110results about How to "Strong tracking ability" patented technology

The invention discloses an aviation direct-current converter online fault combined prediction method based on fractional order wavelet transformation. The method includes: (1) monitoring and collecting output voltage signals of the aviation direct-current converter in real time, calculating output voltage change rate of different moments, and using the output voltage change rate as converter performance degradation parameters; (2) conducting abnormal value rejection and missing data filling on performance degradation data by using a 3 sigma method and an interpolation method; (3) conducting fractional order wavelet transformation on the performance degradation data, the performance degradation data is decomposed into subcomponents with different scales, and determining noise components and removing the noise components by calculating a combination entropy between high-frequency components and environment data; (4) building a predication model of the high-frequency components in decomposition data by using a wavelet neural network, building a prediction model of low-frequency components by using a gray neural network, and conducting time sequence prediction; and (5) stacking predication values of the high-frequency components and the lower-frequency components to obtain a final predication value, conducting performance evaluation and fault predication on the aviation direct-current converter by combining fault threshold values. The aviation direct-current converter online fault combined prediction method removes disturbances caused by environment factor fluctuation in performance degradation data, restores real performance degradation data, simultaneously decomposes the performance degradation data into different frequency subcomponents with strong regularity, predicts the subcomponents by using a combined prediction model, enables prediction risks to be dispersed, and improves online fault prediction correctness.

A target detecting and tracking system using a background differencing method based on FPGA comprises a FPGA device, a SD card, an LED indicator light, a DDR2SDRAM, two seven-segment digital tubes, an LTM display screen and a minimum video detecting and tracking system, wherein the minimum video detecting and tracking system is embedded in the FPGA and is integrated in a SOPCbuilder environment with an AVLON bus as a standard. The target detecting and tracking system using the background differencing method based on FPGA is mainly used for detecting and tracking moving objects in static scenes. A target detecting and tracking method using the background differencing method based on FPGA comprises the steps of a. reading AVI format video files with the SD card, b. converting the AVI format video files into multiframe images, c. detecting and tracking the videos based on the images, d. obtaining the moving objects which are to be tracked, and e. displaying tracking results on the LTM display screen. Due to the fact that an FPGA platform is adopted by the system, the system is high in parallelism performance, and computing speed is improved. The target detecting and tracking method is good in tracking effect and adaptability, particle filter algorithm which is suitable for non-linear non-Gaussian dynamic models, and the target detecting and tracking system is similar with real scenes and is good in adaptability and high in tracking accuracy.

The invention discloses an online calibrating method of a ship-based rotary strapdown inertial navigation system. The method comprises the following steps: establishing an inertial component output error model and an inertial navigation system error equation, and researching the calibration of inertial component parameter errors and determining the quantity of state and the quantity of measuration; determining the position and weight of a cubature point according to dimension of the quantity of state, deducing a state equation and a one-step state prediction and state prediction covariance matrix related to the cubature point, and introducing a multiple time-varying fading factor modified state prediction covariance matrix; and deducing a measuring equation related to the cubature point and the fading factors, a self-correlated covariance matrix, a cross-correlated covariance matrix, a gain matrix, a state estimated value and a state error covariance estimated value, and designing a strong tracking volume Kalman filtering method with strong tracking performance and strong robustness. The method disclosed by the invention estimates the inertial component parameter errors by a filtering algorithm and carries out online calibration and compensates the inertial component parameter errors, so that the navigation precision is effectively improved. The method has strong parameter-varying robustness.

The invention provides an under-actuated AUV (autonomous underwater vehicle) three-dimensional trajectory tracking control method based on biological speed regulation. The method comprises the following steps: (1) giving expected trajectory position information by an AUV according to the current task, and obtaining current position and attitude information; (2) obtaining position and attitude error variables by use of the mathematical model of an under-actuated UUV (unmanned underwater vehicle); (3) calculating a virtual control law by a method of defining a virtual speed error variable; (4) finishing dynamic regulation of the speed error through a biological inspiration model; and (5) deducing a dynamic speed regulation controller. In the method provided by the invention, the speed error of the under-actuated AUV can be dynamically regulated, the performance of the controller is improved while the singular value of the traditional back-stepping method occurring when the heading angle error is 90 degrees is avoided, and accurate tracking of a time-varying three-dimensional trajectory is realized under external constant disturbance.

