42 results about "Curve fitting algorithm" patented technology

The invention discloses a thyroid cancer analysis method and device based on real-time fluorescent quantitative PCR. The method comprises the following steps: acquiring to-be-analyzed data; determining sample exclusive threshold line data according to the to-be-analyzed data; calculating a corrected Ct value based on double inflection point dynamics according to the exclusive threshold line data of the sample so as to obtain a final corrected Ct value; and generating a detection result according to the final corrected Ct value. Through optimization processing of the curve fitting algorithm, the linear characteristic of the fluorescence amplification curve is remarkably improved, the interference of background noise on the result is reduced, and the calculation stability of the Ct value is improved.

Some embodiments of the specification provide a welding method based on coordinate system registration, the method comprising: acquiring, by a sensor, a sequence of spatial coordinate points of a teaching welding path and teaching point cloud data of a standard workpiece; generating, by a curve fitting algorithm, a standard welding path based on the sequence of spatial coordinate points; determining a reference point based on the teaching point cloud data and the sequence of spatial coordinate points; constructing a welding database based on a first workpiece identifier of the standard workpiece, the reference point, and the standard welding path; acquiring a second workpiece identifier and coarse scanning point cloud data of a workpiece to be welded; acquiring, from the welding database, the teaching point cloud data, the reference point, and the standard welding path corresponding to the second workpiece identifier; determining a first registration transformation based on the teaching point cloud data and the coarse scanning point cloud data; determining a scanning range based on the reference point and the first registration transformation; acquiring fine scanning point cloud data based on the scanning range; and determining a welding instruction according to the fine scanning point cloud data and the standard welding path.

This application discloses a temperature curve fitting algorithm, including the following steps: (1) preheating temperature control and feature extraction; (2) constant power curve fitting; (3) volume estimation; (4) anomaly handling. The temperature curve fitting algorithm of this application stabilizes to 50℃±3℃ within 10 minutes, and mines the temperature data features of the preheating stage through sliding residual trend analysis, capturing the weak heat capacity influence of small-volume insects in advance, and providing a sensitivity marker for the slope extraction in the subsequent constant power stage; the fitting mode is automatically switched by the slope reduction threshold, which not only retains the accurate extraction of the core slope in the stable segment, but also avoids the fitting distortion in the slowing segment; the fitting benchmark is dynamically corrected by h, rather than relying on a fixed insect-free curve, which effectively offsets the heat dissipation difference interference caused by room temperature fluctuations; when the heat dissipation coefficient estimation is abnormal, it is automatically downgraded to a linear fitting benchmark, and when historical data is insufficient, a backup formula is called to ensure that effective results can still be output under extreme scenarios such as sensor fluctuations and data loss.

The application discloses a curve segmentation fitting method and device for image segmentation and a storage medium, and comprises the following steps: pre-processing an image to obtain contour data; performing a contour segmentation algorithm of a DCE evaluation standard according to the contour data to obtain a segmentation point list; and performing a curve fitting algorithm based on an Euler curve model according to the segmentation list to obtain a final fitting result. The automatic segmentation of the contour can be realized, and subjective errors caused by human intervention are reduced.

This invention discloses a panoramic image monitoring system, comprising: an image acquisition module for acquiring visual image data through an ultra-wide-angle image acquisition device with at least four sets of ring-shaped first devices; a data processing module for sequentially correcting and stitching the acquired image data to obtain panoramic image data, then performing target detection, recognition, and tracking operations on the processed image, and employing linear interpolation or curve fitting algorithms to improve the detection and recognition accuracy of the target; and a display module for displaying the processed image data and the detection, recognition, and tracking results of the target. This invention solves the problem of potential obstruction of the camera's field of view or severe degradation of image quality, effectively improving the system's stability and reliability. It also enables precise tracking of object motion, improving the accuracy and continuity of dynamic target tracking, and facilitating effective handling of situations where multiple targets move simultaneously or change rapidly.

The application belongs to the technical field of path planning, and particularly relates to a PCB glue tearing machine operation path optimization control method and system, which comprises the following steps: collecting torque and motion state data of a glue tearing machine actuator, performing gravity compensation and filtering on the torque data; calculating accumulated mechanical work according to the processed torque data; obtaining a sticking resistance coefficient according to the accumulated mechanical work, displacement distance and torque change rate; when a continuous preset number of sticking resistance coefficients exceeds a threshold value, sampling control points on an initial path, taking the control point longitudinal coordinate as an optimization variable; constructing a cost function comprising a safety penalty term, a path smoothness penalty term and an efficiency reward term, performing particle swarm optimization on the control points according to the cost function to obtain optimal control points, and generating an avoidance path using a curve fitting algorithm on the optimal control points; and controlling the actuator to complete glue tearing according to the avoidance path. The application adjusts the path by monitoring the sticking state, thereby improving the glue tearing quality.

