Tracking turntable follow-up identification adjustment control method and system

Abstract

The invention is suitable for the field of tracking rotary tables, and provides a follow-up identification adjustment control method and system for a tracking rotary table, and the method comprises the steps: collecting the peripheral data of the tracking rotary table in a multi-modal manner; acquiring and fusing the acquired surrounding data of the tracking turntable; establishing a three-dimensional space error model; generating a prediction control strategy based on the three-dimensional space error model; and generating an intelligent feedback optimization result based on a prediction control strategy, and the method has the beneficial effects that by means of deep fusion and accurate analysis of multi-modal data, the tracking turntable is endowed with excellent environmental adaptability and target tracking capability. Complex environment interference such as atmospheric refraction, illumination change and shielding can be effectively handled, the tracking precision is improved, online self-adaptive adjustment can be achieved through an intelligent feedback optimization mechanism and a control strategy, the stability, reliability and intelligence of the system are remarkably enhanced, and an advanced and efficient technical solution is provided for target tracking application in multiple fields.

Description

Technical Field

[0001] The present invention belongs to the field of tracking turntables, and particularly relates to a follow-up recognition adjustment control method and system for a tracking turntable. Background Art

[0002] As a device that can accurately control the pointing of an object and track a target in real time, a tracking turntable has extremely wide applications in fields such as aerospace astronomy observation and industrial automation.

[0003] Currently, the control methods of tracking turntables mainly have the following deficiencies: on the one hand, in the target recognition link, the recognition accuracy of traditional tracking turntables for targets in complex backgrounds is poor. When the target and background features are similar, or are interfered by factors such as light changes and occlusion, the recognition accuracy will be greatly reduced, resulting in the tracking turntable being unable to lock the target in time. With the rapid development of technology, higher requirements have been put forward for the performance of tracking turntables in various application fields. There is an urgent need for a new method that can improve the target recognition accuracy, optimize the follow-up adjustment control, and enhance the environmental adaptability of the system. Summary of the Invention

[0004] The purpose of the present invention is to provide a follow-up recognition adjustment control method and system for a tracking turntable, aiming to solve the problems raised in the above background art.

[0005] The present invention is implemented as follows. On the one hand, a follow-up recognition adjustment control method for a tracking turntable, the method includes:

[0006] Collect multi-modal data around the tracking turntable;

[0007] Obtain and fuse the collected data around the tracking turntable;

[0008] Establish a three-dimensional space error model;

[0009] Generate a predictive control strategy based on the three-dimensional space error model;

[0010] Generate an intelligent feedback optimization result based on the predictive control strategy.

[0011] As a further solution of the present invention, the specific steps of collecting multi-modal data around the tracking turntable include:

[0012] Realize synchronous acquisition based on a global shutter camera and a six-axis IMU;

[0013] Measure the angular velocity at the level of 0.001° / h through a fiber optic gyroscope, and synchronously obtain the target image sequence, the pose data of the tracking turntable, and the environmental parameters.

[0014] As a further solution of the present invention, the specific steps of obtaining and fusing the collected data around the tracking turntable include:

[0015] Obtain the target image sequence, the pose data of the tracking turntable, and the environmental parameters;

[0016] Extract the target vector parameters in the target grayscale image sequence, the pose data of the tracking turntable, and the environmental parameters respectively to generate a grayscale image feature vector , a turntable state vector and an environmental parameter vector ;

[0017] Construct a three - element mapping model of environment - target - platform;

[0018] The specific form of the three - element mapping model is:

[0019] ;

[0020] In the formula, is the azimuth deviation, is the pitch deviation, is the environmental compensation coefficient;

[0021] Based on Retinex - Net, implement dynamic illumination compensation and correlate the compensation coefficient with the variance of the grayscale image gradient;

[0022] ;

[0023] In the formula, is the variance of the grayscale image gradient, is the preset maximum variance of the grayscale image gradient, is the preset adjustment parameter, is the illumination intensity;

[0024] Adopt a spatio - temporal context model to fuse the information of five consecutive frames of target images, construct an occlusion confidence prediction matrix, and output the occlusion confidence at the current moment :

[0025] ;

[0026] In the formula, is the occlusion confidence of the target at the current moment with a value range of , is the weight coefficient, is the intersection - over - union ratio of the th frame, is the maximum intersection - over - union ratio.

[0027] As a further solution of the present invention, the establishment of the three - dimensional space error model specifically includes:

[0028] Establish a three - dimensional space error model including atmospheric refraction correction;

[0029] The calculation process of the three-dimensional space error model is as follows:

[0030] ;

[0031] ;

[0032] In the formula, is the total azimuth error, is the true azimuth, is the currently measured azimuth of the tracking turntable, is the azimuth correction term caused by atmospheric refraction, is the atmospheric refractive index, is the preset target elevation angle.