Provided is a dynamic evolution modeling method for aluminum electrolysis process electrolytic bath technology energy consumption. The method is characterized by including the following steps of step 1, collecting data [XN, Y], step 2, carrying out normalization processing on the collected data, step 3, carrying out modeling on the data after the normalization processing by strongly tracking a square root trackless Kalman neural network, and step 4, estimating an electrolysis process energy consumption value by applying an established model to obtain a technology energy consumption value of the electrolysis process at the moment. The method has the advantages that advantages of strong tracking filtering and square root filtering are combined, convergence rates of the model and tracking ability on electrolytic bath mutation states are improved, the algorithm is stable, accuracy is high, tracking ability on the electrolytic bath mutation states is strong, therefore, real time estimation on the aluminum electrolysis process electrolytic bath technology energy consumption is achieved, technology operations on the aluminum electrolysis process can be optimized, and the purposes of saving energy and reducing emission can be achieved.

The invention discloses a detection method for a variable step length LMS (Least Mean Square) harmonic current based on versiera. The method comprises the steps of: sampling, estimating an active current of a fundamental wave, estimating a general harmonic current, updating a weight coefficient, updating the step length and iterating. According to the orthogonal characteristics of a trigonometric function, the current error output by a filter and an instantaneous cross-correlation time mean value are subtracted from a load current to eliminate interference of the harmonic current and the active current of the fundamental wave on update of the step length. Furthermore, update of the step length is controlled by an improved versiera function. Meanwhile, in the updating process of the weight coefficient, variation of the weight coefficient is imported so that the steady state error is very small and the calculated amount is smaller. The detection method for the variable step length LMS harmonic current based on versiera provided by the invention is suitable for occasion where the load current has saltus and value of the optimal parameter provided by the invention is less dependent on the load current.

The invention discloses a high energy-consuming enterprise peak regulation control method used in a new energy power generation mode. The method comprises the step of establishing a high energy-consuming load peak regulation model in the new energy power generation mode and the step of conducting analysis of multi-mode and multi-scale high energy-consuming load participation peak regulation on the basis of the established high energy-consuming load peak regulation model in the new energy power generation mode. The high energy-consuming enterprise peak regulation control method used in the new energy power generation mode can overcome the detects that in the prior art, peak regulation pressure is high, tracking capacity is weak and the energy saving effect is poor, thereby having the advantages of being low in peak regulation pressure, strong in tracking capacity and good in energy saving effect.

The invention discloses a novel lithium ion power cell SOC estimation method. The novel lithium ion power cell SOC estimation method is characterized in that a cell equivalent circuit model can be established, and the identification of the parameters of the established cell model can be carried out by adopting the least squares algorithm; according to the cell open-circuit voltage UOCV acquired by identifying the parameters according to the step one and the corresponding SOC relation, a corresponding function can be acquired by adopting a Shepherd model and a Nernst model in a combined manner, and the function is used for the matching of the UOCV and the SOC relation; the state equation and the observation equation of the SOC estimation can be established, and the STF algorithm has the strong robustness related to the model uncertainty and the extremely strong tracking capability related to the breaking state. The verification of the SOC estimation by adopting the EKF algorithm and the STF algorithm can be carried out by the constant current discharging experiment and the UDDS working condition experiment, and according to the result, the accuracy of the SOC estimation is higher by adopting the STF algorithm than adopting the EKF algorithm, and the convergence performance is better.

The invention discloses an unattended satellite ground station tracking control system. The system is suitable for task execution of satellite constellation satellite-ground communication, and specifically comprises an orbit prediction module, a satellite-ground azimuth angle calculation module, a servo control module, an antenna automatic zeroing module and a fault detection module. The orbit data of a satellite in the target constellation can be automatically updated and forecasted in real time, the antenna pointing is automatically controlled to realize the optimal satellite-ground communication gain, the antenna pointing can be automatically returned to zero after the task is finished, and the real-time antenna pointing monitoring can be performed during the task proceeding period. According to the invention, the traditional working mode of manually operating the ground station is changed, the communication effect is optimal, and the reliability is high.