The invention relates to the technical field of atomizer control, in particular to an intelligent atomization control method and system for an atomizer, and the method comprises the steps: obtaining parameters of an equivalent model of a piezoelectric transducer; analyzing parameter fluctuation characteristics of the piezoelectric transducer under different external conditions according to parameters of the equivalent model; constructing an overall equivalent circuit of the atomizer, wherein the overall equivalent circuit comprises an LC filter and a piezoelectric transducer at the output end of an inverter; applying one or more pulse voltages to the overall equivalent circuit to trigger a transient phenomenon, measuring a transient current in the circuit, and determining the overall resonant frequency; based on the nonlinear correlation between the whole resonant frequency and the mechanical resonant frequency of the piezoelectric transducer, fitting related data by using a curve fitting algorithm; and according to the result of the curve fitting algorithm, calculating and tracking the mechanical resonant frequency of the piezoelectric transducer, and applying the tracked mechanical resonant frequency to the atomization system. The problems of slow tracking and low efficiency caused by fluctuation of the resonant frequency of the piezoelectric transducer in an existing atomization system due to external conditions can be solved.

The invention relates to the technical field of image recognition processing, and discloses a direct-injection English read-through card classification method based on image recognition processing, and the method comprises the steps: obtaining a damaged English read-through card image and a historical learning record of a user, recognizing a damaged region, and extracting the features of an individual forgetting mode; calculating a memory attenuation degree and a credibility coefficient based on an improved Ebbinghaus forgetting curve fitting algorithm; utilizing a memory guide content reconstruction algorithm, and taking the memory credibility as a constraint weight to adjust an image restoration parameter; semantic probability distribution is calculated based on the vocabulary features, and restoration candidate schemes are generated in combination with memory constraints; analyzing the matching degree between the candidate scheme and the user memory state by using a time sequence memory verification algorithm; verifying the semantic rationality of the repaired content based on a language model; and performing content identification on the restored image to generate a classification result of comprehensive restoration and time sequence optimization. According to the method, intelligent image restoration of semantic constraint is realized, and integrated processing of restoration and classification is realized.

The present application relates to a kind of plane short weld seam identification method, welding method and device, belong to the field of automated welding;Through the two-step photographing method of single camera autonomous welding model, accurately identify the three-dimensional coordinates of spatial point under the robot base coordinate system, give the autonomy of robot welding non-standard plane weld or batch plane weld;YOLOv5 neural network is used to locate the arc starting point, arc extinguishing point, workpiece corner point and weld area, improve the perception ability of robot to space target;Through brightness and contrast enhancement algorithm, edge detection algorithm, morphological processing algorithm, accurately extract weld point cloud from weld area;Through stereo matching algorithm and B spline curve fitting algorithm, identify and fit welding trajectory from irresistible, dynamic, irregular interference information;The fitted welding trajectory is transmitted to robot system to carry out autonomous welding on plane weld.

The invention provides an air conditioner which comprises an indoor unit, a heat exchange air duct and a fresh air duct are arranged in the indoor unit; the fresh air fan is arranged in the fresh air duct; the purification module is arranged in the heat exchange air duct; the controller is configured to calculate the range of the pollutant concentration sensed by the human body in the current sampling period by adopting an index curve fitting algorithm on the basis of the pollutant concentration in the current sampling period, the reference concentration and the upper limit value of the pollutant concentration in the previous sampling period; linearly converting the range of the human body sensing concentration into a preset numerical value range, wherein numerical values in the preset numerical value range are represented by s; dividing the preset numerical range into a plurality of continuous concentration intervals according to the boundary concentration; and according to the concentration interval where the current numerical value s is located, the fresh air fan and the purification module are controlled to operate according to the preset control logic. In the invention, the fresh air fan and the purification module are gradually adjusted according to the change of the human body perceived concentration, the intelligent level is high, and the user experience is favorably improved.