[0033] As a further solution of the present invention, generating a predictive control strategy based on the three-dimensional space error model specifically includes:

[0034] Extract the actual state vector in the turntable state vector ; ;

[0035] Based on the control instruction of the tracking turntable at the previous moment, obtain the control input vector of the tracking turntable ;

[0036] Based on the LSTM prediction of the future state of the outer loop, generate an outer loop neural network prediction model ;

[0037] The outer loop neural network prediction model specifically includes:

[0038] ;

[0039] In the formula, is the predicted value of the turntable state at the future moment, is the actual state vector of the turntable at the current moment, the control input vector at the current moment, is the environmental parameter vector at the current moment;

[0040] Based on the outer loop neural network prediction model , calculate the inner loop adaptive sliding mode control adjustment parameter ;

[0041] The calculation process of the inner loop adaptive sliding mode control adjustment parameter is as follows:

[0042] ;

[0043] In the formula, is the final output control quantity, is the proportional gain, is the differential gain, is the system error of the tracking turntable target state and the actual state, is the switching gain, is the sign function, which is used to determine the direction of the control quantity based on the sign of the sliding mode surface ;

[0044] The associated switching gain is related to the occlusion confidence ;

[0045] The specific association process is as follows:

[0046] ;

[0047] In the formula, is the preset initial switching gain.

[0048] As a further solution of the present invention, the intelligent feedback optimization result generated based on the predictive control strategy specifically includes:

[0049] Implement three-loop collaborative control, and the three-loop collaborative control specifically includes:

[0050] Position loop: Based on Kalman filtering, realize target trajectory prediction;

[0051] Speed loop: Variable parameter PID control;

[0052] Disturbance compensation loop: Use a disturbance observer to compensate the friction torque in real time ;

[0053] The friction torque The specific calculation process is as follows:

[0054] ;

[0055] In the formula, is the static friction coefficient, is the dynamic friction coefficient, is the rotational speed of the tracking turntable motor.

[0056] As a further solution of the present invention, on the other hand, a tracking turntable servo recognition and adjustment control system, the system includes:

[0057] A multi-modal acquisition module, which is used to multi-modally acquire the data around the tracking turntable;

[0058] An acquisition and fusion module, configured to acquire and fuse the collected data around the tracking turntable;

[0059] A three-dimensional space error model module, configured to establish a three-dimensional space error model;

[0060] A first generation module, configured to generate a predictive control strategy based on the three-dimensional space error model;

[0061] A second generation module, configured to generate an intelligent feedback optimization result based on the predictive control strategy.

[0062] As a further solution of the present invention, the multimodal acquisition module specifically includes:

[0063] A synchronous acquisition unit, configured to achieve synchronous acquisition based on a global shutter camera and a six-axis IMU;

[0064] A measurement unit, configured to achieve measurement of an angular velocity at a level of 0.001° / h through a fiber optic gyroscope;

[0065] A synchronous acquisition unit, configured to synchronously acquire a target image sequence, tracking turntable pose data, and environmental parameters;

[0066] The synchronous acquisition unit achieves synchronous acquisition based on a global shutter camera and a six-axis IMU, the measurement unit achieves measurement of an angular velocity at a level of 0.001° / h through a fiber optic gyroscope, and the synchronous acquisition unit synchronously acquires a target image sequence, tracking turntable pose data, and environmental parameters.

[0067] As a further solution of the present invention, the acquisition and fusion module specifically includes:

[0068] A first acquisition unit, configured to acquire a target image sequence, tracking turntable pose data, and environmental parameters;

[0069] An extraction unit, configured to respectively extract target vector parameters in a target gray image sequence, tracking turntable pose data, and environmental parameters;

[0070] A first generation unit, configured to generate a gray image feature vector , a turntable state vector and an environmental parameter vector ;

[0071] A ternary mapping model unit, configured to construct a ternary mapping model of environment-target-platform;

[0072] A first association unit, configured to achieve dynamic light compensation based on Retinex-Net and associate a compensation coefficient with the gray image gradient variance;

[0073] A fusion unit, configured to fuse the information of five consecutive frames of target images by using a spatio-temporal context model;

[0074] A prediction matrix unit for constructing an occlusion confidence prediction matrix;

[0075] An output unit for outputting the occlusion confidence at the current moment 。

[0076] As a further solution of the present invention, the first generation module specifically includes:

[0077] An extraction unit for extracting the actual state vector in the turntable state vector ; ;

[0078] A second acquisition unit for obtaining the control input vector of the tracking turntable based on the control instruction of the tracking turntable at the previous moment ;

[0079] A second generation unit for generating an outer loop neural network prediction model by predicting the future state of the outer loop based on LSTM ;

[0080] A calculation unit for calculating the inner loop adaptive sliding mode control adjustment parameter based on the outer loop neural network prediction model ; ;

[0081] A second association unit for associating the switching gain with the occlusion confidence ; 。

[0082] A method and system for tracking turntable follow-up recognition and adjustment control provided by the present invention endow the tracking turntable with excellent environmental adaptability and target tracking ability by virtue of the deep fusion and accurate analysis of multi-modal data. It can effectively cope with complex environmental interferences such as atmospheric refraction, light changes, and occlusion, improve the tracking accuracy, and through the intelligent feedback optimization mechanism, the control strategy can be adaptively adjusted online, significantly enhancing the stability, reliability, and intelligence of the system, providing an advanced and efficient technical solution for target tracking applications in multiple fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 is the main flowchart of a method for tracking turntable follow-up recognition and adjustment control.