The invention provides an intelligent medicine cabinet system and relates to the field of intelligent medicine cabinets. The intelligent medicine cabinet system is achieved based on an intelligent medicine cabinet touch screen and an intelligent medicine cabinet lock and comprises a bar code reader, an upper computer module and a lower computer module; the intelligent medicine cabinet touch screen is used for displaying and inputting information; the bar code reader is used for reading identity data information and sending the data information to the upper computer module; the upper computer module is used for performing identity authentication and sending control instruction data, wherein the identity authentication is achieved through a medicine storage location matching library and an identity matching library which are arranged in the upper computer module; the lower computer module is used for receiving the control instruction sent out of the upper computer module and driving the intelligent medicine cabinet lock and accordingly opening, closing and early-warning of the intelligent medicine cabinet lock can be performed. The intelligent medicine cabinet system is applied to the medicine field specifically. According to the intelligent medicine cabinet system, the problems that the service efficiency is low, the security is poor, the medicine storage is lax, and the medicine information cannot be tracked timely in the existing process of traditional population register medicine storage and fetch management are solved.

An improved convex combination decorrelation proportionate self-adaption echo cancellation method comprises the steps that first, far-end signal filtering is carried out, the input vector X(n) of a convex combination self-adaption echo cancellation filter is formed by the discrete value of a far-end signal, and after filtering is carried out on the input vector, a large-step-length filtering value y1(n) and a small-step-length filtering value y2(n) are obtained; second, decorrelation operation is carried out on the input vector X(n), the result of decorrelation operation serves as the weight coefficient updating direction vector Z(n) of the convex combination self-adaption echo cancellation filter; third, convex combination is carried out, the large-step-length filtering value y1(n) and the small-step-length filtering value y2(n) are subjected to convex combination through weight lambda (n), and a combination filter value y(n) is obtained; fourth, echo cancellation is carried out, the combination filter value y(n) is subtracted from a near-end signal d(n) with an echo, and the subtracted near-end signal d(n) is fed back to the far end; fifth, a filter tap weight coefficient is updated; sixth, the weight of the filter is updated; seventh, the weight of the filter is limited; eighth, n is made to be equal to n+1, the first step to the seventh step are repeated till a conversation is over. According to the method, a high rate of convergence can be obtained, the steady state error can be small, and a good anti-jamming capability is achieved.

Disclosed is an active factor set membership proportional sub band self-adaption echo cancellation method. The method comprises the steps: A, dividing an adaptive echo cancellation sub band filter input vector X(n) formed by a discrete value of a far-end signal into a sub band signal X(n), B, performing N extraction on an input sub band signal X(n) and a near-end sub band signal d(n) to obtain a signal X(k) after the extraction and a d(k), C, obtaining a filtering value y(k) of the signal X(k) after the extraction via an adaptive echo cancellation sub band filter, D, subtracting a filtering value y(k) from the near-end sub band signal d(k), with an echo, after the extraction and then sending a different value to a far end, E, obtaining a sub band filter step length u(k) through calculation of an active factor f<l>(k) and a proportional matrix G(k) by utilizing a set membership filtering algorithm, and updating a weight coefficient vector W(k), and F, enabling k to meet an expression k=k+1, and repeating A-E steps until an end of a call. The method has the rapid convergence speed and the low steady-state error at one aspect, and has the rapid tracking capability at the other aspect; and the method has the good cancellation effect on an acoustic echo of a communication system.

The present invention discloses a target relay tracking algorithm based on intelligent street lamp partners. According to the present invention, a target is detected by utilizing a deep network model, and secondly, the features of the target are extracted by utilizing another one deep model. An intelligent street lamp partner can snapshot the images of a plurality of visual angles of an object, and save the features extracted at each visual angle. A system detects the target at set intervals, and fuses a detection result and a tracking result. When the object moves out from the visual field of the current intelligent street lamp partner and enters in the visual field of the next intelligent street lamp partner, the next intelligent street lamp partner similarly detects the target and extracts the features, and the extracted features are compared with the features saved by the last intelligent street lamp partner, so that whether the detected objects are the same object can be determined, and the target relay tracking is realized.

The analog dc voltage conversion method with synchro / resolver comprises: converting the axial angle of synchro / resolver into parallel natural binary O for output; amplifying the digital angle O with FPGA PLC to increase or decrease its output step for signal amplification or minification and output to a digital-analog dc voltage converter with a DAC, a reference power and a operational amplifier to complete the conversion. This invention has high reliability while low consumption.

