tube sorting attribute information.

[0038] Step S330: Establish a dual-path sample anomaly inspection module with the sample type matching anomaly inspection branch and the sample weight matching anomaly inspection branch.

[0039] Step S340: performing a dual-path sample abnormality inspection on the test tube image data and the test tube weight data through the dual-path sample abnormality inspection module.

[0040] Specifically, the configuration of the preset test tube image data is performed based on the sample type contained in the first test tube sorting attribute information, thereby creating a sample type matching abnormality inspection branch. This means that the expected image features will be set in advance for different sample types, for example, a certain blood sample should have a specific color and texture, and a certain tissue fluid should have a specific transparency and morphology. Then, when the image data of the test tube to be inspected is obtained, it is compared with the preset image features of the corresponding sample type. If the actual image does not match the preset features, it indicates that there is an abnormal situation of wrong labeling (that is, the sample type on the label does not match the actual one) or sample mixing (that is, the actual sample in the test tube is not the one it should be). For example, if a test tube is labeled as a specific type of serum sample, but the collected image shows that its color and texture are more like plasma, then there is a problem of wrong labeling or sample mixing.

[0041] The sample weight matching abnormality inspection branch is constructed based on the sample weight in the first test tube sorting attribute information, and the normal weight range corresponding to each sample type is first determined. For example, a blood sample should usually be 3-5 grams, a urine sample should generally be 10-20 ml, etc. After obtaining the actual weight data of the test tube to be inspected, it is compared with the preset normal weight range of the sample type. If the actual weight is not within the preset range, it indicates that an abnormal situation of incorrect sample weight has occurred, such as an error in sample filling, insufficient or excessive filling.

[0042] The previously constructed sample type matching abnormality inspection branch and sample weight matching abnormality inspection branch are integrated to establish a dual-path sample abnormality inspection module. By combining these two branches at the same time, it is possible to comprehensively detect whether the sample is abnormal from two important aspects: the type characteristics and weight of the sample. The sample type matching abnormality inspection branch focuses on judging whether the appearance characteristics of the sample are consistent with its expected type, while the sample weight matching abnormality inspection branch focuses on confirming whether the weight of the sample is within a reasonable range. For example, assuming that a test tube is labeled as a chemical reagent solution of a specific concentration, the sample type matching abnormality inspection branch will check whether the color, transparency, etc. of the solution are consistent with the characteristics of the reagent of this type, while the sample weight matching abnormality inspection branch will verify whether the weight of the solution is within the expected range. Such a dual-path inspection module can more effectively discover existing problems and improve the accuracy and reliability of the inspection. For example, when the sample type looks correct but the weight is abnormal, or when the weight is normal but the type characteristics do not meet the requirements, it can be detected in time, thereby preventing the wrong sample from entering the subsequent processing flow.

[0043] Using the previously constructed dual-channel sample abnormality inspection module, dual-channel abnormality inspection is carried out for the obtained test tube image data and test tube weight data at the same time. First, the dual-channel sample abnormality inspection module will compare the test tube image data with the preset sample type characteristics. For example, if it is a blood sample, it will check whether the blood color, clarity, presence of clots, etc. in the image meet the characteristics of normal blood. At the same time, the module will also compare the test tube weight data with the pre-set reasonable weight range of the sample type. For example, for a urine sample of a specific specification, confirm whether its weight is between the expected upper and lower limits. Through the simultaneous inspection of these two aspects, if the sample type characteristics displayed by the image data do not match the preset, or the weight data exceeds the reasonable range, it can be judged that the sample is abnormal. This dual-channel abnormality inspection method can more comprehensively and accurately discover sample problems, improve the reliability and effectiveness of the inspection, minimize the missed detection or false detection caused by a single inspection method, and ensure the accuracy and scientificity of subsequent processing and analysis.

[0044] In a possible implementation, step S400 further includes:

[0045] Step S410: establishing a reference grabbing area of ​​the first sorting robot arm based on the sample type in the first test tube sorting attribute information.

[0046] Step S420: using the reference grabbing area as an analysis reference point, and locating the initial placement state of the test tube based on the test tube image data.

[0047] Step S430: Based on the initial placement state of the test tube, predict the change of the test tube placement state of the first target position to generate a first predicted placement state.

[0048] Step S440: performing sorting path planning of the first sorting robot arm based on the first predicted placement state to generate the preset sorting plan.