A kind of machining datum self-positioning method of additive repair bearing bush, comprising the following steps: S1. obtaining point cloud data of bearing bush surface;S2. preprocessing point cloud data;S3. point cloud registration aligns bearing bush surface with point cloud model;S4. after surface reconstruction, obtain feature points of bearing bush surface;S5. based on point cloud data, accurately extract feature points, and then calculate the geometric center of bearing bush;S6. determine the machining axis of bearing bush by curve fitting algorithm;S7. real-time acquisition of bearing bush positioning data determines machining datum and axis parameters, and automatically adjusts the position and parameters of machining equipment through compensation logic and three-dimensional reconstruction algorithm, so that the machining equipment matches the actual shape and position of bearing bush, so that the bearing bush is placed on the machining platform, the computer can automatically identify and determine the machining datum, axis and other parameters, and the position of machining equipment can be adjusted in real time through compensation method to match the bearing bush, improve the machining efficiency and machining precision, and have good practical value and economic value.

The application relates to a combined positioning method of a vehicle-mounted platform and a UAV, which comprises the following steps: acquiring LiDAR point cloud and vehicle coordinates from the vehicle-mounted platform, acquiring UAV coordinates and a first relative pose between the vehicle and the UAV from the UAV, generating a semantic map according to the LiDAR point cloud based on a parameterized curve fitting algorithm, determining accurate coordinates of the UAV according to the vehicle coordinates and the first relative pose, sending the semantic map to the UAV to instruct the UAV to shoot image data according to lane lines in the semantic map, obtaining a second relative pose between the UAV and the vehicle-mounted platform according to the image data and the semantic map, and determining accurate coordinates of the vehicle based on the second relative pose and the accurate coordinates of the UAV. Through the application, the problem that the vehicle and the UAV are not accurately positioned is solved, the bidirectional correction reduces the cumulative error, and the lightweight semantic map reduces the transmission bandwidth occupation while maintaining a high recognition rate.

The invention relates to a vehicle-mounted platform and unmanned aerial vehicle combined positioning method, and the method comprises the steps: obtaining a LiDAR point cloud and a vehicle coordinate from a vehicle-mounted platform, obtaining an unmanned aerial vehicle coordinate and a first relative pose between a vehicle and an unmanned aerial vehicle from an unmanned aerial vehicle, generating a semantic map according to the LiDAR point cloud based on a parameterized curve fitting algorithm, determining accurate coordinates of the unmanned aerial vehicle according to the vehicle coordinates and the first relative pose, sending a semantic map to the unmanned aerial vehicle to instruct the unmanned aerial vehicle to shoot according to lane lines in the semantic map to obtain image data, obtaining a second relative pose of the unmanned aerial vehicle and the vehicle-mounted platform according to the image data and the semantic map, and sending the second relative pose to the vehicle-mounted platform. And determining accurate coordinates of the vehicle based on the second relative pose and the accurate coordinates of the unmanned aerial vehicle. Through the method and the device, the problem of inaccurate positioning of the vehicle and the unmanned aerial vehicle is solved, accumulated errors are reduced through bidirectional correction, and the transmission bandwidth occupation is reduced while the lightweight semantic map keeps a high recognition rate.

The application discloses a flight pipeline modeling method for cluster unmanned aerial vehicle control, and establishes a geometric pipeline model of unmanned aerial vehicle flight; a profile model of a convex polygon section is established based on rotation angles and lengths of convex vectors, and the convex polygon section is composed of a plurality of convex vectors; nodes are constructed by positions of the convex polygon section on a reference curve, the rotation angles and the lengths of the convex vectors, the geometric pipeline is discretized into a plurality of node sequences, a curve fitting algorithm is applied to the node sequences to generate a continuous three-dimensional pipeline model; a comprehensive cost function of the geometric pipeline is established according to characteristics of the geometric pipeline; and the optimal geometric flight pipeline model is obtained when the comprehensive cost function is the minimum. The flight pipeline is modeled by the reference curve and the convex polygon section, a polygon geometric pipeline is obtained, and safe flight space modeling in a complex three-dimensional scene is realized.

