Swimming pool water speed self-adaptive control method and system based on human body detection

Abstract

The application relates to the technical field of pool control, in particular to a pool water speed self-adaptive control method and system based on human body detection. The method comprises the following steps: outputting an observation position and a detection confidence of a human body target according to a target detection model; predicting a predicted position according to a motion model, and calculating a difference between the observation position and the predicted position to obtain an observation deviation; weighting and statistically processing the observation deviation sequence by using the detection confidence to obtain a trend factor; determining an observation noise covariance according to the detection confidence, determining a process noise covariance according to the trend factor, calculating a Kalman gain, fusing the observation deviation into the predicted position by using the Kalman gain, and obtaining an optimal state; and adaptively adjusting the output power of a pool propeller according to the optimal state. Through the technical scheme, the pool propeller can output water flow dynamically matched with the swimming speed of a swimmer, and water speed self-adaptive control is realized.

Description

Technical Field

[0001] This application relates to the field of swimming pool control technology, and in particular to a swimming pool water velocity adaptive control method and system based on human body detection. Background Technology

[0002] With the widespread adoption of the national fitness concept, countercurrent swimming pools, as a new type of fitness equipment that provides an unlimited swimming experience within a limited space, have gradually entered high-end homes and professional training venues. As a core component of countercurrent swimming pools, the pool's water speed control system directly determines the user's swimming experience and training results. In complex underwater environments, how to adjust the propeller's output power according to the swimmer's real-time position and speed to achieve adaptive water speed control has become a pressing technical problem to be solved.

[0003] Existing adaptive pool water speed control typically employs methods based on visual detection or simple positional errors. A camera mounted above the pool captures the swimmer's position, and the deviation between this position and a preset position at the pool center is directly used as a feedback signal, inputting it into a PID controller or other logic controller. When the controller detects a swimmer deviating from the preset position, it linearly adjusts the propeller speed based on the magnitude and direction of the deviation, thereby controlling the pool water speed.

[0004] However, during swimming, the splashing water generated by the swimmer's strokes can severely interfere with the accuracy of visual detection, resulting in significant errors in the visual detection results. Furthermore, the swimmer's speed is not constant; the interference from splashing water, combined with sudden accelerations or decelerations, causes frequent fluctuations in the propeller's output power. This makes it impossible to capture the true movement trend while eliminating environmental interference, leading to substantial errors in pool water speed control. Summary of the Invention

[0005] To address the technical problem of large errors in pool water speed control, this application provides a pool water speed adaptive control method and system based on human body detection. This method ensures that the pool propeller can output water flow that dynamically matches the swimmer's swimming speed, achieving smooth and rapid adaptive water speed control.

[0006] In a first aspect, this application provides a swimming pool water speed adaptive control method based on human detection. The control method includes: real-time acquisition of swimming pool images and inputting them into a target detection model, outputting the observed position and detection confidence of the human target; inputting the optimal state of the previous moment into a motion model to predict the predicted position at the current moment, and calculating the difference between the observed position and the predicted position to obtain the observation deviation, wherein the optimal state includes the optimal position and the optimal speed; acquiring the observation deviation sequence within a preset time window, and using the detection confidence to perform weighted statistics on the observation deviation sequence to obtain a trend factor characterizing the true movement intention of the human body; determining the observation noise covariance based on the detection confidence, and determining the process noise covariance based on the trend factor, wherein the observation noise covariance is negatively correlated with the detection confidence, and the process noise covariance is positively correlated with the trend factor; calculating the Kalman gain based on the observation noise covariance and the process noise covariance, and using the Kalman gain to fuse the observation deviation into the predicted position to obtain the optimal state at the current moment; and adaptively adjusting the output power of the swimming pool propeller based on the optimal state at the current moment.

[0007] By introducing observation noise covariance negatively correlated with detection confidence and process noise covariance positively correlated with trend factor, the Kalman filter is improved. It can automatically shield unreliable observation positions to maintain stability when visual clarity decreases, and actively increase process noise when a clear motion trend is captured to quickly track the human body position, thus achieving precise adjustment of the output power of the pool propeller.

[0008] Preferably, the step of using detection confidence to perform weighted statistics on the observation deviation sequence to obtain a trend factor characterizing the true movement intention of the human body includes: within a preset time window, calculating the product of the observation deviation of each frame of pool image and the normalized detection confidence of that frame of pool image to obtain a weighted deviation; calculating the average value of all weighted deviations within the preset time window, and using the absolute value of the average value as the trend factor.