[0049] Specifically, the reference grabbing area is established for the first sorting robot arm according to the sample type contained in the first test tube sorting attribute information. This is because the material, shape, size and characteristics of the test tubes of different sample types may be different. For example, if the sample type is a precious cell culture fluid in fragile glass material, the reference grabbing area will be set at the thicker position of the upper part of the test tube to reduce the pressure on the test tube during grabbing and reduce the risk of breakage. For another example, for a test tube made of plastic material, thicker and filled with conventional liquid, the reference grabbing area will be selected in the middle, which can not only ensure stable grabbing, but also improve operational efficiency. For another example, if the sample type is a small and irregularly shaped test tube, the reference grabbing area will be set near the opening to more accurately control the grabbing force and angle. By determining the reference grabbing area according to the sample type, the safety of the test tube during the grabbing process can be guaranteed to the greatest extent, while improving the accuracy and stability of the operation of the sorting robot arm.

[0050] The reference grab area is used as the key analysis reference point, and the initial placement state of the test tube is accurately determined based on the acquired test tube image data. Here, special attention is paid to some specific features of the test tube on the sorting line, such as the orientation of the test tube head. Through detailed analysis of the image, if the reference grab area is located in the middle of the test tube, and the image shows that the test tube head is facing the upper left, then this initial orientation feature can be clearly determined. Alternatively, the reference grab area is at the bottom of the test tube, and the image shows that the test tube head is vertically downward, but with a certain angle deviation. In addition to the head orientation, other information related to the initial placement state can also be obtained, such as the inclination angle of the test tube, whether there is any bending deformation, and whether the position on the sorting line is biased to the left or right. For example, if the reference grab area is at the top of the test tube, the image shows that the test tube head is facing right and tilted 30 degrees, which is a clear description of the initial placement state. This precise positioning of the initial placement state of the test tube provides accurate starting information for subsequent operations, which helps to plan and execute subsequent sorting actions more effectively.

[0051] Based on the initial placement state of the test tube obtained above, the first target position is predicted in combination with the decision tree algorithm in machine learning, and the possible changes in the placement state of the test tube from the initial position to the first target position are analyzed to generate the first predicted placement state. First, a large amount of training data is prepared for the decision tree algorithm. These data include test tube information of different initial placement states, such as initial position coordinates, angles, postures, and various conditions in the corresponding transmission process, such as transmission speed, vibration amplitude, friction, etc., as well as the actual placement state when it finally reaches the first target position. These data are preprocessed and feature extracted to convert the initial placement state and the conditions in the transmission process into feature vectors that the decision tree can understand and process. Then, the decision tree model is trained using these prepared data. The decision tree will branch and split according to the importance and threshold of the features, and gradually form a structure that can predict the placement state according to the input features. When predicting the placement state of a new test tube, first obtain its initial placement state and related transmission condition information, and convert this information into feature vectors and input them into the trained decision tree model. The decision tree will make judgments and inferences along its branches according to the values ​​of the input features. For example, if the initial angle is large, the decision tree will continue to judge the impact of the transmission speed along a certain branch; if the vibration amplitude is small, it will make subsequent judgments along another branch. Through such layer-by-layer reasoning and judgment, the predicted placement state of the test tube when it reaches the first target position is finally obtained, that is, the first predicted placement state is generated. By continuously optimizing and adjusting the decision tree model with new data, the accuracy and reliability of its predictions are improved, providing more valuable references for subsequent sorting operations. This kind of prediction can prepare for response in advance, adjust the action of the robot arm, or take additional measures to ensure that the test tube can eventually be accurately and stably placed at the first target position.

[0052] According to the first predicted placement state obtained above, the sorting path is planned for the first sorting robot arm, and then a preset sorting scheme is generated. Because the state of the test tube when it reaches the first target position is known, the action path of the robot arm can be planned in a targeted manner. For example, if the predicted placement state is that the test tube will have a certain degree of inclination, the sorting path will include appropriate adjustment actions after grasping to correct the inclination, and then plan a path to move straight to the target position. When planning the sorting path, the changes in the movement speed and acceleration of the robot arm should also be considered to ensure the smoothness and accuracy of the action. For example, for a test tube that is initially placed with an inclination and the target position space is narrow, the preset sorting scheme is: the robot arm first grasps the test tube at a slower speed, then quickly makes a small rotation adjustment to correct the inclination, and then moves to the target position at a moderate speed along a curved path that avoids surrounding obstacles, and finally slowly puts down the test tube. Through such a fine sorting path planning, the generated preset sorting scheme can maximize the sorting efficiency and accuracy while ensuring the safety of the test tube and the surrounding environment.