[0084] Figure 2 is the specific flowchart of collecting multi-modal data around the tracking turntable in a method for tracking turntable follow-up recognition and adjustment control.

[0085] Figure 3 is the flowchart of obtaining and fusing the collected data around the tracking turntable in a method for tracking turntable follow-up recognition and adjustment control.

[0086] Figure 4 It is a flowchart of generating a predictive control strategy based on a three-dimensional space error model in a tracking turntable servo recognition adjustment control method.

[0087] Figure 5 It is the main structure diagram of a tracking turntable servo recognition adjustment control method system.

[0088] Figure 6 It is the structural block diagram of the multi-modal acquisition module in a tracking turntable servo recognition adjustment control method system.

[0089] Figure 7 It is the structural block diagram of the acquisition fusion module in a tracking turntable servo recognition adjustment control method system.

[0090] Figure 8 It is the structural block diagram of the first generation module in a tracking turntable servo recognition adjustment control method system. Specific implementation mode

[0091] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0092] The following describes the specific implementation of the present invention in detail with reference to specific embodiments.

[0093] A tracking turntable servo recognition adjustment control method and system provided by the present invention solve the technical problems in the background art.

[0094] As Figure 1 shown, it is the main flowchart of a tracking turntable servo recognition adjustment control method provided by an embodiment of the present invention. The tracking turntable servo recognition adjustment control method includes:

[0095] Step S100: Multimodal acquisition of data around the tracking turntable;

[0096] Step S200: Obtain and fuse the data around the tracking turntable collected;

[0097] Step S300: Establish a three-dimensional space error model;

[0098] Step S400: Generate a predictive control strategy based on the three-dimensional space error model;

[0099] Step S500: Generate an intelligent feedback optimization result based on the predictive control strategy.

[0100] In the application of this embodiment, advanced devices such as a global shutter camera, a six-axis IMU, and a fiber optic gyro are selected in the multi-modal acquisition link to construct a spatio-temporal synchronous acquisition system. With a synchronization accuracy of nanosecond level, high-definition image information of the target object, accurate pose data of the turntable, and all-round environmental parameters including light intensity, temperature and humidity, and atmospheric visibility are acquired in parallel. During this process, the camera acquires images at a resolution of 4K and a frame rate of 60fps, and the IMU acquires pose information at a frequency of 1kHz to ensure the integrity and timeliness of the data. After obtaining the data, it enters the intelligent fusion stage. The Retinex-Net network is used to achieve dynamic light compensation, and the compensation coefficient is accurately adjusted according to the correlation between the light intensity and the image gradient variance to improve the image quality. At the same time, a spatio-temporal context model is used to fuse the information of five consecutive frames of images to construct an occlusion confidence prediction matrix to accurately evaluate the target occlusion situation. In addition, an environment-target-platform ternary mapping model is constructed to establish the mapping relationship between image features, turntable states, and environmental parameters to azimuth deviation , pitch angle deviation and environmental compensation coefficient , and deeply fuse multi-source information. Subsequently, a three-dimensional space error model is established. Considering factors such as atmospheric refraction, the pointing deviation of the target in the spherical coordinate system is accurately calculated. Based on the three-dimensional space error model, a predictive control strategy is generated. A dual-loop control architecture based on MPC is designed. The outer loop uses a neural network prediction model, an LSTM network, to predict the future system state according to the current system state, control input, and environmental parameters, and plan the control sequence in advance. The inner loop uses an adaptive sliding mode controller to adjust the control quantity in real time according to the error and its change rate, and associates the switching gain with the occlusion confidence to enhance the robustness in complex scenarios.

[0101] As Figure 2 shown, as a preferred embodiment of the present invention, the multi-modal acquisition and tracking of the data around the turntable specifically includes:

[0102] Step S101: Synchronous acquisition is realized based on the global shutter camera and the six-axis IMU;

[0103] Step S102: Angular velocity measurement at the level of 0.001° / h is realized through the fiber optic gyro, and the target image sequence, the pose data of the tracking turntable, and the environmental parameters are synchronously acquired.

[0104] In the application of this embodiment, a synchronous acquisition system is constructed with the help of a global shutter camera and a six-axis IMU to accurately synchronously acquire the target image sequence and the pose data of the tracking turntable, ensuring the consistency and accuracy of the data from the spatio-temporal dimension. At the same time, a precision measurement device, the fiber optic gyro, is introduced to achieve ultra-high-precision measurement at the level of 0.001° / h. On this basis, all-round environmental parameters are collected, covering information such as light intensity, temperature and humidity.