The invention relates to a multi-model self-adaptive fusion filtering method of a ship dynamic positioning system and belongs to the technical field of ship dynamic positioning. The method includes the steps: (1), building a ship three-degree-of-freedom low-frequency and high-frequency motion model, and acquiring a filter state formula and a measuring formula; (2), utilizing a differential global positioning system and a platform compass to measure position information and a heading angle, and collecting information in real time; (3), utilizing prior information and posterior information to initialize input of a model-based filter; (4), on the basis of a system model, utilizing a strong tracking filter and a Sage-Husa filter for parallel filtering; (5), subjecting the model to probability updating, and utilizing residual covariance output by the filters to calculate model probability matched with the model; (6), according to the model probability, acquiring fusion output of multi-model state estimation, namely ship position and heading information. The multi-model self-adaptive fusion filtering method has the advantages of strong robustness, high accuracy in Sage-Husa filter state estimation, stable system, high positioning accuracy and the like.

A fluorine-containing compound represented by a general formula (c-1) shown below:RX-AN-(OR2)a   [Chemical Formula 1](c-1)[wherein, RX represents an organic group, AN represents a naphthalene ring that may have a substituent, R2 represents a base dissociable group, and a represents 1 or 2, provided that at least one among AN and said a R2 groups contains a fluorine atom].

Discloses is an adaptive method, applied to echo cancellation, of an active factor proportional sub band. The method comprises the steps: A, dividing an adaptive echo cancellation sub band filter input vector X(n) formed by a discrete value of a far-end signal into a sub band signal X(n), B, through N extraction, obtaining an input sub band signal X(k) after the extraction and a near-end sub band signal d(k), C, obtaining a filtering value y(k) of the input sub band signal X<i(k) after the extraction via an adaptive echo cancellation sub band filter, D, sending a sub band error signal e(k), obtained through calculation, after the echo cancellation to a far end, E, obtaining a sub band filter step length u(k) through calculation of an active factor f<l>(k) and a proportional matrix G(k) by utilizing a set membership filtering algorithm, and updating a weight coefficient vector W(k), and F, enabling k to meet an expression k=k+1, and repeating A-E steps until an end of a call. The method has the advantages of the high identification ability for a sparse system of call communication, the rapid convergence speed, the low steady-state error, the high tracking capability for a jump system and the good echo cancellation effect; and the method is easy to implement.

A self-adaptive sparse system identification method based on impact-interference-resistance of an independent activity factor mainly comprises the following steps: A, acquiring an output signal and a desired signal of a self-adaptive FIR (finite impulse response) filter; B, subtracting an output signal vector (FORMULA) from a desired signal vector d(n) to obtain an output error vector e(n) of the self-adaptive FIR filter; C, calculating a scaling factor column vector g(n) of a weight vector of the self-adaptive FIR filter; D, calculating a weight column vector w(n+1) of the self-adaptive FIR filter at the next moment; E, assigning n=n+1, repeating operations in steps A, B, C and D, namely continuously obtaining identification values w(n) of the sparse system at different moments. The impact-interference-resistant self-adaptive sparse system identification method is very fast in converging speed, strong in traceability and low in calculation complexity.

The invention belongs to the field of robots, and discloses a depth backstepping control method for an underactuated autonomous underwater vehicle on the basis of a dynamic surface slip form. The method comprises the following steps that: S(1) combining with the kinematic model of the autonomous underwater vehicle on a vertical plane to determine a control target as the convergence of a tracking error; S(2) on the basis of a backstepping method, designing a Lyapunov function, importing a virtual control variable, and designing a dynamic surface to eliminate a differential explosion phenomenoncaused by a traditional backstepping method; S(3) combining S(1) with S(2) to design a slip form surface and an adaptive control law to solve a depth control problem; and S(4) combining data in S(2) and S(3), according to a Lyapunov theory of stability and a comparison principle, using a closed-loop tracking error to regulate gain convergence to a compressed boundary set which is almost zero, andguaranteeing the semi-global uniform boundedness of a control system. By use of the method, the problems of model uncertainty and environment interference are solved, and the method is high in traceability for an expected path.

A triaxial holder includes a base, a horizontal roll shaft motor, a pitch shaft motor and a heading shaft motor which is mounted on a base. The base is provided with a third connector, a second control module connected to the heading shaft motor is arranged in the third connector, and the heading shaft motor drives the third connector to rotate circumferentially. The pitch shaft motor is mounted on the third connector and is also provided with a second connector, and the pitch shaft motor drives the second connector to rotate circumferentially. The horizontal shaft motor is installed in the second connector which is internally provided with a first control module, the first control module is connected with the horizontal shaft motor, the first control module is also connected with a first connector used for connecting with the load, and the horizontal shaft motor drives the first connector to rotate circumferentially. The tracking shooting effect is good, and it is possible to quickly and stably control the shooting device to be at the determined posture when the moving target is shot.

