373results about How to "Achieve correction" patented technology

The invention discloses a stationary pollution source data acquisition and calibration method and system, a medium and a program product, and the method comprises the steps: carrying out the segmentation processing of a plurality of concentration change data, and obtaining a plurality of data segments; determining the concentration change trend of various pollutants, and calculating a correlation coefficient matrix; the Euclidean distance value between the correlation coefficient matrix of the current data segment and each correlation coefficient matrix before sorting is calculated; when the Euclidean distance is greater than a preset threshold value, marking the corresponding data segment as an abnormal data segment; determining the pollutant as a target pollutant, and establishing a material balance equation; based on the material balance equation, generating a theoretical concentration change curve of the target pollutant; calculating a deviation value, and when the deviation value exceeds a preset dynamic error limit value, determining a sensor correction coefficient according to the process parameters; and carrying out calibration processing on the acquired data of the target pollutant according to the sensor correction coefficient. According to the invention, the accuracy of data acquisition and calibration of the stationary pollution source is improved, so that the influence on formulation and execution of pollution prevention and control measures is reduced.

The invention relates to the technical field of biological information, in particular to a method and system for online integration of multi-source single cell data. The method comprises the following steps: acquiring multi-source single cell transcriptome data; performing data quality evaluation based on gene expression quantity, mitochondrial gene proportion and cell data depth on the single cell transcriptome data to obtain initial quality control data; setting standardized filtering parameters according to the initial quality control data, and performing quality control parameter filtering on the single cell transcriptome data to obtain filtered cell data; performing gene expression characteristic analysis on the filtered cell data to generate high-variation gene expression characteristic data; and performing gene expression spatio-temporal distribution statistics according to the high-variation gene expression feature data to generate gene expression statistical feature data. According to the method, through refined quality control and batch effect correction, data from different sources can be seamlessly integrated on the premise of ensuring data quality, so that noise caused by different batches and technical platforms is eliminated.

The invention belongs to the technical field of fundus imaging, and discloses an ultra-wide-angle ellipsoid confocal laser scanning fundus imaging system, which comprises a light splitting module, a light condensing module, a scanning module, a transmitting mirror group, an ellipsoid reflector and a detection module, wherein the center of the scanning module and the center of the transmitting mirror group are located on the same optical axis A, the pupil of the human eye is located at the first focus of the ellipsoidal reflector, and the entrance pupil of the transmitting mirror group is located at the second focus of the ellipsoidal reflector; and the transmitting lens group comprises a movable focusing lens and is used for changing the position of a first image surface of light reflected by the retina of the human eyes when moving along the optical axis A so as to realize clear imaging of the retina of the human eyes with different diopters. Under the condition that a free-form surface is not introduced, ultra-wide-angle multi-wavelength confocal laser scanning fundus imaging with the single intraocular field angle of 200 degrees is achieved, and the full-field optical resolution reaches 15 micrometers. And by moving the focusing lens, imaging compensation can be carried out on human eyes with diopters of-15D to + 15D.

The invention relates to the technical field of color correction, and discloses a publication printing color correction method and system based on deep learning, and the method achieves the precise correction and compensation of the color of a publication printed matter through a series of steps of cosine distance calculation, color-environment matching model, color deviation calculation, operation parameter adjustment and the like. The core of the method lies in that the accuracy of the printing color is effectively improved and the color quality of the printed matter is ensured to meet the expected standard through careful analysis and processing of the printing color by using the deep learning technology. The correction method provided by the invention has obvious advantages, not only can realize automatic color correction, reduce human intervention and improve the stability and consistency of correction, but also can optimize the operation parameters of the printing equipment, so that the printing equipment can keep a good working state in different printing environments, and the adaptability of the printing process is enhanced.

The invention provides a frame-mounted miniaturized high-speed ground penetrating radar jitter compensation method and system, and relates to the technical field of ground penetrating radar detection.The method comprises the steps that a data processor receives tilt attitude information and calculates a frame tilt angle, and a compensation processor receives the frame tilt angle and calculates the frame tilt angle; a compensation processor performs road condition-based adaptive noise and angle threshold adjustment and multi-model switching Kalman filtering processing on the vehicle frame inclination angle information, predicts attitude information of a vehicle frame platform at the next moment and updates an angle threshold; meanwhile, the compensation processor generates a compensation motion command according to the predicted data and sends the compensation motion command to a displacement controller of the mechanical compensation device; the distance compensation algorithm is used for processing to obtain motion information, and the motion information is transmitted to the mechanical compensation device to execute compensation action. According to the invention, Kalman filtering processing, an adaptive noise and angle threshold adjustment mechanism and a linear and nonlinear compensation switching mechanism are used for prediction updating, adaptive adjustment is carried out according to road surface changes, and the accuracy of data acquisition is improved.