The present application belongs to the technical field of digital measurement, and particularly relates to a high-precision three-dimensional measurement method for local features of an aircraft shape. First, the imaging center of a camera is aligned with a region to be measured on the surface of the aircraft, and the longitudinal position of a scanning system is adjusted so that the focal plane of the camera is located at an effective scanning position. Then, the included angle between the measurement light path and the surface to be measured is adjusted so that the number of imaging fringe periods is less than 1. The scanning system is controlled to move in a single direction to achieve longitudinal scanning of the surface to be measured. After scanning is completed, the longitudinal light intensity curve is extracted for each pixel point, and the Fourier transform technique and Gaussian curve fitting algorithm are further combined to obtain the accurate focusing position of the pixel point. Finally, height mapping is performed in combination with the scanning step distance to restore the three-dimensional topography of the region to be measured. The technical solution can improve the measurement accuracy, support three-dimensional measurement of small local features of an aircraft shape, and has the characteristics of high demodulation efficiency and wide adaptability.

Embodiments estimate the trajectory of an airborne or aquatic object. Embodiments comprise obtaining (204) track data of an airborne or aquatic object; fitting (210) at least one motion model to the track data using a curve- fitting algorithm to determine parameter values of the at least one motion model, and estimating (210) a trajectory of the object using the parameter values. The at least one motion model estimates the trajectory of the object as a plurality of linearly separable sections.

This invention discloses a novel continuum robot and its shape prediction method. The continuum robot includes a handle body, a snake skeleton, a drive device, and two drive cables. The snake skeleton is fixed to the front end of the handle body, and the drive device is located at the rear end inside the handle body. The two drive cables are symmetrically arranged axially within cable constraint grooves on both sides of the snake skeleton. The drive device controls the deformation of the snake skeleton through the drive cables. The snake skeleton includes several sequentially adjacent guide joints. Each guide joint includes a flexible hinge in the middle and a nested ring structure on each of the left and right sides that rotates around the flexible hinge. The shape prediction method utilizes the real-time collected changes in the length of the drive cables to obtain the predicted coordinates of several control points of the snake skeleton through a trained prediction model. The shape curve of the snake skeleton is reconstructed using a spline curve fitting algorithm. The continuum robot of this invention possesses good bending consistency, torsional resistance, and resistance to local disturbances, and exhibits high shape prediction accuracy.

The invention discloses a main transformer differential protection automatic calculation method and system, and belongs to the technical field of power system relay protection. Traditional main transformer differential protection calculation depends on manpower, is high in complexity, is easily influenced by parameter errors and is low in efficiency. Multi-source parameters such as a main transformer nameplate parameter, a CT / PT transformation ratio and a winding wiring mode are automatically acquired and verified through the parameter analysis module; self-adaptively calculating the balance coefficient of each side by using a dynamic balance coefficient calculation algorithm; a phase compensation matrix is automatically generated based on the wiring mode, and compensation effect verification is carried out; characteristic inflection points of the proportional braking characteristic curve are automatically searched through a six-phase current synchronous injection method, and an optimized proportional braking characteristic curve is generated through a curve fitting algorithm; and finally, carrying out optimization recommendation and dynamic adjustment on the differential fixed value by means of an AI auxiliary decision-making module. According to the method, full-process automation from parameter input to constant value generation is realized, the calculation efficiency and precision are remarkably improved, and rapid and accurate setting calculation support is provided for differential protection of the main transformer.

An apparatus, system, and method for compensating for bone loss of a patient's knee joint during the performance of an orthopaedic surgical procedure includes determining an estimated amount of bone loss of a bone of a patient's knee joint, determining an anatomical distance between a femoral anatomical feature of a patient's femur and a tibial anatomical feature of a patient's tibia, and adjusting the determined anatomical distance based on the estimated amount of bone loss. The determined anatomical distance is indicative of a ligament length of a collateral ligament of the patient's knee joint. The bone loss may be estimated based on, for example, a surface curve fitting algorithm applied to a three-dimensional model of the bone, a defined bone loss amount provided by a user, or a comparison of the three-dimensional model of the bone to a statistical shape model that approximates the bone in a healthy state.

The application provides a laser radar aliasing waveform decomposition method based on waveform characteristics, constructs a waveform mathematical model according to seed light characteristics of a laser radar, combines an LM curve fitting algorithm, decomposes aliasing echoes into multiple single echo data, and achieves the purpose of accurately extracting multi-target ranging information. The technical basis of the application is a laser radar full waveform detection technology, that is, a laser radar detection system can sample analog electric signals converted from optical signals into digital waveform signals, and the processing object of the application is a waveform sampling point sequence of seed light and return light detected by the laser radar.

