An open-loop target locking and tracking system and method based on hybrid stepper motors

Abstract

The present application relates to a kind of open-loop target locking and tracking system and method based on hybrid stepper motor, camera acquires target image, and camera is rotated by hybrid stepper motor and is driven, realize target locking and tracking, method includes image processing and target detection, prediction and filtering, position deviation calculation, step angle mapping and control.The system of the present application uses this method, including image processing module, prediction and filtering module, position deviation calculation module and step control module.The present application combines image recognition and motor control technology, realize stable, efficient locking and tracking to target object under the condition of not needing position feedback.Compared with the traditional scheme based on servo motor, it greatly reduces the system cost, is suitable for cost-sensitive and for the control precision requirement moderate target tracking scene, is suitable for intelligent camera, robot vision tracking, automatic aiming system and other application scenarios, especially suitable for high cost requirement consumer products.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automatic control and intelligent tracking, and particularly relates to an open-loop target locking and tracking system and method based on a hybrid stepper motor. BACKGROUND

[0002] In the fields of intelligent tracking systems, automated detection, consumer-level security, and other modern intelligent fields, target locking and tracking technology is a key support for realizing the core functions of the system, and its performance directly determines the intelligent level and application effect of the system. Currently, there are two main technical solutions for realizing target locking and tracking functions. One is a closed-loop control system solution based on a servo motor, and the other is an open-loop control solution.

[0003] The closed-loop control system solution based on a servo motor combines a servo motor with a closed-loop control logic, which can meet the technical requirements of high-precision positioning and high-response speed to some extent, and therefore is widely used in scenarios with high performance requirements. However, with the rapid development of low-cost, lightweight, and embedded intelligent devices, the traditional target tracking solution based on servo motor closed-loop control has gradually exposed some technical defects that are difficult to overcome, as follows.

[0004] 1. High hardware cost: High-performance servo motors, as the core execution components, have high manufacturing costs, resulting in high hardware costs for the entire system, which makes it difficult to meet the needs of low-cost application scenarios such as consumer-level applications.

[0005] 2. Complex system structure: The solution relies on a complex encoder feedback mechanism to achieve closed-loop control, which not only increases the complexity of the hardware structure but also makes the system debugging and maintenance process more difficult, reducing the practicality of the system.

[0006] 3. Large controller resource consumption: The closed-loop control logic has strict requirements for the performance indicators such as the operation capacity and response speed of the motor controller, which results in the controller needing to occupy a large amount of hardware resources, limiting its application in resource-constrained embedded platforms.

[0007] 4. Poor scene adaptability: The solution is inherently difficult to adapt to low-cost, lightweight embedded platforms, and cannot meet the integration needs of target tracking functions for such platforms.

[0008] The traditional open-loop control solution also has some applications in some scenarios, but it also has the following shortcomings:

[0009] 1. Insufficient dynamic tracking performance: Due to the use of a fixed step frequency control strategy, the system is prone to step loss when facing fast-moving targets, resulting in target tracking failure.

[0010] 2. Control delay is obvious: lack of effective target position prediction mechanism, the generation and execution of control instructions have significant delay, and real-time accurate tracking of the target cannot be realized.

[0011] 3. Low energy utilization efficiency: without adopting a dynamic adjustment strategy to match different target motion states, the motor operates in a non-optimal working condition, the system energy efficiency is low, and it does not meet the energy-saving and environmental protection technology development trend.

[0012] Therefore, in view of the above defects of the existing target tracking scheme based on servo motor closed-loop control, a low-cost, low-complexity, high adaptability and dynamic tracking demand meeting target locking and tracking technology scheme is needed to promote the further development and application of intelligent tracking system in embedded and other low-cost fields. SUMMARY

[0013] The technical problem to be solved by the present application is: in order to overcome the technical problem of how to realize high reliability target tracking function without encoder feedback in an open-loop control system, the present application provides an open-loop target locking and tracking system and method based on a hybrid stepping motor, which combines image recognition and motor control technology to realize stable and efficient locking and tracking of target objects without position feedback. Compared with the traditional servo motor based scheme, the system cost is greatly reduced, which is suitable for target tracking scenes that are sensitive to cost and have moderate control accuracy requirements, and is suitable for application scenes such as intelligent camera, robot vision tracking, automatic aiming system, etc., especially for consumer-grade products with high cost requirements.