[0053] In a possible implementation, step S430 further includes:

[0054] Step S431: Acquire the test tube transmission characteristics and the sorting line transmission characteristics, wherein the test tube transmission characteristics are the outer surface roughness of the test tube to be sorted, and the sorting line transmission characteristics are the surface roughness and vibration characteristics of the test tube sorting line.

[0055] Step S432: using the test tube transmission characteristics and the sorting line transmission characteristics and combining the test tube weight data to collect test tube transmission samples, and establishing a test tube position vector prediction model for transmission distance.

[0056] Step S433: collecting data collection points of the test tube image data, and calculating a first transmission distance from the data collection points to the first target position.

[0057] Step S434: inputting the first transmission distance into the test tube position vector prediction model to generate the first predicted placement state.

[0058] Specifically, two key feature information are obtained, namely, the test tube transmission feature and the sorting line transmission feature. For the test tube transmission feature, the outer surface roughness of the test tube to be sorted is focused. If the outer surface roughness of the test tube is high, the friction between the test tube and the sorting line will be large during the transmission process, which will affect the stability of the test tube movement, and even cause jamming or irregular movement. The sorting line transmission feature includes the surface roughness and vibration characteristics of the sorting line. The surface roughness of the sorting line will directly affect the friction between the test tube and the test tube. If the surface is rough, it will increase resistance and change the transmission speed and direction of the test tube. The vibration feature is also very important. If there is obvious vibration in the sorting line, the test tube will jump and shake during the transmission process, thereby changing its initial placement state. For example, a test tube with low outer surface roughness is placed on a sorting line with high surface roughness and strong vibration, and a large change in position and posture will occur during the transmission process. For another example, if the outer surface of the test tube is rough, and the surface of the sorting line is smooth but vibrates frequently, the transmission state of the test tube will also be more complicated and difficult to accurately predict. The purpose of acquiring these characteristic information is to more accurately analyze and predict the placement status changes of the test tube during the transmission process.

[0059] Using the obtained test tube transmission features (the outer surface roughness of the test tube to be sorted), the sorting line transmission features (the surface roughness and vibration characteristics of the sorting line), and the weight data of the test tube, a large number of test tube transmission samples are collected. By combining these factors, a model that can predict the test tube position vector according to the transmission distance is constructed. The random forest algorithm is used to construct a test tube position vector prediction model for the transmission distance. First, a large number of labeled sample data are prepared. Each sample contains the test tube transmission features (the outer surface roughness of the test tube to be sorted), the sorting line transmission features (the surface roughness and vibration characteristics of the sorting line), the test tube weight data, the corresponding transmission distance and the final test tube position vector. For the construction of the random forest algorithm: a certain number of samples are randomly extracted from the original data with replacement as the training set for each decision tree. Assume that N samples are extracted. For each decision tree, in the process of building the tree, not all features are used, but a part of the features (for example, m, m is less than the total number of features) are randomly selected to determine the node split. Starting from the root node, the sample data is split into left and right child nodes according to the selected features and the corresponding thresholds, so that the data in the child nodes is as "pure" as possible, that is, more homogeneous. This process is recursive until the stopping condition is reached, such as the number of samples in the node is less than a certain value or the depth of the tree reaches the preset maximum value. Repeat steps 1-3 to build multiple decision trees to form a random forest. When predicting a new sample: the features of the new sample are input into each decision tree in the random forest, and each decision tree gives a predicted test tube position vector. Combining the prediction results of all decision trees, the final prediction result is obtained by taking the average or majority vote, that is, the test tube position vector of the new sample at a given transmission distance. By continuously adjusting the hyperparameters such as the number of decision trees in the random forest, the maximum depth of each tree, and the number of randomly selected features, and evaluating the performance of the model on the validation set, the optimal model structure and parameters are finally determined. This model considers how the different surface roughness of the test tube interacts with the surface roughness and vibration of the sorting line, while also taking into account the impact of the weight of the test tube on transmission. When a new test tube needs to be transported, as long as its relevant features and transmission distance are input, the model can predict the position vector of the test tube after that distance, that is, predict the placement status.