The invention discloses a building electric energy meter energy consumption acquisition error identification and correction method, which is characterized in that an error identification and correction model is established by improving a BP neural network, the influence of manual intervention is eliminated, and the correction of loss data of power equipment is realized. Meanwhile, the models are refined and applied to power sub-meter equipment and general meter equipment, and a sub-meter equipment BP neural network model and a general meter equipment BP neural network model are established. The method comprises the following steps: firstly, correcting the deviation of sub-table equipment data through an improved BP neural network, adding the sub-table data according to lines to obtain total table data, and then correcting the deviation of the total table equipment data through the improved BP neural network to further reduce the error. In addition, data collected by different devices are processed in a unified mode through a data standardization method, the influence of different device collection modes is reduced, and finally error correction of the energy consumption data is achieved.

The invention discloses an image content recognition system based on deep learning, which relates to the technical field of image content recognition and comprises an image processing module, an edge point acquisition module, a boundary straight line fitting module, an inclination angle acquisition module, an inclination judgment module and an inclination correction module. The method comprises the following steps: judging whether an image is inclined or not, generating a corresponding identifier, judging whether correction is needed or not, calculating a correction coordinate according to a central point coordinate, a pixel point coordinate, an inclination angle and a slope of the image after judging that the image is inclined, and adjusting the position of a pixel point in the inclined image, so as to realize the correction of the image. The image is restored to a relatively normal state through the correction operation, and subsequent image content recognition is facilitated, so that the difficulty and the influence of inclination on subsequent image content recognition are reduced, and the image recognition accuracy is improved.

The present invention relates to the field of waterway monitoring technology, and specifically, to a real-time monitoring method and system for water depth of a port waterway. It includes: S1, selecting a monitoring point in the waterway, installing an intelligent buoy for monitoring waterway data at the monitoring point, the intelligent buoy is used to monitor waterway data, and deploying an acoustic Doppler current profiler to monitor the bottom flow velocity; S2, collecting waterway data collected by a sonar system and an acoustic Doppler current profiler; S3, constructing a waterway sedimentation model, and combining the real-time waterway data after preprocessing, introducing a time-dependent function, and correcting the measurement error of the real-time water depth data caused by silt silting; S4, based on the above steps S1-S3, constructing a network including multiple monitoring points to form a three-dimensional monitoring system for the waterway. By establishing a preliminary waterway sedimentation model of sedimentation rate and average bottom flow velocity, and introducing a time-dependent function and terrain factors, the prediction accuracy of future water depth changes in the waterway is improved.

The invention belongs to the technical field of valve detection, and discloses a detection device for valve processing and a detection method thereof.The detection device comprises a base, a support is fixedly arranged at the top of the base, a photosensitive assembly is nested in the top surface of the base, and a motor is fixedly arranged at the top of the support. By adapting to the change of the internal size of the to-be-detected valve, when the internal diameter of the valve is changed, the position of the closely attached detection assembly is changed, so that the position of the corresponding laser emitting end is changed, the change of the laser irradiation position is sensed through the light sensing assembly, and the detection of the internal diameter size of the valve is realized; the size change of the pipe diameters at the two ends of the valve is judged by matching the sensing positions of the standard photosensitive area, the inner side photosensitive area and the outer side photosensitive area, the reduced and expanded areas can be accurately detected, the quality condition of the valve can be accurately identified and detected and quickly judged according to the increase or decrease of the local size of the inner wall of the valve, the detection is simple, and the detection efficiency is high. The use effect is good.