[0014] The technical solution adopted by the present application to solve the technical problem is: an open-loop target locking and tracking method based on a hybrid stepping motor, a camera captures target images, and the camera is driven to rotate by a hybrid stepping motor to realize target locking and tracking, comprising the following steps:

[0015] Step 1, image processing and target detection: a lightweight YOLOv network is used for target detection of images captured by the camera, and a Sigmoid function is used for target confidence calculation in the output layer;

[0016] Step 2, prediction and filtering: predict the target motion state based on a Kalman filter, and use image prediction lead time to compensate for control instruction execution delay through pre-judgment of the target motion trajectory;

[0017] Step 3, position deviation calculation: establish a mapping between target coordinates and hybrid stepping motor angle, and calculate the deviation amount to get the angle that the hybrid stepping motor needs to compensate for rotation through the mapping algorithm;

[0018] Step 4, step angle mapping and control: dynamically switch microstep mode according to deviation, generate pulse interval of hybrid stepper motor, and adaptively adjust step frequency based on target speed.

[0019] In step 1, a lightweight YOLOv3-tiny network is used for target detection, the input resolution is 416*416, dynamic frame interval processing is used, and 8-bit integer quantization network parameters are used.

[0020] In step 1, the target confidence calculation formula is Pc=1 / (1+e -z ), where z is the network output value; the target screening condition is: confidence threshold Pc th =0.7, non-maximum suppression IoU threshold =0.5, and minimum detection box area =1 / 100 of image area.

[0021] In step 2, the Kalman filter state equation is:

[0022] x k =F k ·x k-1 +B k ·u k +w k ;

[0023] z k =H k ·x k +v k ;

[0024] Where:

[0025] x k : the system state vector at time k, i.e. the estimated quantity;

[0026] x k-1 : a column vector containing all state variables of the system at time k-1;

[0027] F k : state transition matrix, describing how the state evolves from time k-1 to time k;

[0028] B k : control input matrix, mapping external control input u k , such as motor instructions, external forces, etc., to the state space; this item is optional, depending on whether the system has known external control input, if the system has no known external control input, or the external input is too complex and random to be accurately modeled, this item can be omitted.

[0029] w k : process noise, assumed to be Gaussian white noise with covariance Q k , representing the uncertainty of the model;

[0030] Q k : Process Noise Covariance Matrix, used to quantify the uncertainty or error in the model prediction process;

[0031] z k : k : Observation vector at time k, i.e. sensor measurement;

[0032] H k : Observation matrix, mapping state space to observation space;

[0033] v k : Observation noise, assumed to be Gaussian white noise with covariance R k , representing sensor error;

[0034] State vector x = [p x , p y , v x , v y ] T , (p x , p y ) is the target position state vector, (v x , v y ) is the velocity state vector;

[0035] Process noise covariance Q = diag([0.1, 0.1, 0.5, 0.5]);

[0036] Observation noise covariance R = diag([5, 5]).

[0037] In step 2:

[0038] When the target velocity > 10 pixels / frame, increase the velocity component noise covariance in the Q matrix;

[0039] Through phase current waveform anomaly detection, when the motor is out of step, reset the Kalman filter covariance matrix.

[0040] In step 3, the mapping algorithm is:

[0041] Delta theta = k * arctan(Delta x / f);

[0042] Where Delta theta is the angle the motor needs to turn; Delta x is the lateral offset of the target in the image; f is the equivalent focal length of the camera in pixels, not physical millimeters unless converted; k is the calibration coefficient, obtained through calibration.

[0043] In step 4, the pulse interval T = T0 / (1 + alpha * |Delta theta|);

[0044] Wherein, T0 is the basic pulse interval, and a is the acceleration factor;

[0045] The dynamic micro-step adjustment logic is:

[0046] When |Δθ|>5°, the micro-step subdivision mode is enabled;

[0047] When 1°<|Δθ|≤5°, the half-step mode is used;

[0048] When |Δθ|≤1°, the full-step mode is used.

[0049] In step 4, it also includes:

[0050] Speed feedforward compensation: when the rate of change of angle Δθ>10° / s, the step frequency is increased by 20% in advance;

[0051] Step-out detection: by monitoring the motor phase current waveform, when the current is abnormal, the step frequency is automatically reduced by 30%.