[0060] The data acquisition point of the test tube image data is collected. This data acquisition point is a specific position in the test tube transmission process. It is determined by relevant detection equipment or sensors, and the distance from this acquisition point to the first target position, that is, the first transmission distance, is calculated. In order to calculate this distance, the coordinate system in the physical space is used. For example, a three-dimensional coordinate system is established for the entire transmission system, and the coordinate values ​​of the acquisition point and the first target position in the coordinate system are determined. Through the difference in coordinate values ​​and combined with unit conversion (such as conversion from pixel distance to actual physical distance), mathematical formulas (such as the Pythagorean theorem, etc.) are used to calculate the straight-line distance between the two points. By accurately calculating the first transmission distance, key parameters are provided for the subsequent pre-test tube placement status.

[0061] The first transmission distance calculated previously is input into the test tube position vector prediction model that has been constructed. This prediction model has been trained and optimized with a large amount of data, and can infer the placement state of the test tube after passing this transmission distance based on the input transmission distance and previously acquired test tube transmission characteristics, sorting line transmission characteristics, test tube weight and other information. After the first transmission distance is input, the model will perform a series of complex calculations and analyses. Ultimately, the model outputs the first predicted placement state, which includes detailed information such as the test tube's position coordinates, tilt angle, and rotation angle. By generating such a first predicted placement state, an important reference is provided for subsequent sorting operations, which helps to make adjustments and preparations in advance and improve the accuracy and efficiency of sorting.

[0062] In a possible implementation, step S400 further includes:

[0063] Step S450: An abnormal aggregation terminal area is provided at the end point of the test tube sorting line, wherein the abnormal aggregation terminal area is used to store unsorted test tubes on the test tube sorting line.

[0064] Step S460: collecting the labels of the missed-sorting test tubes in the abnormal aggregation terminal area, matching them with the first sorting point, and collecting the time nodes when the unsorted test tubes reach the matching sorting point.

[0065] Step S470: identifying the parallel clustering coefficient of the same type of test tubes at the time node, and if the parallel clustering coefficient is greater than a preset threshold, controlling the upstream test tube flow of the first sorting point.

[0066] Specifically, in order to deal with possible abnormalities in test tube sorting, a specific area is specially set up at the end of the test tube sorting line, namely the abnormal aggregation end area. This area has clear functions and uses, and is specifically used to store test tubes that have not been sorted in a timely and correct manner in the normal sorting process. The reason for setting up such an area is to prevent unsorted test tubes from arbitrarily piling up or causing chaos on the sorting line, thereby ensuring the orderly operation of the entire sorting line. For example, if the sorting system fails, or the characteristics of some test tubes are difficult to identify and cannot be sorted in time, these test tubes will be transported to this abnormal aggregation end area for temporary storage. For another example, when the sorting speed cannot keep up with the input speed of the test tubes, causing some test tubes to be too late to be sorted, they will also be directed to this area for subsequent processing. By setting up the abnormal aggregation end area, the unsorted test tubes can be effectively managed centrally, which is convenient for further processing and analysis of these test tubes, and the reliability and stability of the entire test tube sorting system can be improved.

[0067] A series of operations are taken for the missed sorting test tubes in the abnormal aggregation terminal area. First, the label information of these missed sorting test tubes will be collected. The label contains important identification information such as the type, number, and source of the test tube. Then, these labels are matched with the first sorting point to determine that these missed sorting test tubes should have been processed at the first sorting point, but due to some reasons, they were not sorted here. At the same time, the time node when the unsorted test tubes arrived at the matching sorting point will be accurately collected. The collection of this time node is very critical. It can help us understand the specific time when the missed sorting occurred, so as to further analyze the possible causes and impacts. For example, if the label of a missed sorting test tube shows that it should arrive at the first sorting point at 10 am, but it is actually found in the abnormal aggregation terminal area at 10:15 am, then 10 am is the time node when the collected test tube should arrive at the matching sorting point. Through such collection and matching operations, accurate and detailed data support is provided for subsequent problem analysis, process optimization, and responsibility tracing, which helps to improve the efficiency and accuracy of the entire test tube sorting system.