[0105] As Figure 3 shown, as a preferred embodiment of the present invention, the obtaining and fusing the collected data around the tracking turntable specifically includes:

[0106] Step S201: Obtain the target image sequence, the pose data of the tracking turntable, and the environmental parameters;

[0107] Step S202: Extract the target vector parameters in the target gray image sequence, the pose data of the tracking turntable, and the environmental parameters respectively, and generate the gray image feature vector , the turntable state vector and the environmental parameter vector ;

[0108] Step S203: Construct a three - element mapping model of environment - target - platform;

[0109] The three - element mapping model is specifically:

[0110] ;

[0111] In the formula, is the azimuth deviation, is the pitch deviation, is the environmental compensation coefficient

[0112] The three - element mapping model is specifically: Step S204: Implement dynamic light compensation based on Retinex - Net, and associate the compensation coefficient with the gray - image gradient variance:

[0113] ;

[0114] In the formula, is the gray - image gradient variance, is the preset maximum gray - image gradient variance, is the preset adjustment parameter, is the light intensity;

[0115] Step S205: Adopt the spatio - temporal context model to fuse the information of five consecutive frames of target images, construct an occlusion confidence prediction matrix, and output the occlusion confidence at the current moment ;

[0116] ;

[0117] In the formula, is the occlusion confidence of the target at the current moment , and the value range is , is the weight coefficient, is the intersection - over - union ratio of the th frame, is the maximum intersection over union;

[0118] When this embodiment is applied, it comprehensively obtains the target image sequence, the pose data of the tracking turntable, and the environmental parameters, laying a foundation for subsequent analysis. Subsequently, the target vector parameters are accurately extracted from various types of data to generate the gray image feature vector, the turntable state vector, and the environmental parameter vector, realizing the feature processing of the data. On this basis, a three-element mapping model of the environment-target-platform is constructed, and through the precise mapping relationship, the azimuth deviation, the pitch angle deviation, and the environmental compensation coefficient are output, effectively correlating the relationship among the environment, the target, and the platform. In terms of light compensation, dynamic light compensation is realized based on Retinex-Net. Retinex-Net is a neural network model based on the Retinex theory, which can be trained on a large number of image data to enable the network to learn the illumination and reflection characteristics of the images. The compensation coefficient is cleverly correlated with the gray image gradient variance, and the compensation degree is flexibly adjusted according to the richness of image details and the illumination intensity. At the same time, the spatio-temporal context model is used to fuse the information of five consecutive frames of target images to construct an occlusion confidence prediction matrix, and the occlusion confidence at the current moment is accurately output, providing key data support for dealing with complex environments.

[0119] As a preferred embodiment of the present invention, the establishment of the three-dimensional space error model specifically includes:

[0120] Establish a three-dimensional space error model including atmospheric refraction correction;

[0121] The calculation process of the three-dimensional space error model is as follows:

[0122] ;

[0123] ;

[0124] In the formula, is the total azimuth error, is the true azimuth, is the currently measured azimuth of the tracking turntable, is the azimuth correction term caused by atmospheric refraction, is the atmospheric refractive index, is the preset target elevation angle;

[0125] It should be understood that since the atmospheric refraction phenomenon in the actual environment will significantly affect the light propagation path, and thus cause a deviation in the azimuth measurement of the tracking turntable, this technology is committed to constructing a set of high-precision error calculation systems. To solve the above problems, a three-dimensional space error model including atmospheric refraction correction is established.. Among them, As the total azimuth error, it is the core index for evaluating the azimuth accuracy of the tracking turntable, which is composed of the true azimuth , tracking the current measured azimuth angle of the turntable and the azimuth correction term caused by atmospheric refraction together constitute. The azimuth correction term caused by atmospheric refraction is calculated based on the atmospheric refractive index and the preset target elevation angle for calculation. The atmospheric refractive index is affected by factors such as temperature, humidity, and air pressure in the environment and will change dynamically, while the target elevation angle is directly related to the degree of refraction of light in the atmosphere. Through calculation, the change in the propagation direction of light caused by atmospheric refraction can be accurately quantified, and then the azimuth correction value caused by atmospheric refraction can be obtained. The establishment of this model realizes the accurate capture and quantification of the azimuth angle error of the tracking turntable, effectively integrates the actual measurement value, the true value, and the environmental interference correction amount, provides highly reliable error data support for the subsequent control strategy of the tracking turntable, and significantly improves the azimuth angle measurement accuracy and target tracking accuracy of the tracking turntable in a complex atmospheric environment.