[0052] The step frequency adaptive adjustment formula is:

[0053] f new =f old +β·(Δθ / θ max ) ·f max ;

[0054] Wherein, f new represents the adjusted target step frequency, i.e. the future value, which is calculated in real time according to the formula;

[0055] f old represents the step frequency before adjustment, i.e. the historical value, which is updated after each iteration;

[0056] θ max represents the control output, such as the maximum allowed value of motor rotation angle or rudder deflection;

[0057] The constraint condition is: f min ≤f new ≤f max ;

[0058] Wherein: β is the adjustment coefficient, 0.3-0.7; f max is the maximum allowed step frequency of the motor; and f min is the minimum step frequency to ensure torque.

[0059] The open-loop target locking and tracking system based on the hybrid stepping motor of the application adopts the open-loop target locking and tracking method based on the hybrid stepping motor, and comprises:

[0060] Image processing module: a lightweight YOLOv network is used for target detection on the image collected by the camera, and the output layer uses a Sigmoid function to calculate the target confidence;

[0061] Prediction and filtering module: predict the target motion state based on Kalman filter;

[0062] Position deviation calculation module: establish a mapping based on the target motion state and the target position, and calculate the deviation amount to obtain the angle of rotation that needs to be compensated by the hybrid stepping motor through the mapping algorithm;

[0063] Stepping control module: dynamically switch the micro-step mode according to the deviation amount, generate pulse intervals, and adaptively adjust the step frequency based on the target speed.

[0064] The open-loop target locking and tracking system and method based on the hybrid stepping motor has the following remarkable beneficial effects:

[0065] 1. Greatly reduce system cost and improve performance-price ratio advantage: the open-loop control architecture is built by using the hybrid stepping motor, without the need for expensive high-performance servo motor and complex encoder feedback components in the traditional servo motor closed-loop system, thereby reducing the system cost from the hardware source. The controller resource demand of the present application is low, and it can be realized based on a middle-end processor such as STM32H743, without the need for high-performance hardware support; the lightweight design makes it well adapted to low-cost, lightweight embedded platforms, and it is suitable for target tracking scenarios that are sensitive to cost and have moderate control precision requirements, suitable for application scenarios such as smart cameras, robot vision tracking, automatic aiming systems, etc., especially for consumer-grade products with high cost requirements.

[0066] 2. Simplify system structure and reduce development and maintenance difficulty: the present application replaces physical position feedback with image recognition, eliminating the need for hardware deployment and debugging of the encoder. At the same time, the lightweight YOLOv3-tiny network and optimized control logic reduce the resource demand on the controller, making the system debugging and maintenance process more convenient.

[0067] 3. Improve dynamic tracking precision, reduce delay and error: through the Kalman filtering algorithm of the prediction and filtering module, the present application can accurately estimate the target position and speed, and compensate for control delay based on image prediction, meeting the precision requirements of most consumer-level scenarios.

[0068] 4. Adaptive matching target motion state, optimizing dynamic tracking performance: With the help of step frequency adaptive adjustment mechanism, the application can adjust the step frequency parameters of the stepper motor in real time according to the target motion state (such as motion speed, acceleration, etc.). When facing fast-moving targets, the system can automatically increase the step frequency to enhance the response speed, avoiding the out-of-step phenomenon caused by fixed step frequency in traditional open-loop control; when the target moves slowly or is stationary, the step frequency can be reduced to reduce resource consumption, effectively solving the problem of insufficient dynamic tracking performance, while balancing tracking efficiency and resource utilization.

[0069] 5. Optimize step frequency adjustment strategy, reduce out-of-step phenomenon: Based on the dynamic step frequency adjustment mechanism, the application can adaptively adjust the step frequency parameters according to the target motion state, and enhance the fast response ability through speed feedforward compensation. At the same time, according to the deviation, the microstep mode is automatically switched, combined with current adaptive adjustment, effectively solving the problem of easy out-of-step when tracking fast targets caused by fixed step frequency in traditional open-loop scheme, reducing high-speed tracking delay.

[0070] 6. Realize soft out-of-step detection, guarantee tracking stability: By monitoring the motor phase current waveform, the application can detect soft out-of-step in real time, and automatically reduce the step frequency by 30% and trigger the adjustment strategy when the current is abnormal. This mechanism fills the technical gap that traditional open-loop systems cannot perceive the out-of-step state, avoids tracking failure caused by out-of-step accumulation, and significantly improves the running stability of the system in complex scenes such as fast-moving targets.