[0068] At the time node collected earlier, the parallel clustering coefficient of the same type of test tubes will be identified and analyzed. The parallel clustering coefficient is an indicator used to measure the degree of aggregation of the same type of test tubes in a specific time and space. It comprehensively considers factors such as the number, distribution density, and relationship between the same type of test tubes. If the identified parallel clustering coefficient is greater than the preset threshold, it means that the aggregation degree of the same type of test tubes at this time node is too high, which will have an adverse effect on the sorting process, such as causing congestion and increasing the probability of wrong sorting. At this time, it is necessary to control the upstream test tube flow of the first sorting point. There are many ways to control it, such as reducing the input speed of the upstream test tube, temporarily stopping the input of the upstream test tube, or adjusting the working rhythm of the upstream conveying device. For example, assuming that the preset threshold is 0.6, the parallel clustering coefficient identification result of a certain type of test tube at a certain time node is 0.8. This indicates that the aggregation degree of this type of test tube is too high, which will cause excessive local pressure on the sorting line. Therefore, measures will be taken, such as reducing the input speed of upstream test tubes by 50% to reduce the number of test tubes entering the first sorting point, thereby alleviating the aggregation and ensuring the normal sorting work. Through this real-time monitoring and control mechanism, sorting system failures caused by excessive test tube aggregation can be effectively avoided, improving sorting efficiency and accuracy.

[0069] In a possible implementation, step S470 further includes:

[0070] Step S471: if the parallel clustering coefficient is less than or equal to the preset threshold, locate the first missed detection rate indicator of the first sorting point.

[0071] Step S472: If the first missed detection rate indicator is greater than a preset missed detection rate threshold, a first sorting abnormality signal is generated.

[0072] Specifically, when the parallel clustering coefficient is less than or equal to the preset threshold, it means that the degree of clustering of the same type of test tubes at the current time node is within an acceptable range. At this point, attention will be turned to the first missed detection rate indicator for locating the first sorting point. The positioning of this indicator is to evaluate the test tubes that are not correctly sorted at the first sorting point. The calculation of the missed detection rate indicator is usually calculated by counting the number of test tubes that pass through the first sorting point but are not sorted correctly within a certain period of time, and comparing it with the total number of test tubes passing through this point to obtain the ratio. For example, within an hour, a total of 100 test tubes pass through the first sorting point, of which 10 are not sorted correctly, then the first missed detection rate indicator is 10%. By locating and determining this missed detection rate indicator, the work efficiency and accuracy of the first sorting point can be more accurately understood, providing key data support for subsequent analysis and improvement.

[0073] If the first missed detection rate index obtained by the previous positioning is greater than the preset missed detection rate threshold, this indicates that the missed detection situation of the first sorting point exceeds the acceptable level, and the first sorting abnormality signal is generated accordingly. The preset missed detection rate threshold is a standard value set in advance according to the requirements and expected performance of the sorting system. When the first missed detection rate index exceeds this threshold, it means that there are more significant problems in the sorting work. For example, the preset missed detection rate threshold is 5%, and the first missed detection rate index actually calculated is 8%, which exceeds the threshold, and the first sorting abnormality signal will be generated immediately. This abnormal signal can be a specific sound and light alarm, a prominent prompt displayed on the system interface, or a notification message sent to relevant staff. Its purpose is to attract attention in time so that relevant personnel can take measures quickly, such as checking whether the sorting equipment is faulty, recalibrating the identification system, increasing manual intervention, etc., to solve the missed detection problem of the first sorting point as soon as possible, restore the normal sorting workflow, and ensure the accuracy and efficiency of sorting.

[0074] In a possible implementation, step S400 further includes:

[0075] Step S480: When the dual-channel sample abnormality inspection fails, matching abnormal sorting points.

[0076] Step S490: sorting the test tube to be sorted to the abnormal sorting point according to the preset sorting scheme, and generating a sample abnormality signal.

[0077] Specifically, if the abnormal inspection of the two-way sample finds a problem, such as the label is wrong or the sample weight does not meet the requirements, the operation of matching the abnormal sorting point will be started. When the inspection result shows that the label is wrong, according to the preset rules, find a sorting point specifically used to handle abnormalities such as label errors. This is because label errors will affect the subsequent identification, classification and processing processes, and require specific operations to correct. If the sample weight does not meet the standard, it means that the sample is missing, excessive or otherwise inconsistent with the expected weight. It will also be matched to the sorting point for abnormal sample weight according to the pre-set settings. For example, if the label of a sample is found to have the wrong product batch information in the two-way inspection, it will be quickly matched to an abnormal sorting point that specializes in handling label errors and can recheck and correct the label. For another example, if the sample weight is far below the normal range, indicating that the sample is incomplete, it will be matched to a sorting point used to check and handle incomplete samples. Through such precise abnormal type identification and corresponding abnormal sorting point matching, it can ensure that different types of abnormal samples can be properly and efficiently processed.