The invention discloses a raster scanning correction method and system for an electromagnetic MEMS micromirror, and belongs to the technical field of raster scanning correction, and the method comprises the steps: collecting the real-time three-axis attitude information of an electromagnetic MEMS micromirror product in a three-dimensional space, and the fast and slow axis piezoresistive signals of the micromirror; carrying out PID control by taking the angle difference value as an error signal, outputting a driving voltage of a fast axis and a slow axis of the MEMS micro-mirror after the scanning view field correction, and further mapping to obtain a correction angle of micro-mirror scanning; and taking the correction angle as the input of PID control, and correcting the rotation angle of the micromirror by minimizing the tracking error between the rotation angle of the micromirror and the correction angle. Three-axis attitude information of an electromagnetic MEMS micro-mirror product is collected, and an angle difference value between the three-axis attitude information and standard three-axis attitude information is introduced into PID control to adjust a fast-slow axis driving signal, so that the rotation angle of the micro-mirror is corrected, and correction and compensation of grating distortion in grating scanning are realized. And the accuracy of the reconstruction quality of the three-dimensional point cloud is ensured.

The invention discloses a fault diagnosis method based on multi-dimensional knowledge migration. The method comprises the following steps: acquiring a source domain data set and a multi-target domain data set; constructing a double-branch neural network model based on guidance of data knowledge and semantic knowledge, training the double-branch neural network model, and optimizing model parameters by taking minimization of a total loss function of the double-branch neural network model as an optimization target; wherein the model comprises M + 1 data knowledge guidance branch networks and M + 1 semantic knowledge guidance branch networks; and performing fault diagnosis tasks of different target domains by using the trained double-branch neural network model. According to the method, the problem of negative migration caused by forced alignment distribution in traditional migration learning can be solved, and effective knowledge migration from a source domain to a multi-target domain is realized from the perspective of multi-dimensional knowledge fusion, interaction and driving; the method can be applied to industrial fault diagnosis tasks of industrial operation and maintenance and rapid migration deployment in various migration learning scenes, and has a good learning effect and generalization ability.

The invention relates to the field of steel bar rust removal equipment, in particular to a steel bar pretreatment device and method for a precast beam. The steel bar pretreatment device comprises a steel bar rust removal machine, the steel bar rust removal machine comprises a machine shell, a control panel is installed in the middle of one side of the machine shell, and material ports are formed in the middle of the inlet end and the middle of the outlet end of the machine shell; the device comprises a machine shell, mounting frames are mounted on the two sides of the middle of the interior of the machine shell, the ends, close to each other, of the mounting frames are fixedly connected with evenly-distributed clamping discs, rotating shafts are rotationally connected to the middles of the clamping discs, the tail ends of the rotating shafts are fixedly connected with n-shaped frames, and servo motors are fixedly connected to the tops of the n-shaped frames; and the driving ends of the bottoms of the servo motors are fixedly connected with rotating shafts, and the rotating shafts penetrate through the [-shaped frame. And the situation that the grinding effect is affected due to the fact that the steel bar body is bent and cannot make full contact with the grinding component in the grinding process of the steel bar is avoided.

The present invention discloses a photoacoustic microscopy imaging system based on a high-speed focuser. The photoacoustic microscopy imaging system includes a light source component, a single-mode fiber coupling component, a high-speed focuser component, a light beam scanning component, a reflective bracket, a signal acquisition component and a computer. The present invention adopts a mobile focusing design. During the photoacoustic imaging point scanning process, the displacement of the double-cemented lens is controlled by the movement of a fast linear motor, and then the focal position of the light beam is adjusted, so that during the scanning process, the focal plane at any position always coincides with the imaging surface. The present invention establishes an imaging method for automatically adjusting the focal plane by adopting a high-speed focuser, which effectively improves the image signal-to-noise ratio and resolution, and is low in price and easy to operate. The present invention adjusts the imaging in real time according to the photoacoustic signal fed back during the scanning process, and can perform photoacoustic microscopy imaging of any unknown surface, which has broad research prospects and application value.

The invention relates to the technical field of data processing, in particular to a data processing method for a high-dynamic physical simulation system. The method comprises the following steps: acquiring continuous multi-frame infrared image data of a high-speed moving object; performing region division on the continuous multi-frame infrared image data to obtain real object region data and background region data of the object; identifying ghosting data in the real object region data; and correcting the ghost data to obtain target simulation analysis data. According to the embodiment of the invention, the method can achieve the data correction of the ghost shadow of the high-speed moving object, provides more accurate simulation experiment data, and improves the accuracy of a simulation result.