[0071] 7. Optimize energy utilization efficiency, meet the trend of technological development: Through dynamic microstep adjustment and current adaptive strategy, the application makes the motor always work in the optimal working condition, avoiding the energy waste caused by fixed control parameters in traditional schemes. Under the premise of ensuring tracking performance, the energy utilization efficiency of the system is significantly improved, which meets the modern technological development trend of energy saving and environmental protection. BRIEF DESCRIPTION OF DRAWINGS

[0072] Figure 1 is the flowchart of the open-loop target locking and tracking method based on the hybrid stepper motor of the application. DETAILED DESCRIPTION

[0073] The application will now be further described in conjunction with the accompanying drawings. It should be emphasized that the following description is only exemplary and is not intended to limit the scope of the application and its applications.

[0074] As shown in Figure 1 A kind of open-loop target locking and tracking method based on the hybrid stepper motor of the application, camera captures target image, and camera is driven to rotate by hybrid stepper motor, realizes target locking and tracking, including the following steps:

[0075] Step 1, image processing and target detection: a lightweight YOLOv network is used for target detection on the images collected by the camera, and the output layer uses a Sigmoid function to calculate the target confidence;

[0076] Step 2, prediction and filtering: the Kalman filter is used to predict the target motion state, and the image prediction lead time is used to compensate for the control command execution delay through the prediction of the target motion trajectory;

[0077] Step 3, position deviation calculation: a mapping between the target coordinates and the hybrid stepper motor angle is established, and the deviation amount is calculated to obtain the angle that the hybrid stepper motor needs to compensate for rotation;

[0078] Step 4, step angle mapping and control: the microstep mode is dynamically switched according to the deviation amount, the pulse interval of the hybrid stepper motor is generated, and the step frequency is adaptively adjusted based on the target speed. The microstep number is automatically switched according to the deviation amount, which can reduce power consumption by 30% while maintaining accuracy. When the deviation is large, the driving current is increased, for example, from 60% to 80%, to prevent step loss, and when the deviation is small, the current is reduced to reduce heat generation.

[0079] In step 1, a lightweight YOLOv3-tiny network is used for target detection, the input resolution is 416x416, dynamic frame interval processing is used, still targets are processed every 5 frames, and 8-bit integer quantization network parameters are used.

[0080] In step 1, the target confidence calculation formula is Pc=1 / (1+e -z ), where z is the network output value; the target screening conditions are: confidence threshold Pc th =0.7, non-maximum suppression IoU threshold =0.5, and minimum detection box area =1 / 100 of the image area.

[0081] In step 2, the Kalman filter state equation is:

[0082] x k =F k ·x k-1 +B k ·u k +w k ;

[0083] z k =H k ·x k +v k ;

[0084] Where:

[0085] x k : the system state vector at time k, i.e. the estimated quantity;

[0086] xk-1 : is a column vector containing all state variables of the system at time k-1;

[0087] F k : state transition matrix, describing how the state evolves from time k-1 to time k;

[0088] B k : control input matrix, mapping external control input u k to the state space;

[0089] w k : process noise, assumed to be Gaussian white noise with covariance Q k , representing the uncertainty of the model;

[0090] Q k : process noise covariance matrix, used to quantify the uncertainty or error in the model prediction;

[0091] z k : k observation vector at time k, i.e., the sensor measurement;

[0092] H k : observation matrix, mapping the state space to the observation space;

[0093] v k : observation noise, assumed to be Gaussian white noise with covariance R k , representing the sensor error;

[0094] State variables x = [p x , p y , v x , v y ] T , (p x , p y ) are the target position state variables, and (v x , v y ) are the velocity state variables;

[0095] Process noise covariance Q = diag([0.1, 0.1, 0.5, 0.5]);

[0096] Observation noise covariance R = diag([5, 5]).

[0097] In step 2:

[0098] When the target velocity > 10 pixels / frame, increase the noise covariance of the velocity component in the Q matrix;

[0099] Through phase current waveform anomaly detection, when the motor is out of step, reset the Kalman filter covariance matrix.