[0078] When it is determined that the test tubes to be sorted are abnormal and matched to the abnormal sorting points, the corresponding sorting mechanism is started to accurately transport these test tubes to be sorted to the designated abnormal sorting points. This sorting process involves the grabbing and movement of the robotic arm, the steering of the conveyor belt, or the coordinated work of other automated equipment to ensure that the test tubes can smoothly reach the abnormal sorting points. At the same time, a sample abnormality signal is generated. The generation of this signal is of great significance. It will be presented in various forms, such as popping up a warning window on the display, issuing a specific sound alarm, and sending a notification message to the terminal device of the relevant staff. Through such a sorting and signal generation mechanism, abnormal samples can be effectively managed to ensure the normal progress and quality control of the entire sorting process.

[0079] Embodiment 2

[0080] Based on the same inventive concept as the test tube sorting intelligent identification method based on multi-source sensors in the aforementioned embodiment, Figure 2 As shown, the present application provides a test tube sorting intelligent identification device based on a multi-source sensor, and the device and method embodiments in the present application are based on the same inventive concept. The device includes:

[0081] The multi-source data acquisition module 10 is used to perform multi-source data acquisition on the test tubes to be sorted through the multi-source sensor modules on the test tube sorting line, and generate test tube label analysis data, test tube image data and test tube weight data.

[0082] The first test tube sorting attribute information reading module 20 performs test tube sample type matching based on the test tube label parsing data, locates the first sorting point, and reads the first test tube sorting attribute information, wherein the first test tube sorting attribute information includes the sample type and the sample weight.

[0083] The dual-path sample anomaly detection module 30 is used to perform dual-path sample anomaly detection in combination with the test tube image data, the test tube weight data and the first test tube sorting attribute information.

[0084] The preset sorting scheme generating module 40 is used to read the first target position of the first sorting point when the two-way sample abnormality inspection passes, and when the test tube to be sorted reaches the first target position, make a test tube sorting scheme decision based on the test tube image data to generate a preset sorting scheme.

[0085] The test tube sorting module 50 is used to control the first sorting robot arm to sort the test tubes to be sorted to the first test tube storage rack according to a preset sorting scheme.

[0086] Furthermore, the first test tube sorting attribute information reading module 20 also includes:

[0087] A sorting point acquisition unit, the sorting point acquisition unit is used to acquire multiple sorting points distributed on the test tube sorting line, wherein the multiple sorting points have a sorting sample type identifier and a preset sample weight identifier; a first sorting point generation unit, the first sorting point generation unit performs point matching based on the test tube label analysis data and the sorting sample type identifier to generate the first sorting point, and generates the first test tube sorting attribute information.

[0088] Furthermore, the dual-path sample abnormality inspection module 30 further includes:

[0089] a type matching abnormality inspection branch establishing unit, wherein the type matching abnormality inspection branch establishing unit presets the test tube image data configuration based on the sample type in the first test tube sorting attribute information, and establishes a sample type matching abnormality inspection branch; a weight matching abnormality inspection branch establishing unit, wherein the weight matching abnormality inspection branch establishing unit establishes a sample weight matching abnormality inspection branch based on the sample weight in the first test tube sorting attribute information; a dual-path sample abnormality inspection module establishing unit, wherein the dual-path sample abnormality inspection module establishing unit is used to establish a dual-path sample abnormality inspection module with the sample type matching abnormality inspection branch and the sample weight matching abnormality inspection branch; and a dual-path sample abnormality inspection unit, wherein the dual-path sample abnormality inspection unit is used to perform a dual-path sample abnormality inspection on the test tube image data and the test tube weight data through the dual-path sample abnormality inspection module.

[0090] Furthermore, the preset sorting scheme generating module 40 further includes:

[0091] a reference grabbing area establishing unit, the reference grabbing area establishing unit establishing a reference grabbing area of ​​a first sorting robot arm based on the sample type in the first test tube sorting attribute information; a test tube initial placement state positioning unit, the test tube initial placement state positioning unit being used to locate the test tube initial placement state based on the test tube image data with the reference grabbing area as an analysis reference point; a first predicted placement state generating unit, the first predicted placement state generating unit predicts a test tube placement state change for the first target position based on the test tube initial placement state, and generates a first predicted placement state; and a preset sorting scheme generating unit, the preset sorting scheme generating unit performing sorting path planning of the first sorting robot arm based on the first predicted placement state, and generating the preset sorting scheme.