[0126] Such as Figure 4 shown, as a preferred embodiment of the present invention, the generating of the predictive control strategy based on the three-dimensional space error model specifically includes:

[0127] Step S401: Extract the actual state vector in the state vector of the turntable ;

[0128] Step S402: Based on the control instruction of the tracking turntable at the previous moment, obtain the control input vector of the tracking turntable;

[0129] Step S403: Based on the LSTM prediction of the future state of the outer loop, generate the outer loop neural network prediction model ;

[0130] The outer loop neural network prediction model specifically includes:

[0131] ;

[0132] In the formula, is the predicted value of the turntable state at the future moment, is the actual state vector of the turntable at the current moment, the control input vector at the current moment, is the environmental parameter vector at the current moment;

[0133] Step S404: Based on the outer-loop neural network prediction model , calculate the adjustment parameters of the inner-loop adaptive sliding mode control ;

[0134] The calculation process of the adjustment parameters of the inner-loop adaptive sliding mode control is as follows:

[0135] ;

[0136] In the formula, is the final output control quantity, is the proportional gain, is the differential gain, is the system error between the tracking turntable target state and the actual state, is the switching gain, is the sign function, which is used to determine the direction of the control quantity based on the sign of the sliding mode surface ;

[0137] Step S405: Correlate the switching gain with the occlusion confidence ;

[0138] The specific correlation process is as follows:

[0139] ;

[0140] In the formula, is the preset initial switching gain;

[0141] In this embodiment, when applied, the actual state vector is separated from the turntable state vector , which covers the key operating parameters of the turntable such as position, angle, and speed. At the same time, according to the control instruction for tracking the turntable at the previous moment, the control input vector containing information such as motor drive signals and power is obtained. Subsequently, a long short-term memory network (LSTM) is introduced to construct an outer-loop neural network prediction model. This model is calculated through this formula , fully integrating the actual state vector of the turntable at the current moment, the control input vector , and the environmental parameter vector , where the environmental parameter vector includes environmental factors such as atmospheric refractive index, light intensity, temperature, and humidity that affect the operation of the turntable. Through the time-series data processing ability of LSTM, the predicted value of the turntable state at the future moment is effectively predicted., it provides forward-looking information for the system to plan control strategies in advance. Based on the results output by the outer-loop neural network prediction model, the inner-loop adaptive sliding mode control adjustment parameters are further calculated , in this process, the proportional gain is used to amplify the system error between the target state and the actual state of the tracking turntable, so as to quickly respond to the error change; through the differential gain , the error change rate is processed to suppress the system overshoot and ensure the stability of the turntable operation. At the same time, the switching gain is closely related to the occlusion confidence , and the switching gain is dynamically adjusted according to the degree of target occlusion, where is the preset initial switching gain. When the occlusion confidence increases, the switching gain is increased to enhance the robustness of the sliding mode controller in the occlusion scenario and optimize the control strategy. Finally, the control quantity is output to achieve precise control of the tracking turntable.

[0142] As a preferred embodiment of the present invention, the generation of the intelligent feedback optimization result based on the predictive control strategy specifically includes:

[0143] Implement three-closed-loop collaborative control, and the three-closed-loop collaborative control specifically includes:

[0144] Position loop: Based on Kalman filtering, target trajectory prediction is realized;

[0145] Speed loop: Variable parameter PID control;

[0146] Disturbance compensation loop: An interference observer is used to compensate the friction torque in real time;

[0147] The specific calculation process of the friction torque is as follows:

[0148] ;

[0149] In the formula, is the static friction coefficient, is the dynamic friction coefficient, is the rotational speed of the tracking turntable motor;

[0150] In the application of this embodiment, in the follow-up turntable servo recognition adjustment control method, when generating the intelligent feedback optimization result based on the predictive control strategy, a three-closed-loop collaborative control architecture is adopted. The position loop utilizes the data fusion and prediction capabilities of Kalman filtering to comprehensively process multi-modal sensor data, accurately predict the target trajectory, and guide the turntable to accurately point to the target. The speed loop adopts variable-parameter PID control, which flexibly adjusts the proportional, integral, and differential parameters according to the operating conditions of the turntable, realizes the efficient tracking of the turntable speed to the target motion speed, and ensures the smoothness of the tracking process. The disturbance compensation loop monitors the friction torque in real time through a disturbance observer, considering the static friction coefficient and the dynamic friction coefficient as well as the rotational speed of the motor of the follow-up turntable , and compensates the friction torque in real time to eliminate the interference of friction factors in mechanical transmission on the operation of the turntable. The three closed loops cooperate with each other, comprehensively improving the control accuracy and stability of the follow-up turntable from multiple key dimensions such as target trajectory prediction, speed tracking control, and disturbance compensation, realizing intelligent feedback optimization, and ensuring the reliable operation of the system in complex environments.