[0100] In step 3, the mapping algorithm is:

[0101] Δθ = k · arctan(Δx / f);

[0102] Δθ is the angle (radian) the motor needs to rotate; Δx is the lateral offset (pixel) of the target in the image; f is the equivalent focal length of the camera (pixel unit), typical value f = 600; k is the calibration coefficient, range 0.8-1.2, obtained by calibration. Among them, the parameter range is: maximum tracking deviation: ±30°, corresponding to the image edge; minimum step angle of motor half-step mode: 0.072°; dead zone setting: ±3 pixels, about ±0.3°, when the deviation is in the dead zone, the motor action is not triggered.

[0103] In step 4, the pulse interval T = T0 / (1+α·|Δθ|);

[0104] Where T0 is the basic pulse interval, corresponding to the rated speed of the motor, and α is the acceleration coefficient, typical value 0.05-0.2;

[0105] The dynamic micro-step adjustment logic is:

[0106] When |Δθ|>5°, the micro-step subdivision mode (32 subdivisions) is enabled;

[0107] When 1°<|Δθ|≤5°, the half-step mode is used;

[0108] When |Δθ|≤1°, the full-step mode is used.

[0109] In step 4, it also includes:

[0110] Speed feedforward compensation: when the rate of change of angle Δθ>10° / s, increase the step frequency by 20% in advance;

[0111] Out-of-step detection: by monitoring the motor phase current waveform, when the current is abnormal, automatically reduce the step frequency by 30%.

[0112] The step frequency adaptive adjustment formula is:

[0113] f new =f old +β·(Δθ / θ max ) ·f max ;

[0114] Where f new represents the adjusted target step frequency, i.e. the future value, calculated in real time according to the formula;

[0115] f old represents the step frequency before adjustment, i.e. the historical value, updated after each iteration;

[0116] θmax representing the control output, such as the maximum allowed value of the motor rotation angle, rudder deflection, etc.

[0117] The constraint condition is: f min ≤ f new ≤ f max ;

[0118] Wherein: β is the adjustment coefficient, 0.3-0.7; f max is the maximum allowed step frequency of the motor, such as 10 kHz; f min is the minimum step frequency to ensure torque, such as 1 kHz.

[0119] The open-loop target locking and tracking system based on the hybrid stepping motor of the application adopts the open-loop target locking and tracking method based on the hybrid stepping motor of the application, comprising:

[0120] The image processing module: adopts a lightweight YOLOv network to detect the target in the image collected by the camera, and uses a Sigmoid function to calculate the target confidence in the output layer;

[0121] The prediction and filtering module: predicts the target motion state based on a Kalman filter;

[0122] The position deviation calculation module: establishes a mapping based on the target motion state and the target position, and calculates the deviation amount to obtain the angle of rotation that needs to be compensated by the hybrid stepping motor through the mapping algorithm;

[0123] The stepping control module: dynamically switches the micro-step mode according to the deviation amount, generates pulse intervals, and adaptively adjusts the step frequency based on the target speed.

[0124] Embodiment

[0125] This embodiment verifies the actual effect of the open-loop target locking and tracking method and system based on the hybrid stepping motor of the application based on an intelligent cloud head camera system. Through hardware configuration optimization, scene testing and performance comparison, the technical advantages of the algorithm in low-cost and high-reliability tracking scenarios are fully demonstrated.

[0126] Hardware configuration and module adaptation:

[0127] Core execution component: hybrid stepping motor 17HS19-2004S1, step angle 1.8°, holding torque 0.4N·m, cooperates with TMC5160 driver to realize precise driving. The driver supports 256 micro-step subdivision, can adapt to the dynamic micro-step adjustment strategy of enabling 32 subdivision micro-step mode when |Δθ|>5°, using half-step mode when 1°<|Δθ|≤5°, and using full-step mode when |Δθ|≤1°, and realizes the control logic of increasing the driving current from 60% to 80% to prevent step loss when the deviation is large, and reducing the current to reduce heat when the deviation is small.

[0128] Processing core: STM32H743 processor (480 MHz clock frequency) is selected, responsible for running the algorithm full-process module. Its hardware computing power can support real-time inference of YOLOv3-tiny lightweight target detection network, input resolution 416x416, while meeting the real-time requirements of Kalman filter state estimation, position deviation calculation and dynamic step frequency adjustment.