[0092] Furthermore, the first predicted placement state generating unit further includes:

[0093] a transmission feature acquisition unit, the transmission feature acquisition unit is used to acquire the test tube transmission feature and the sorting line transmission feature, wherein the test tube transmission feature is the outer surface roughness of the test tube to be sorted, and the sorting line transmission feature is the surface roughness and vibration feature of the test tube sorting line; a prediction model establishment unit, the prediction model establishment unit is used to collect test tube transmission samples based on the test tube transmission feature and the sorting line transmission feature and in combination with the test tube weight data, and establish a test tube position vector prediction model about the transmission distance; a first transmission distance calculation unit, the first transmission distance calculation unit is used to collect data collection points of the test tube image data, and calculate the first transmission distance from the data collection point to the first target position; a first predicted placement state generation unit, the first predicted placement state generation unit is used to input the first transmission distance into the test tube position vector prediction model to generate the first predicted placement state.

[0094] Furthermore, the preset sorting scheme generating module 40 further includes:

[0095] An abnormal aggregation terminal area setting unit, the abnormal aggregation terminal area setting unit is used to set an abnormal aggregation terminal area at the terminal of the test tube sorting line, wherein the abnormal aggregation terminal area is used to store unsorted test tubes on the test tube sorting line; a time node collection unit, the time node collection unit is used to collect the labels of missed sorted test tubes in the abnormal aggregation terminal area, match them with the first sorting point, and collect the time nodes when the unsorted test tubes arrive at the matching sorting point; a test tube flow control unit, the test tube flow control unit is used to identify the parallel aggregation coefficient of the same type of test tubes at the time node, and if the parallel aggregation coefficient is greater than a preset threshold, the upstream test tube flow of the first sorting point is controlled.

[0096] Furthermore, the test tube flow control unit also includes:

[0097] a first missed detection rate indicator locating unit, the first missed detection rate indicator locating unit being used to locate the first missed detection rate indicator of the first sorting point if the parallel clustering coefficient is less than or equal to the preset threshold; a first sorting abnormality signal generating unit, the first sorting abnormality signal generating unit being used to generate a first sorting abnormality signal if the first missed detection rate indicator is greater than a preset missed detection rate threshold.

[0098] Furthermore, the preset sorting scheme generating module 40 further includes:

[0099] An abnormal sorting point matching unit, the abnormal sorting point matching unit is used to match the abnormal sorting point when the dual-path sample abnormality test fails; a sample abnormality signal generating unit, the sample abnormality signal generating unit is used to sort the test tube to be sorted to the abnormal sorting point according to the preset sorting scheme, and generate a sample abnormality signal.

[0100] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0101] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

[0102] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A test tube sorting intelligent identification method based on multi-source sensors, characterized in that: include: The multi-source sensor module on the test tube sorting line collects multi-source data of the test tubes to be sorted, generating test tube label analysis data, test tube image data and test tube weight data; Perform test tube sample type matching based on the test tube label parsing data, locate the first sorting point, and read the first test tube sorting attribute information, wherein the first test tube sorting attribute information includes the sample type and the sample weight; performing a dual-path sample abnormality inspection in combination with the test tube image data, the test tube weight data and the first test tube sorting attribute information; When the two-way sample abnormality inspection passes, the first target position of the first sorting point is read, and when the test tube to be sorted reaches the first target position, a test tube sorting scheme decision is made based on the test tube image data to generate a preset sorting scheme; Controlling the first sorting robot arm to sort the test tubes to be sorted to the first test tube storage rack according to a preset sorting scheme; When the test tube to be sorted reaches the first target position, a test tube sorting scheme decision is made based on the test tube image data to generate a preset sorting scheme, including: Based on the sample type in the first test tube sorting attribute information, establishing a reference grabbing area of ​​the first sorting robot arm; Using the reference grabbing area as an analysis reference point, locating the initial placement state of the test tube based on the test tube image data; Based on the initial placement state of the test tube, predicting a change in the placement state of the test tube at the first target position to generate a first predicted placement state; Perform sorting path planning of the first sorting robot arm based on the first predicted placement state to generate the preset sorting plan; Based on the initial placement state of the test tube, predicting a change in the placement state of the test tube at the first target position to generate a first predicted placement state includes: Acquire test tube transmission characteristics and sorting line transmission characteristics, wherein the test tube transmission characteristics are the outer surface roughness of the test tube to be sorted, and the sorting line transmission characteristics are the surface roughness and vibration characteristics of the test tube sorting line; Using the test tube transmission characteristics and the sorting line transmission characteristics, and combining the test tube weight data to collect test tube transmission samples, a test tube position vector prediction model for transmission distance is established; Collecting data collection points of the test tube image data, and calculating a first transmission distance from the data collection points to the first target position; The first transmission distance is input into the test tube position vector prediction model to generate the first predicted placement state.