The invention relates to the technical field of image processing, in particular to a coastline automatic extraction method and system based on deep learning and satellite images. Comprising the steps of obtaining multi-temporal high-resolution image data of a target coast, and performing preprocessing to obtain an enhanced image sequence; extracting a connected component of the salinization erosion area, and generating a refined boundary pixel set; a distance field is generated, the minimum distance from each boundary pixel to the nearest boundary is calculated, an initial center point sequence is generated, and a spline curve fitting algorithm is adopted for optimization; generating a polygon buffer area, and performing superposition verification on the polygon buffer area and the enhanced image sequence to obtain a verified central point sequence; a continuous vector path is generated through fitting, coastline vector data are extracted, multi-temporal change monitoring is carried out, and high-salinity soil accumulation and erosion rate indexes are obtained; and updating the pixel-level classification to generate a final coastline vector data sequence. According to the method, the salinization erosion coastline can be efficiently and accurately extracted, and reliable data support is provided for ecological environment monitoring.

Aiming at the problems of wear and blockage in lane line detection, a lane line robust detection and extraction method based on road constraint is proposed. Firstly, a large number of Dongfeng park scene images are collected to construct a lane line dataset, and the road area is labeled while labeling the lane line. Secondly, the ENet-SAD network model is improved, and the lane line segmentation loss function of joint road constraint is defined to generate accurate lane line segmentation prediction results. Finally, the full connection conditional random field algorithm is introduced to post-process the output lane line probability map, correct the probability map, select N offline coordinate points on each lane line, and use the cubic spline curve fitting algorithm to obtain the fitted lane line. The experimental results show that compared with the ENet-SAD method, the method in this paper achieves good results in different scenes such as conventional, strong light, shadow, blockage, etc. At the same time, based on the loss function defined in this paper, the performance of existing lane line detection models (such as ENet-SAD, SCNN, etc.) can be further improved.

The invention provides a mushroom stipe phenotypic feature recognition method and device, and relates to the technical field of image recognition, and the method comprises the steps: obtaining an original image containing a to-be-detected mushroom stipe; the original image is input into a semantic segmentation network, a binarization mask of a stipe area is obtained, and the semantic segmentation network integrates a triple attention mechanism; the original image is input to a key point detection network, position coordinates of a plurality of key points distributed in the axial direction of the stipe are obtained, and the key point detection network integrates a spatial pyramid attention mechanism; based on the position coordinates of the plurality of key points, generating a stipe center line through a curve fitting algorithm; and in combination with the binarized mask, the stipe center line and the original image, calculating morphological features, axial geometric features and color features of the mushroom stipe, and obtaining a multi-dimensional feature vector of the mushroom stipe.

This application discloses a method and apparatus for thyroid cancer analysis based on real-time quantitative PCR. The method includes: acquiring data to be analyzed; determining sample-specific threshold line data based on the data to be analyzed; calculating a corrected Ct value based on dual-inflection-point kinetics based on the sample-specific threshold line data to obtain the final corrected Ct value; and generating a detection result based on the final corrected Ct value. This application significantly improves the linearity of the fluorescence amplification curve, reduces the interference of background noise on the results, and improves the stability of Ct value calculation through optimized curve fitting algorithm processing.

The invention discloses a hazardous chemical substance storage environment data analysis system, and relates to the technical field of hazardous chemical substance safety management, and the system comprises a photocatalytic reaction spectrum acquisition module, an environmental parameter sensing and transmission module, a multi-source data preprocessing module, a self-purification capability dynamic evaluation module and an intelligent early warning linkage control module. A composite photocatalyst coating is deployed by adopting a high-pressure airless spraying process, a photocatalytic reaction characteristic spectrum is monitored in real time in combination with an optical fiber spectrometer, meanwhile, multiple types of sensors are integrated to construct a three-dimensional sensing network, and a multi-source data preprocessing module corrects humidity parameters through a temperature compensation quadratic curve fitting algorithm; the time stamp alignment technology is combined, the data space-time consistency is ensured, the self-cleaning capacity dynamic evaluation module proposes three core indexes of real-time degradation efficiency, risk accumulation indexes and self-cleaning completion time, the influences of coating attenuation and environment variables on the degradation effect are dynamically reflected through a quantification formula, and a scientific early warning basis is provided for storage safety.