[0129] Visual acquisition module: OV5640 camera is used, 5 million pixels, 60 fps frame rate, providing target motion image data. After calibration of the camera intrinsic parameters, the focal length parameter f is set to 600 pixel units, providing accurate input parameters for the mapping algorithm of the position deviation calculation module Δθ=k·arctan(Δx / f), where the calibration coefficient k is determined to be 1.0 through field calibration.

[0130] The test scene is a personnel tracking scene, and the specific test conditions are as follows:

[0131] Target characteristics: walking human body, motion speed range 0.5-1.5 m / s (corresponding to target lateral movement speed 5-30 pixels / frame in image), including uniform walking, acceleration, turning and other dynamic behaviors.

[0132] Environmental parameters: target distance from camera 3-5 meters, light intensity 300-1000 lux (covering typical scenes such as overcast indoor, sunny shade, etc.), avoiding strong light direct radiation or dark light noise interference.

[0133] Core control flow:

[0134] Image processing stage: after the camera outputs each frame of image, it is preprocessed by STM32H743 and input into the YOLOv3-tiny network. For stationary targets (position change ≤3 pixels for 3 consecutive frames), dynamic frame interval processing is adopted, detecting every 5 frames, and the calculation amount is reduced through 8-bit integer quantization network parameters. Target screening strictly follows the conditions of confidence threshold 0.7, IoU threshold 0.5 and minimum detection box area ≥ image area 1 / 100, ensuring that the target detection accuracy is ≥95%.

[0135] Prediction and filtering stage: based on Kalman filter, the target state x=[p x ,p y ,v x ,v y ] T is estimated. When the target speed > 10 pixels / frame, corresponding to actual motion speed > 1 m / s, the speed component in the process noise covariance Q is increased from 0.5 to 1.0, improving the dynamic response sensitivity; if the motor is out of step through current monitoring, the phase current waveform abnormal fluctuation > 20%, immediately reset the covariance matrix to avoid error accumulation.

[0136] Bias calculation and driving control stage: according to the lateral offset of the target in the image, the motor rotation angle is calculated by mapping algorithm, the dead zone is set to ±3 pixels, corresponding to ±0.3°, to reduce frequent start and stop. The step pulse generation adopts the pulse interval formula T=T0 / (1+α·|Δθ|)(α=0.1), and the dynamic step frequency adjustment follows f new =f old +β·(Δθ / θ max )·f max , where β=0.5, f max =10kHz, f min =1kHz, when the change rate of angle Δθ>10° / s, increase 20% step frequency in advance, the high-speed tracking delay is reduced from 150ms of traditional open-loop scheme to 90ms.

[0137] Performance index comparison and analysis:

[0138]

[0139] The actual test verification of the intelligent cloud head camera system of the embodiment shows that the method can realize stable target locking and tracking function on a low-cost hardware platform. The dynamic step frequency adjustment, prediction compensation and soft loss of step detection mechanism effectively overcome the defects of the traditional open-loop scheme, while the cost and power consumption are greatly reduced compared with the servo closed-loop scheme, and it is especially suitable for consumer-level security, home intelligent camera and other scenes sensitive to cost and moderate precision requirements.

[0140] Based on the above ideal embodiments according to the application, through the above description, relevant personnel can make various changes and modifications without deviating from the technical idea of the application. The technical scope of the application is not limited to the contents of the specification, and the technical scope must be determined according to the scope of the claims.