2. The test tube sorting intelligent identification method based on multi-source sensors as claimed in claim 1, characterized in that: The test tube sample type is matched based on the test tube label parsing data, the first sorting point is located, and the first test tube sorting attribute information is read, including: Acquire a plurality of sorting points distributed on the test tube sorting line, wherein the plurality of sorting points have sorting sample type identifiers and preset sample weight identifiers; Point matching is performed based on the test tube label parsing data and the sorting sample type identifier to generate the first sorting point and the first test tube sorting attribute information.

3. The test tube sorting intelligent identification method based on multi-source sensors as claimed in claim 1, characterized in that: Combining the test tube image data, the test tube weight data and the first test tube sorting attribute information to perform a dual-path sample abnormality inspection, including: Preset the test tube image data configuration based on the sample type in the first test tube sorting attribute information, and establish a sample type matching abnormality inspection branch; Establishing a sample weight matching abnormality inspection branch based on the sample weight in the first test tube sorting attribute information; Establishing a dual-path sample anomaly inspection module with the sample type matching anomaly inspection branch and the sample weight matching anomaly inspection branch; The dual-path sample abnormality inspection module performs a dual-path sample abnormality inspection on the test tube image data and the test tube weight data.

4. The test tube sorting intelligent identification method based on multi-source sensors as claimed in claim 1, characterized in that: Also includes: The end point of the test tube sorting line is provided with an abnormal aggregation end point area, wherein the abnormal aggregation end point area is used to store the test tubes that have not been sorted on the test tube sorting line; Collect the labels of the missed-sorting test tubes in the abnormal aggregation endpoint area, match them with the first sorting point, and collect the time nodes when the unsorted test tubes reach the matching sorting point; The parallel clustering coefficient of the same type of test tubes is identified at the time node, and if the parallel clustering coefficient is greater than a preset threshold, the upstream test tube flow of the first sorting point is controlled.

5. The test tube sorting intelligent identification method based on multi-source sensors as claimed in claim 4, characterized in that: Also includes: If the parallel clustering coefficient is less than or equal to the preset threshold, locating a first missed detection rate indicator of a first sorting point; If the first missed detection rate indicator is greater than a preset missed detection rate threshold, a first sorting abnormality signal is generated.

6. The test tube sorting intelligent identification method based on multi-source sensors as claimed in claim 1, characterized in that: Also includes: When the two-way sample abnormality inspection fails, matching abnormal sorting points; The test tube to be sorted is sorted to the abnormal sorting point according to the preset sorting scheme, and a sample abnormality signal is generated.

7. A test tube sorting intelligent identification device based on multi-source sensors, characterized in that: The device is used to execute the method according to any one of claims 1 to 6, and the device comprises: A multi-source data acquisition module, which is used to perform multi-source data acquisition on the test tubes to be sorted through the multi-source sensor module on the test tube sorting line, and generate test tube label analysis data, test tube image data and test tube weight data; a first test tube sorting attribute information reading module, which performs test tube sample type matching based on the test tube label parsing data, locates the first sorting point, and reads the first test tube sorting attribute information, wherein the first test tube sorting attribute information includes sample type and sample weight; A two-way sample anomaly inspection module, the two-way sample anomaly inspection module is used to perform a two-way sample anomaly inspection in combination with the test tube image data, the test tube weight data and the first test tube sorting attribute information; A preset sorting scheme generating module, wherein the preset sorting scheme generating module is used to read the first target position of the first sorting point when the two-way sample abnormality inspection passes, and when the test tube to be sorted reaches the first target position, make a test tube sorting scheme decision based on the test tube image data to generate a preset sorting scheme; A test tube sorting module is used to control the first sorting robot arm to sort the test tubes to be sorted to the first test tube storage rack according to a preset sorting scheme.

Images