The invention discloses a positioning method of a suspension rail type immersed tunnel inspection platform, which is characterized in that the inspection platform is suspended on a rail of a rail structure in a walking manner, the inspection platform is provided with two clamping driving wheel mechanisms symmetrically arranged on two sides of the rail, and the clamping driving wheel mechanisms can adjust the wheel diameters of driving wheels; an inspection platform auxiliary marking system is designed, and a two-dimensional code label is arranged on the bottom surface of the track corresponding to each detection point; when the inspection platform runs to a detection point position, reading absolute position information of the two-dimensional code label and label information of the detection point position, calculating a deviation between an absolute position obtained by reading the two-dimensional code label and a current displacement amount obtained according to data of the rotary encoder, and then according to the deviation, determining whether the detection point position is a detection point position or not; the wheel diameter of the driving wheel is adjusted by adjusting the pressure of the driving wheel of the clamping driving wheel mechanism on the rail, so that the moving position of the inspection platform is corrected, and the positioning precision of the inspection platform is remarkably improved.

The invention relates to a clock calibration method and device. The method comprises the following steps: obtaining a corresponding relation between a reference clock frequency generated by an internal oscillator and a baud rate frequency of a universal asynchronous transceiver; obtaining a time sequence error of each bit time in a transmission period of the universal asynchronous transceiver according to the corresponding relationship, and accumulating the time sequence errors into an accumulated error; and compensating the accumulative error to calibrate the Baud rate frequency.

The invention relates to an intelligent acquisition, analysis, prediction and automatic violation correction method and system based on public traffic violation big data, and belongs to the field of traffic informatization and intelligent management. The objective of the invention is to identify complex and diverse traffic illegal behaviors and analyze massive illegal data. The system comprises data acquisition, processing and storage, analysis and prediction, automatic violation correction and data security. Through multi-source data access, edge calculation and distributed storage, real-time acquisition and efficient processing of illegal behaviors are realized; deep analysis is carried out on the traffic violation data by using deep learning and big data technologies; and through real-time video stream analysis, tracking, recording and evidence obtaining are performed on suspected illegal behaviors. The system adopts a multi-level security protection mechanism, data privacy and security are ensured, and flexible expansion and multi-system collaboration are realized through modular design. Compared with a traditional system, the system has remarkable advantages in the aspects of data acquisition comprehensiveness, processing efficiency, prediction accuracy, violation correction automation, data security and the like.

The invention discloses a control sheet generation and geometric positioning method fusing satellite-borne laser and an orthoimage. The method comprises the following steps: firstly, screening satellite-borne laser point clouds to obtain a high-quality laser point set; secondly, projecting high-quality satellite-borne laser points obtained by screening onto an orthoimage through coordinate transformation, and removing low-feature response points; and finally, generating a control sheet based on the orthoimage and the laser control point, and performing geometric correction on the target image by using the constructed control sheet to realize high-precision geometric positioning of the image. Different from the mode that only laser height measurement data is adopted for elevation control, the satellite-borne laser data and the orthoimage are fused, accurate horizontal height control over the remote sensing image is achieved, meanwhile, the defect that the DEM precision is limited is overcome, and the method is more suitable for geometric positioning of sub-meter images.

The invention relates to the technical field of image enhancement three-dimensional reconstruction, in particular to an urban building three-dimensional reconstruction method and system based on virtual modeling, and the method comprises the steps: obtaining a gray-scale map and a depth map of an urban building at different visual angles through employing a camera carried by an unmanned plane; obtaining the depth homogeneity of each pixel point in each depth map based on the similarity degree of the depth value change in the local window of each pixel point in each depth map; based on the spatial difference and gray level distribution difference of point cloud data of each visual angle and the adjacent visual angle thereof, and the gray level difference condition of pixel points at the same position in the gray level images of each visual angle and the adjacent visual angle thereof, the reflection standard difference and the depth homogeneity are combined; and obtaining a luminosity compensation weight of each pixel point in each visual angle grey-scale map to construct a loss function, and training a deep learning model to obtain a three-dimensional model of the urban building. According to the invention, the accuracy of virtual modeling three-dimensional reconstruction is improved.

The utility model provides a transmission-type imaging spectrometer. The transmission-type imaging spectrometer comprises a slit, a collimating lens group, a light splitting module, an imaging group and a detector which are arranged in sequence, the light splitting module comprises a grating and a prism; polychromatic light enters the collimating lens group through the slit, is collimated into parallel light, is projected to the grating for diffraction light splitting, then enters the prism for refraction light splitting, and is converged on the detector through the imaging group after light splitting. According to the utility model, the light transmission efficiency can be improved, aberration can be corrected, and large-area image correction can be realized.