Claims

1. An open loop target locking and tracking method based on hybrid stepper motor, characterized by, The camera captures target images, and the camera is driven to rotate by a hybrid stepper motor to realize target locking and tracking, including the following steps: Step 1, image processing and target detection: a lightweight YOLOv network is used to detect the target in the image captured by the camera, and the output layer uses a Sigmoid function to calculate the target confidence; Step 2, prediction and filtering: the Kalman filter is used to predict the target motion state, and the image prediction lead time is used to compensate for the control instruction execution delay through the prediction of the target motion trajectory; Step 3, position deviation calculation: a mapping between the target coordinates and the angle of the hybrid stepper motor is established, and the deviation is calculated to obtain the angle of the hybrid stepper motor that needs to be compensated for rotation; Step 4, step angle mapping and control: the microstep mode is dynamically switched according to the deviation, the pulse interval of the hybrid stepper motor is generated, and the step frequency is adaptively adjusted based on the target speed; In step 3, the mapping algorithm is: Δθ=k·arctan(Δx / f); Where Δθ is the angle that the motor needs to rotate; Δx is the horizontal offset of the target in the image; f is the equivalent focal length of the camera, in pixels; k is the calibration coefficient, obtained through calibration; In step 4, the pulse interval T=T0 / (1+α·|Δθ|); Where T0 is the basic pulse interval, and α is the acceleration coefficient; The dynamic microstep adjustment logic is: When |Δθ|>5°, the microstep subdivision mode is enabled; When 1°<|Δθ|≤5°, the half-step mode is used; When |Δθ|≤1°, the full-step mode is used; The step frequency adaptive adjustment formula is: f new =f old +β·(Δθ / θ max ) ·f max ; where f new represents the adjusted target step frequency, which is calculated in real time according to the formula f old represents the step size before adjustment, which is updated after each iteration; θ max represents the maximum allowed value of the control output; The constraint is: f min ≤ f new ≤ f max ; Wherein: β is the adjustment coefficient, 0.3-0.7; f max is the maximum step frequency allowed for the motor; f min is the minimum step frequency to ensure torque.

2. The open loop target lock and track method based on hybrid stepper motor as claimed in claim 1 wherein, In step 1, a lightweight YOLOv3-tiny network is used for target detection, the input resolution is 416×416, dynamic frame interval processing is used, and 8-bit integer quantization network parameters are used.

3. The open loop target lock and track method based on hybrid stepper motor as claimed in claim 1 wherein, In step 1, the target confidence calculation formula is Pc = 1 / (1+e -z ), where z is the network output value; Target screening conditions are: confidence threshold Pc th = 0.7, non-maximum suppression IoU threshold = 0.5, minimum detection frame area = 1 / 100 of the image area.

4. The hybrid stepper motor based open loop target locking and tracking method as claimed in claim 1, wherein, In step 2, the state equation of the Kalman filter is: x k =F k ·x k-1 +B k ·u k +w k ; z k =H k ·x k +v k ; Where: x k : system state vector at time k, i.e. the quantity to be estimated; x k-1 : is a column vector containing all the state variables of the system at time k - 1; F k : state transition matrix, describing how the state evolves from time k-1 to time k; B k : control input matrix, optional quantity, maps external control input u k to state space; w k : process noise, assumed to be Gaussian white noise with covariance Q k , representing the uncertainty of the model; Q k : process noise covariance matrix, used to quantify the uncertainty or error in the model prediction process; z k : k the observation vector at the time instant, i.e. the sensor measurements; H k : observation matrix, mapping the state space to the observation space; v k : observation noise, assumed to be Gaussian white noise with covariance R k , represents sensor error; State variables x = [p x ,p y ,v x ,v y ] T , (p x ,p y ) are target position state variables, (v x ,v y ) are velocity state variables; The process noise covariance Q=diag([0.1,0.1,0.5,0.5]); The observation noise covariance R=diag([5,5]).

5. The open loop target lock and track method based on hybrid stepper motor as claimed in claim 4, wherein, In step 2: When the target speed is greater than 10 pixels / frame, increase the speed component noise covariance in the Q matrix; When the motor is out of step, reset the Kalman filter covariance matrix.

6. The hybrid stepper motor based open loop target locking and tracking method as claimed in claim 1, wherein, In step 4, it also includes: Speed feedforward compensation: when the rate of change of the angle Δθ is greater than 10° / s, increase the step frequency by 20% in advance; Step loss detection: by monitoring the motor phase current waveform, when the current is abnormal, automatically reduce the step frequency by 30%.

7. An open loop target locking and tracking system based on hybrid stepper motors, characterized by, The open-loop target locking and tracking method based on a hybrid stepper motor according to any one of claims 1-6, comprising: An image processing module: a lightweight YOLOv network is used to detect the target in the image captured by the camera, and the output layer uses a Sigmoid function to calculate the target confidence; A prediction and filtering module: the Kalman filter is used to predict the target motion state; A position deviation calculation module: a mapping based on the target motion state and the target position is established, and the deviation is calculated to obtain the angle of the hybrid stepper motor that needs to be compensated for rotation; A step control module: the microstep mode is dynamically switched according to the deviation, the pulse interval is generated, and the step frequency is adaptively adjusted based on the target speed.

Images