[0151] As Figure 5 shown, as another preferred embodiment of the present invention, on the other hand, a follow-up turntable servo recognition adjustment control system, the system includes:

[0152] A multi-modal acquisition module 100 for multi-modally acquiring data around the follow-up turntable;

[0153] An acquisition fusion module 200 for acquiring and fusing the data around the follow-up turntable collected;

[0154] A three-dimensional space error model module 300 for establishing a three-dimensional space error model;

[0155] A first generation module 400 for generating a predictive control strategy based on the three-dimensional space error model;

[0156] A second generation module 500 for generating an intelligent feedback optimization result based on the predictive control strategy.

[0157] In the application of this embodiment, the multi-modal acquisition module 100 multi-modally acquires data around the follow-up turntable, the acquisition fusion module 200 acquires and fuses the data around the follow-up turntable collected, the three-dimensional space error model module 300 establishes a three-dimensional space error model, based on the three-dimensional space error model, the first generation module 400 generates a predictive control strategy, and based on the predictive control strategy, the second generation module 500 generates an intelligent feedback optimization result.

[0158] As Figure 6 shown, as another preferred embodiment of the present invention, the multi-modal acquisition module 100 specifically includes:

[0159] The synchronous acquisition unit 101 is used to achieve synchronous acquisition based on a global shutter camera and a six-axis IMU;

[0160] The measurement unit 102 is used to achieve angular velocity measurement at the level of 0.001° / h through a fiber optic gyroscope;

[0161] The synchronous acquisition unit 103 is used to synchronously acquire the target image sequence, the pose data of the tracking turntable, and the environmental parameters;

[0162] When this embodiment is applied, the synchronous acquisition unit 101 achieves synchronous acquisition based on a global shutter camera and a six-axis IMU, the measurement unit 102 achieves angular velocity measurement at the level of 0.001° / h through a fiber optic gyroscope, and the synchronous acquisition unit 103 synchronously acquires the target image sequence, the pose data of the tracking turntable, and the environmental parameters.

[0163] As Figure 7 shown, as another preferred embodiment of the present invention, the acquisition and fusion module 200 specifically includes:

[0164] The first acquisition unit 201 is used to acquire the target image sequence, the pose data of the tracking turntable, and the environmental parameters;

[0165] The extraction unit 202 is used to respectively extract the target vector parameters in the target gray image sequence, the pose data of the tracking turntable, and the environmental parameters;

[0166] The first generation unit 203 is used to generate the gray image feature vector , the turntable state vector and the environmental parameter vector ;

[0167] The three-element mapping model unit 204 is used to construct the three-element mapping model of environment-target-platform;

[0168] The first association unit 205 is used to achieve dynamic light compensation based on Retinex-Net and associate the compensation coefficient with the gray image gradient variance;

[0169] The fusion unit 206 is used to fuse the information of five consecutive frames of target images by using the spatio-temporal context model;

[0170] The prediction matrix unit 207 is used to construct the occlusion confidence prediction matrix;

[0171] The output unit 208 is used to output the occlusion confidence at the current moment .

[0172] In the application of this embodiment, the first acquisition unit 201 acquires the target image sequence, the pose data of the tracking turntable, and the environmental parameters. The extraction unit 202 extracts the target vector parameters in the target grayscale image sequence, the pose data of the tracking turntable, and the environmental parameters respectively. The first generation unit 203 generates the grayscale image feature vector , the turntable state vector and the environmental parameter vector . The three - element mapping model unit 204 constructs a three - element mapping model of environment - target - platform, and realizes dynamic light compensation based on Retinex - Net. The first association unit 205 associates the compensation coefficient with the grayscale image gradient variance. The fusion unit 206 fuses the information of five consecutive frames of target images by using the spatio - temporal context model. The prediction matrix unit 207 constructs an occlusion confidence prediction matrix, and the output unit 208 outputs the occlusion confidence at the current moment .

[0173] As Figure 8 shown, as another preferred embodiment of the present invention, the first generation module 400 specifically includes:

[0174] An extraction unit 401, configured to extract the actual state vector in the turntable state vector ;

[0175] A second acquisition unit 402, configured to acquire the control input vector of the tracking turntable based on the control instruction of the tracking turntable at the previous moment;

[0176] A second generation unit 403, configured to generate an outer - loop neural network prediction model by predicting the future state of the outer loop based on LSTM;

[0177] A calculation unit 404, configured to calculate the inner - loop adaptive sliding - mode control adjustment parameter based on the outer - loop neural network prediction model ;

[0178] A second association unit 405, configured to associate the switching gain with the occlusion confidence .

[0179] In the application of this embodiment, the extraction unit 401 extracts the actual state vector in the turntable state vector . Based on the control instruction of the tracking turntable at the previous moment, the second acquisition unit 402 acquires the control input vector of the tracking turntable. Based on predicting the future state of the outer loop by LSTM, the second generation unit 403 generates an outer - loop neural network prediction model , based on the outer-loop neural network prediction model , the computing unit 404 calculates the inner-loop adaptive sliding mode control adjustment parameters , the second correlation unit 405 correlates the switching gain with the occlusion confidence .