The invention provides a weak optical power detection method and detection circuit. The method comprises the following steps: acquiring a plurality of voltage digital quantities of an optical signal to be detected and a temperature digital quantity corresponding to a current environment temperature; based on an optical power-voltage-temperature mapping function, a measured optical power value is calculated according to the average value of the plurality of voltage digital values and the temperature digital value; obtaining an equivalent optical power value deviation by combining the independent component of each noise in the optical signal to be measured; and correcting the measured optical power value based on the equivalent optical power value deviation to obtain an optical power value to be measured. A multi-point calibration and curve fitting algorithm is adopted to construct an optical power-voltage-temperature nonlinear mapping model, device nonlinearity and temperature excursion are effectively compensated, and weak light detection precision and environmental adaptability are improved; meanwhile, by combining the independent component of each noise in the to-be-detected optical signal, the equivalent optical power value deviation is obtained, sub-item identification and real-time compensation of various noise components are realized, and the weak optical signal detection precision is improved.

The invention provides an energy storage system insulation resistance self-adaptive detection method and device and a medium, and the method comprises the steps: building an unbalanced bridge insulation resistance detection circuit, switching the state of a detection switch to access different detection elements, building an unbalanced state between a positive bus and a negative bus and the ground, and executing a voltage detection sub-process. The output voltage Vout is obtained through sectional threshold detection, and the insulation resistance of the positive electrode and the negative electrode to the ground is calculated according to the output voltage Vout obtained through the circuit and the voltage detection sub-process. According to the method, the self-adaptive judgment logic is introduced, so that monitoring delay caused by large capacitor charging delay in a traditional method is avoided, and the insulation state information can be quickly obtained; by adopting an RC charging curve fitting algorithm based on historical sampling data, the collected voltage data is corrected and compensated, errors caused by charging and discharging effects and random interference are reduced, and thus a more accurate insulation detection result is obtained.

The invention discloses a vehicle body special-shaped space curve welding seam tracking method and system. The method comprises the steps that target workpiece image information is collected through a preset scanning track, a welding seam point position center line is marked, and then a welding seam data set is formed; enhancing the weld data set, and setting a reflective area as an inverted funnel-shaped highlight distribution structure for data enhancement of reflective characteristics to obtain an initial weld data set; constructing a welding seam point location recognition model to recognize the initial welding seam data set to obtain initial welding seam point locations, filtering the initial welding seam point locations to remove outliers, and generating a reference welding seam point location sequence by using a curve fitting algorithm; constructing a curve welding seam tracking model; during welding, real-time data of a welding seam are synchronously collected, the actual pose of the welding seam is recognized through a curve welding seam tracking model, and a control instruction is generated in combination with a reference welding seam point position sequence to track welding in real time. According to the method, high-precision real-time identification and tracking of the complex space curve welding seam can be realized on the premise that the production takt is not increased.

The application discloses a novel continuum robot and a shape prediction method thereof, wherein the continuum robot comprises a handle main body, a snake bone, a driving device and two driving cables; the snake bone is fixed to the front end of the handle main body, the driving device is arranged at the tail end inside the handle main body, the two driving cables are symmetrically arranged in the cable constraint grooves on the two side walls of the snake bone along the axial direction, and the driving device controls the deformation of the snake bone through the driving cables; the snake bone comprises a plurality of guide joints which are sequentially adjacent; each guide joint comprises a flexible hinge in the middle and one nested ring buckle structure on the left and right sides which rotates around the flexible hinge. The shape prediction method uses the real-time collected length change of the driving cable, obtains the predicted coordinates of a plurality of control points of the snake bone through a trained prediction model, and reconstructs the shape curve of the snake bone through a spline curve fitting algorithm. The continuum robot has good bending consistency, torsion resistance and local interference resistance, and has high shape prediction precision.

The invention discloses a compensation milling method for curved surface boundary reconstruction, and belongs to the field of numerical control precision machining of complex curved surface parts. The method aims to solve the problems of low efficiency, unstable quality and easy out-of-tolerance caused by the fact that tool connecting traces generated after complex curved surface milling depend on manual polishing and repairing in the prior art. The method comprises the steps that firstly, a small plane body model reflecting the actual tool mark appearance is obtained through machining simulation; secondly, intersecting the small plane body model with the theoretical curved surface model to obtain a discrete initial intersection line; then, representative points are selected from the discrete lines, and a continuous and smooth curve is reconstructed through a curve fitting algorithm to serve as a compensation boundary; and finally, an offset finish machining tool path is generated based on the boundary, and the butt-joint tool mark area is milled at a time. According to the method, the problems possibly caused by manual grinding are avoided, and the fairness and geometric accuracy of the curved surface switching area are guaranteed.