The invention relates to the technical field of spine correction instruments, in particular to a spine multi-dimensional unbalance corrector. Comprising an upper connecting frame and a lower connecting frame which are arranged on the back side, a first connecting sleeve is arranged on the upper connecting frame, and the first connecting sleeve is used for being connected with the position, corresponding to the chest section, of the trunk; the lower connecting frame is arranged below the upper connecting frame, a waist support is arranged at the front end of the lower connecting frame, the front end of the waist support is supported on the back waist, a second connecting sleeve is arranged on the lower side, corresponding to the waist support, of the lower connecting frame, and the second connecting sleeve is connected with the waist or below the waist; the power-assisted connecting piece is connected between the upper connecting frame and the lower connecting frame, the power-assisted connecting piece is connected with the lower connecting frame through a universal joint, the power-assisted connecting piece is connected with a pair of elastic dragging pieces which are arranged in a bilateral symmetry mode, and the elastic dragging pieces pull the power-assisted connecting piece obliquely downwards in the side rear direction. And correction of multi-dimensional unbalance of the spine can be realized.

The invention provides a signal lamp data processing method and equipment. The method is applied to a terminal and comprises the steps of sending signal lamp data request information to a cloud, receiving first signal lamp data, and determining second signal lamp data of a to-be-passed intersection according to the first signal lamp data. Wherein the signal lamp data request information comprises a vehicle position, the first signal lamp data is fed back by the cloud according to the signal lamp data request information, and the first signal lamp data comprises a current lamp color and lamp color timing information. The method is used for achieving the effect of correcting the signal lamp data at the terminal to make the signal lamp data more accurate.

The invention discloses a distorted image correction method based on a fast multi-scale residual network. Relates to the field of image geometric distortion correction, in particular to the technical field of distortion image correction based on a fast multi-scale residual network. Comprising the following steps: S1, constructing a data set, and generating a paired data set of a distorted image and a standard image and a deformation displacement field file of a corresponding number of the distorted image by using a geometric distortion model; s2, a distorted image correction model based on a fast multi-scale residual network is constructed, the model is divided into a backbone network and a tributary network, the backbone network is used for recovering image features, and the tributary network is used for outputting image classification features; s3, training the distorted image correction model based on the fast multi-scale residual network by using the pairwise data set of the distorted image and the standard image and the deformation displacement field file; and S4, processing the to-be-corrected image by using the trained distortion image correction model based on the fast multi-scale residual network to obtain the corrected image and the corresponding classification thereof.

The invention relates to the technical field of English phoneme personalized correction systems, in particular to an English phoneme personalized correction system and method based on acoustic feature analysis. Comprising a learner identity recognition module, a learner pronunciation information input module, a learner pronunciation information analysis module, a learner pronunciation difficulty point evaluation module, a learner personalized correction strategy generation module, a learner pronunciation difficulty correction module, a voice model evaluation module and a learning effect feedback module which are connected in sequence. According to the technical scheme of the invention, voice learners are served from multiple angles and in all directions from links such as information acquisition, problem diagnosis, strategy making, process monitoring and effect evaluation, and a set of intelligent and personalized voice learning solution is provided for the learners through innovative fusion of acoustic voice analysis and artificial intelligence technology. The method is of great significance in improving the voice level of English learners and promoting cross-cultural communication.

The invention discloses a drilling machine drill bit position correction method based on a laser tracker. The drilling machine drill bit position correction method comprises the steps that firstly, a detection system is built; step 2, correcting the spatial position of a drill bit: step 2.1, correcting the indexing of a drilling machine; 2.2, a drilling machine shaft is determined; step 2.3, equally dividing the root of the blade into a plurality of measurement areas, and selecting some preset hole sites in each measurement area; 2.4, axial drill bit axis correction and radial drill bit axis correction are sequentially carried out on each selected preset hole position; step 2.5, correcting the planeness of the root of the blade; 2.6, preset hole site correction is carried out; and step 2.7, correcting the surface deviation of the axial drill bit and the radial drill bit corresponding to each preset hole site. Therefore, it can be ensured that the axial bolt hole and the radial bolt hole formed by drilling in the blade root position of the wind power blade are matched with the theoretical bolt hole of the blade, and the drilling forming quality and precision are improved.