[0180] In the above embodiments of the present invention, a tracking turntable servo recognition and adjustment control method is provided, and a tracking turntable servo recognition and adjustment control method system is provided. In the multi-modal acquisition link, advanced devices such as a global shutter camera, a six-axis IMU, and an optical fiber gyro are selected to construct a spatio-temporal synchronous acquisition system. With a synchronization accuracy of nanoseconds, high-definition image information of the target object, accurate pose data of the turntable, and all-round environmental parameters including light intensity, temperature and humidity, and atmospheric visibility are collected in parallel. During this process, the camera collects images at a resolution of 4K and a frame rate of 60fps, and the IMU obtains pose information at a frequency of 1kHz to ensure the integrity and timeliness of the data. After obtaining the data, it enters the intelligent fusion stage. The Retinex-Net network is used to achieve dynamic light compensation. According to the correlation relationship between the light intensity and the image gradient variance, the compensation coefficient is accurately adjusted to improve the image quality. At the same time, the spatio-temporal context model is used to fuse the information of five consecutive frames of images to construct an occlusion confidence prediction matrix to accurately evaluate the target occlusion situation. In addition, an environment-target-platform ternary mapping model is constructed to establish the mapping relationship from image features, turntable state and environmental parameters to azimuth deviation , pitch angle deviation and environmental compensation coefficient , and deeply fuse multi-source information. Subsequently, a three-dimensional space error model is established. Considering factors such as atmospheric refraction, the pointing deviation of the target in the spherical coordinate system is accurately calculated. Based on the three-dimensional space error model, a predictive control strategy is generated. A double-loop control architecture based on MPC is designed. The outer loop uses a neural network prediction model, an LSTM network, and predicts the future system state based on the current system state, control input, and environmental parameters, and plans the control sequence in advance. The inner loop uses an adaptive sliding mode controller to adjust the control quantity in real time according to the error and its change rate, and correlates the switching gain with the occlusion confidence to enhance the robustness in complex scenarios; this method and system rely on the deep fusion and accurate analysis of multi-modal data to endow the tracking turntable with excellent environmental adaptability and target tracking ability. It can effectively cope with complex environmental interferences such as atmospheric refraction, light changes, and occlusions, improve the tracking accuracy, and through the intelligent feedback optimization mechanism, the control strategy can be adaptively adjusted online, significantly enhancing the stability, reliability, and intelligence of the system, and providing an advanced and efficient technical solution for target tracking applications in multiple fields.

[0181] In order to enable the smooth operation of the above-mentioned method and system, in addition to the various modules described above, the system may further include more or fewer components than those described above, or combine certain components, or different components. For example, it may include input / output devices, network access devices, buses, processors, and memories, etc.

[0182] The so-called processor may be a central processing unit, or may also be other general-purpose processors, digital signal processors, application-specific integrated circuits, off-the-shelf programmable gate arrays, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The above-mentioned processor is the control center of the above-mentioned system, connecting various parts through various interfaces and lines.

[0183] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0184] The above-described embodiments merely represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the invention patent of the present invention shall be subject to the appended claims.

[0185] The above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for tracking, servo-identifying, adjusting and controlling a turntable, characterized in that The method includes: Multimodal acquisition and tracking of the data around the turntable; Obtaining and fusing the acquired data around the turntable; Establishing a three-dimensional space error model; Generating a predictive control strategy based on the three-dimensional space error model; Generating an intelligent feedback optimization result based on the predictive control strategy.

2. The follow-up turntable servo recognition adjustment control method according to claim 1, characterized in that, The multimodal acquisition and tracking of the data around the turntable specifically includes: Realizing synchronous acquisition based on a global shutter camera and a six-axis IMU; Measuring the angular velocity at the level of 0.001° / h through a fiber optic gyroscope, and synchronously obtaining the target image sequence, the pose data of the turntable, and the environmental parameters.

3. The tracking turntable follow-up identification adjustment control method according to claim 1, characterized in that The obtaining and fusing of the acquired data around the turntable specifically includes: Obtaining the target image sequence, the pose data of the turntable, and the environmental parameters; Extract the target vector parameters in the target grayscale image sequence, tracking turntable pose data, and environmental parameters respectively to generate a grayscale image feature vector , turntable state vector and environmental parameter vector ; Constructing a three-element mapping model of environment-target-platform; The specific three-element mapping model is: ； In the formula, is the azimuth deviation, is the pitch angle deviation, is the environmental compensation coefficient; Dynamic illumination compensation based on Retinex-Net, associated compensation coefficient and the gradient variance of the grayscale image; ； Wherein, is the variance of the gradient of the grayscale image, is the preset maximum variance of the gradient of the grayscale image, is the preset adjustment parameter, is the light intensity; The spatio-temporal context model is adopted to fuse the information of five consecutive frames of target images, construct an occlusion confidence prediction matrix, and output the occlusion confidence at the current moment : ； wherein, is the occlusion confidence of the target at the current moment, and the value range is , , is the weight coefficient, is the intersection over union of the th frame, is the maximum intersection over union.