The invention provides a solid rocket engine thrust control method based on a radial basis function neural network. The method is characterized in that the method comprises the following steps: step 1, determining a control object, enabling an execution mechanism to employ a gas adjustment system and a pneumatic servo system, and selecting a solid rocket engine as a power device; step 2, establishing a mathematical model of a controlled object of the solid rocket engine; step 3, determining a reference model and an actual model according to expected control performance requirements; step 4, designing an RBF neural network controller for solid rocket engine thrust self-adaptive control, and selecting learning indexes of a neural network; and step 5, optimizing RBF neural network controllerparameters by using a genetic algorithm. According to the method, the neural network parameter setting is optimized through the genetic algorithm, so that the solid rocket engine control system basedon the RBF neural network has good tracking capability on the characteristics of the given reference model; and the method has relatively good anti-interference performance and high robustness.

A proportional self-adaption telephone echo cancellation method based on decorrelation mainly includes the steps that (A) far-end signal filtering is carried out to obtain an output value y (n) of an adaptive filter; (B) a decorrelation operation of an input vector X (n) is conducted to obtain an innovation vector Z (n) of a tap weight vector w (n) of the adaptive filter, namely Z (n)= X (n) - a (n) X (n-1); (C) echo cancellation is carried out, wherein the output value y (n) of the adaptive filter is subtracted from a near-end signal d (n) with an echo picked up by a near-end microphone, then the near-end signal d (n) with the echo is sent back to a far-end, and a loopback signal d (n) is a useful signal e (n) of an identification cancellation echo obtained by subtracting d (n) from y (n), namely e (n)= d (n) - y (n); (D) the tap weight vector of the filter is updated; (E) the equation that n= n+1 is made, and the steps from A to D are repeated until a call is ended. The method has high identification capacity for sparse systems like telephone communication, and is high in convergence rate, low in steady state error and good in echo cancellation effect.

The invention discloses a bus net load prediction method based on ARIMA and phase-space reconstruction (SVR). The method comprises the following steps: 1, determining the parameters of an ARIMA modeland establishing the ARIMA model; 2, subjecting the residual sequence of the ARIMA model to SVR based on a CC method; 3, based on a reconstructed residual matrix, creating a net load nonlinear part prediction model by using SVR; and 4, obtaining a final result of net load prediction. The prediction model provided by the invention makes full use of the good tracking ability of the ARIMA model to linear changes and the good capture ability of the SVR to nonlinear changes. The SVR via CC method is based on a theory that the embedding space dimension and time delay are interdependent, has low calculated quantity, high reliability and good anti-noise performance. The SVR model has good tracking and catching ability for nonlinear changes, and can effectively analyze the nonlinear components of the net load.

A positive resist composition including a resin component (A) which exhibits increased solubility in an alkali developing solution under the action of acid and an acid-generator component (B), the resin component (A) including a polymeric compound (A1) having a structural unit (a1) containing an acid dissociable, dissolution inhibiting group, a structural unit (a5) containing a base dissociable group an a structural unit (a6) represented by general formula (a6-1) (R represents a hydrogen atom, an alkyl group of 1 to 5 carbon atoms or a halogenated alkyl group of 1 to 5 carbon atoms; each of R2 and R3 independently represents a hydrogen atom or an alkyl group that may contain an oxygen atom at an arbitrary position, or R2 and R3 are bonded together to form an alkylene group; and W represents a cyclic alkylene group that may include an oxygen atom at an arbitrary position).

The invention discloses a lithium ion power battery state-of-charge estimation method and device, a medium and equipment, and the method comprises the steps: firstly carrying out the importing of a battery model, and carrying out the parameter recognition of the imported battery model; constructing a power battery SOC estimation basic formula according to the identified model parameters, and fusing the power battery SOC estimation basic formula into a Kalman filtering algorithm; initializing assignment of Kalman filtering state vectors and error covariance matrix parameters; during SOC actualestimation, collecting open-circuit voltage and load current of the power battery in real time; introducing a fading factor into the fused Kalman filtering, and estimating the SOC value of the power battery based on the Kalman filtering introduced with the fading factor. According to the method, estimation of the state of charge of the battery is realized based on an optimized nonlinear system filtering technology, the method has excellent characteristics of a strong tracking filter when a system model is uncertain, and real-time, accurate and dynamic estimation of the state of charge of the lithium ion power battery can be realized.