The invention provides a digital tee joint extrusion production method and equipment, and belongs to the technical field of tee joint extrusion production control. The method comprises the following steps: determining a corresponding extrusion operation parameter group sequence in real time; and inputting the extrusion operation parameter group sequence into the three-way production simulation model to determine defect prediction development information. Determining defect correction development information based on the defect prediction development information and preset standard operation regulation and control parameter information; and sending the defect correction development information and the extrusion operation parameter group sequence of the correction time period to a management terminal as defect correction information, so as to store the defect correction information and the evaluation label to a preset database based on the feedback evaluation information. And when a second extrusion material instruction is received, determining a defect correction deviation based on a preset database and the extrusion operation parameter group sequence of the second extrusion material, and updating preset standard operation regulation and control parameter information and / or sending the defect correction deviation to the user terminal based on the defect correction deviation so as to control the three-way extrusion device to produce.

The utility model discloses a feeding and discharging structure for steel pipes, which comprises a discharging box and a conveying belt, a stepping motor is adopted by a conveying belt driving structure, the conveying belt is supported by a rack, the conveying belt is located at the middle end of the rack, the stepping motor is located at the bottom end of the rack, a material guiding bin is arranged at the top end of the discharging box, and the upper end of the material guiding bin is funnel-shaped. The diameter of the through groove in the middle end of the guide bin is matched with the outer diameter of the steel pipe, positioning blocks used for separating the steel pipe are arranged on the outer surface of the conveying belt, a mounting plate is arranged on the outer surface of one end of the discharging box, and an air cylinder is mounted in the mounting plate. According to the feeding and discharging structure for the steel pipes, the stepping motor is adopted to drive the conveying belt, automatic feeding and discharging of the steel pipes are achieved, the production efficiency is improved, the labor cost is reduced, the industrial injury risk caused by manual operation is reduced, and the distance between the steel pipes can be effectively controlled through the positioning blocks arranged on the conveying belt; and the stability and safety of the steel pipes in the conveying process are guaranteed.

The present invention relates to a preform rod bending detection device, correction system and correction method, the device comprises: a first detection assembly comprising a first laser emitter and a first laser receiver, the first laser emitter emits a first light beam along a first linear direction, the first laser receiver is used for receiving the first light beam, and the first laser receiver is used for receiving a second light beam along a second linear direction; the first laser transmitter and the first laser receiver are arranged on the two sides of the reference axis in the radial direction, and the reference axis is perpendicular to the first linear direction and the second linear direction and located in a light path of the first light beam; the second detection assembly comprises a second laser emitter and a second laser receiver, the second laser emitter is used for emitting a second light beam in the second linear direction, the second laser receiver is used for receiving the second light beam, and the second laser emitter and the second laser receiver are arranged on the two sides of the reference axis in the radial direction; the reference axis is located in a light path of the second light beam. According to the invention, accurate detection of the bending direction and the bending degree of the handle bar to be detected can be realized.

The invention discloses a radiotherapy equipment alignment precision detection device which comprises a beam output component and a light limiting cylinder which is arranged below the beam output component and needs to be aligned, the beam output end face of the beam output component is provided with at least three laser ranging components, and the at least three laser ranging components are distributed on a circle coaxial with the beam output component; a laser detection part is arranged on the mounting disc at the top of the light limiting cylinder, and is used for detecting the irradiation position of a laser beam of the laser ranging part; the invention also discloses a radiotherapy equipment alignment precision detection method. The automatic high-precision alignment of the radiotherapy equipment can be realized.

The invention discloses a blowing and sucking type automatic yarn doubling and plying yarn supply device. The device comprises a rotary supporting module and a plurality of spindle supporting modules arranged on the rotary supporting module in a surrounding mode. The thread end blowing modules are arranged corresponding to the spindle supporting modules and blow air around yarn rolls on the spindle supporting modules; and the thread end sucking and yarn gathering module is located above the rotary supporting module and concentrates thread ends blown out by the thread end blowing modules together and blows the thread ends upwards. Automatic feeding of yarn rolls, automatic finding of initial thread ends and automatic gathering operation can be achieved, and the production efficiency of the doubling process is greatly improved.