4. The tracking turntable follow-up identification adjustment control method according to claim 1, characterized in that, The establishment of the three-dimensional space error model specifically includes: Establishing a three-dimensional space error model including atmospheric refraction correction; The calculation process of the three-dimensional space error model is: ； ； In the formula, is the total azimuth error, is the true azimuth, is the currently measured azimuth of the tracking turntable, is the azimuth correction term caused by atmospheric refraction, is the atmospheric refractive index, is the preset target elevation angle.

5. The tracking turntable servo recognition and adjustment control method according to claim 3, characterized in that The generation of the predictive control strategy based on the three-dimensional space error model specifically includes: Extract the state vector of the turntable from the actual state vector ; Obtain the control input vector of the tracking turntable based on the control command of the tracking turntable at the previous moment ; Predict the future state of the outer loop based on LSTM to generate an outer loop neural network prediction model ; The outer ring neural network prediction model Specifically include: ； In the formula, is the predicted value of the turntable state at the future moment, is the actual state vector of the turntable at the current moment, is the control input vector at the current moment, is the environmental parameter vector at the current moment; Based on the outer-loop neural network prediction model , calculate the adjustment parameters of the inner-loop adaptive sliding mode control ; The calculation process of the parameters adjusted by the inner-loop adaptive sliding mode control is: ； wherein, is the final output control quantity, is the proportional gain, is the differential gain, is the system error for tracking the target state and the actual state of the turntable, is the switching gain, is the sign function, which is used to determine the direction of the control quantity based on the sign of the sliding mode surface ; Associated switching gain and occlusion confidence ; The specific association process is: ； Wherein, is a preset initial switching gain.

6. The follow-up turntable servo recognition adjustment control method according to claim 1, characterized in that The generation of the intelligent feedback optimization result based on the predictive control strategy specifically includes: Implementing a three-closed-loop collaborative control, and the specific three-closed-loop collaborative control includes: Position loop: Realizing target trajectory prediction based on Kalman filtering; Velocity loop: Variable parameter PID control; Disturbance compensation loop: A disturbance observer is used to compensate for the friction torque in real time ; The frictional torque The specific calculation process is as follows: ； In the formula, is the static friction coefficient, is the dynamic friction coefficient, is the rotational speed of the motor of the tracking turntable.

7. A tracking turntable servo recognition and adjustment control system, characterized in that, Applying the follow-up identification and adjustment control method of the turntable according to any one of claims 1-6, the system includes: A multimodal acquisition module for multimodal acquisition of the data around the turntable; An acquisition and fusion module for obtaining and fusing the acquired data around the turntable; A three-dimensional space error model module for establishing a three-dimensional space error model; A first generation module for generating a predictive control strategy based on the three-dimensional space error model; A second generation module for generating an intelligent feedback optimization result based on the predictive control strategy.

8. The tracking turntable servo recognition and adjustment control system according to claim 7, characterized in that, The multimodal acquisition module specifically includes: A synchronous acquisition unit for realizing synchronous acquisition based on a global shutter camera and a six-axis IMU; A measurement unit for measuring the angular velocity at the level of 0.001° / h through a fiber optic gyroscope; A synchronous acquisition unit for synchronously obtaining the target image sequence, the pose data of the turntable, and the environmental parameters.

9. The tracking turntable servo recognition and adjustment control system according to claim 7, characterized in that, The acquisition and fusion module specifically includes: A first acquisition unit for obtaining the target image sequence, the pose data of the turntable, and the environmental parameters; An extraction unit for respectively extracting the target vector parameters in the target gray image sequence, the pose data of the turntable, and the environmental parameters; The first generation unit is used to generate a grayscale image feature vector , a turntable state vector and an environmental parameter vector ; A three-element mapping model unit for constructing a three-element mapping model of environment-target-platform; The first correlation unit is used to implement dynamic light compensation based on Retinex-Net and correlate the compensation coefficient with the variance of the gradient of the grayscale image; A fusion unit for fusing the information of five consecutive frames of target images by using a spatio-temporal context model; A prediction matrix unit for constructing an occlusion confidence prediction matrix; An output unit for outputting the occlusion confidence at the current moment .

10. The tracking turntable servo recognition and adjustment control system according to claim 8, characterized in that, The first generation module specifically includes: An extraction unit for extracting the actual state vector in the turntable state vector from the actual state vector ; A second acquisition unit, configured to obtain a control input vector of the tracking turntable based on a control command of the tracking turntable at a previous moment ; A second generation unit, configured to generate an outer loop neural network prediction model based on an LSTM prediction of a future state of the outer loop ; A calculation unit, configured to calculate an inner-loop adaptive sliding mode control adjustment parameter based on an outer-loop neural network prediction model ; ​ The second association unit is used to associate and switch gains with the occlusion confidence .

Images