The invention relates to a hit and run-prevention vehicle positioning system. The hit and run-prevention vehicle positioning system comprises a server, an on-board memory, an on-board driving condition recording device, a GPS positioning terminal, a wireless transceiving module, and a microprocessor. The on-board memory, the on-board driving condition recording device, the GPS positioning terminal, and the wireless transceiving module are connected with the microprocessor, and the wireless transceiving module is connected with a server by a network. The on-board driving condition recording device comprises a pressure sensor disposed on a vehicle brake; a tachograph and an acceleration transducer, which are disposed in the vehicle. The on-board driving condition recording device is used for the real-time recording of the driving information of the vehicle and the driving information of the driver of the vehicle. The hit and run-prevention vehicle positioning system has advantages of good reliability, high practicability, strong identification and tracking capability, and ability of tracking traffic accident vehicle effectively in guaranteed way.

The invention provides a method for predicting the health state of a lithium battery based on SREKF. The method comprises the following steps: firstly, establishing a mathematical model of a state parameter of the lithium battery, and obtaining a state equation of ohmic internal resistance and an observation equation of the ohmic internal resistance; secondly, identifying an offline parameter of alithium battery model, and obtaining an initial value of the SREKF; meanwhile, obtaining an output sequence of a predicted end voltage Uc; then improving EKF to obtain the SREKF; and finally, inputting the measured voltage, current and margin sequence of the lithium battery into the SREKF to no update the state equation and the observation equation, updating an optimal value of the current stateof the lithium battery system of the SREKF by using the output sequence of the predicted end voltage Uc and a measured end voltage sequence, iterating the SREKF according to the number of experiment measured values to obtain a predicted value sequence of the ohmic internal resistance, that is, the quantity of state of the health state of the lithium battery. By adoption of the method disclosed bythe invention, when the internal resistance of the lithium battery is estimated by the traditional EKF, the problems of large estimation errors, low precision and poor robustness are solved.

The invention relates to an unmanned aerial surveying and mapping vehicle posture control method based on an RBF neural network. The control method is characterized by comprising the following steps of creating a non-linear dynamic model of an aerial vehicle serving as a controlled object; designing an unmanned aerial surveying and mapping vehicle PID controller according to the non-linear model;utilizing the non-linear mapping capability of the neural network to obtain the parameter adjustment variable quantity of the controller; combining an RBF neural network method with a PID control method to obtain a self-adapting PID control method based on the RBF neural network, and conducting simulation on the non-linear model of the unmanned aerial surveying and mapping vehicle.

A combined step size echo cancellation method with a high tracking performance comprises the steps of: A, collecting far-end input signals u(n) and a close-range desired signals d(n); B, inputting aninput signal discrete value u(n) into an adaptive filter to obtain output signals y(n); C, subtracting y(n) from d(n) to perform echo cancellation, and transmitting error signals e(n) to a far end; D,calculating error vectors of the adaptive filter; E, updating a tapping weight value w(n) of the adaptive filter, if the relative intensity value of the error signals at the current moment as shown in the description is larger than a set threshold value, and an absolute intensity value r(n-1)=log10(e2(n-1)) is smaller than the set threshold value, forcibly resetting a weight value error matrix track J(n) to be a large initial value; F, updating a weight value error matrix track J(n+1); and G, allowing n to be equal to n+1, and repeating the steps A, B, C, D, E and F until the conversation isfinished. The method has a good tracking performance when an acoustics echo channel is changed.

The invention discloses a dynamic gesture tracking method based on a convolutional neural network, relates to the technical field of computer vision, and aims to solve the problems of poor tracking effect, insufficient tracking real-time performance and the like caused by skin color interference in a gesture tracking process in a complex scene, and the method comprises the following steps: the step 1, taking dynamic gesture tracking under a complex background as a visual task; the step 2, selecting a gesture image sample for filtering processing, and then making a gesture training set; the step 3, determining a YOLOv3-gesture gesture detection network structure, wherein the structural formula of the YOLOv3-gesture gesture detection network is shown in the specification; the step 4, completing dynamic gesture tracking by using a planning area detection framework; the step 5, training the YOLOv3-gesture model to obtain a dynamic gesture tracking model; and the step 6, completing dynamicgesture tracking by utilizing the obtained model. When skin color interference is encountered in the gesture tracking process in a complex scene, the tracking effect is strong, and the tracking real-time performance is sufficient.