[0009] Because swimming strokes cause the body to sway back and forth periodically, the weighted statistics within the time window cause the alternating positive and negative invalid periodic fluctuations to cancel each other out, resulting in a result close to zero. When real acceleration occurs, the unidirectional continuous positional deviation is accumulated and amplified, allowing the trend factor to accurately determine the human body's true movement intention.

[0010] Preferably, the length of the preset time window is negatively correlated with the human swimming speed; the control method further includes: obtaining the optimal speed in the optimal state at the previous moment, and calculating the window length at the current moment based on the inverse relationship.

[0011] When swimming at a faster speed, the window length is automatically shortened to increase the refresh rate of the trend factor and improve the system's sensitivity to sprinting movements. When swimming at a slower speed, the window length is automatically extended to cover a longer stroke cycle, preventing large body movements during slow strokes from being misinterpreted as speed change signals. At the same time, it ensures low-latency response during high-speed swimming and improves adaptability to different swimming speeds.

[0012] Preferably, determining the observation noise covariance based on the detection confidence level includes: calculating the ratio of the baseline observation noise to the detection confidence level at the current time, and using the calculation result as the observation noise covariance at the current time.

[0013] When splashing water reduces the detection confidence, the observation noise covariance increases sharply. This causes the Kalman filter to reduce its confidence in the current observation position and rely more on the predicted position of the motion model, effectively shielding visual measurement abrupt changes caused by environmental noise.

[0014] Preferably, determining the process noise covariance based on the trend factor includes: calculating the product of the sensitivity coefficient and the trend factor at the current time, and adding the product to 1 to obtain the adjustment coefficient; and using the product of the baseline process noise and the adjustment coefficient as the process noise covariance at the current time.

[0015] When the trend factor increases significantly, indicating a genuine intention to accelerate, the process noise covariance is actively increased to inform the filter that the current uniform velocity prediction model has failed. This causes the filter to increase the response weight to new observation locations, quickly correcting prediction bias and thus avoiding control lag caused by sudden changes in human speed.

[0016] Preferably, the real-time acquisition of pool images and input into the target detection model to output the observation position of the human target includes: using the target detection model to identify the human bounding box in the pool image and extracting the center point of the human bounding box; using a preset perspective transformation matrix to map the center point from the image pixel coordinate system to the pool physical coordinate system to obtain the observation position of the human target.

[0017] Preferably, the target detection model is a lightweight target detection model.

[0018] Preferably, the step of inputting the optimal state of the previous moment into the motion model to predict the predicted position at the current moment includes: extracting the optimal position and optimal velocity from the optimal state of the previous moment; calculating the time interval between the current moment and the previous moment; and adding the product of the optimal position, the optimal velocity, and the time interval to obtain the predicted position at the current moment.

[0019] Preferably, the step of adaptively adjusting the output power of the pool propeller based on the optimal state at the current moment includes: inputting the optimal state at the current moment into the PID controller to obtain the target control value; and converting the target control value into the drive frequency of the frequency converter to drive the motor of the pool propeller.

[0020] In a second aspect, this application also provides a swimming pool water speed adaptive control system based on human body detection, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the swimming pool water speed adaptive control method based on human body detection according to the first aspect of this application.

[0021] The technical solution of this application has the following beneficial technical effects:

[0022] By dynamically adjusting the observation noise covariance using detection confidence, the weight of observations is automatically reduced when visual signals are unstable, such as when water splashes obscure the view. The process noise covariance is dynamically adjusted using a trend factor based on time window statistics, which improves the system response speed when a true acceleration intention is identified. Meanwhile, the positive and negative cancellation principle is used to filter oscillation signals within the normal planned water cycle. This effectively eliminates the influence of visual signal noise and changes in motion trends on water speed control, thereby ensuring that the pool propeller can output water flow that dynamically matches the swimmer's speed, achieving smooth and rapid adaptive water speed control. Attached Figure Description

[0023] Figure 1 This is a flowchart of a swimming pool water velocity adaptive control method based on human body detection according to an embodiment of this application.

[0024] Figure 2 This is a comparison chart of the errors of conventional Kalman filtering and improved Kalman filtering according to embodiments of this application.

[0025] Figure 3 This is a structural block diagram of a swimming pool water speed adaptive control system based on human body detection, according to an embodiment of this application. Detailed Implementation

[0026] According to a first aspect of this application, this application provides a swimming pool water velocity adaptive control method based on human body detection. Figure 1 This is a flowchart of a swimming pool water velocity adaptive control method based on human body detection, according to an embodiment of this application. Figure 1 As shown, the swimming pool water speed adaptive control method based on human detection includes steps S101 to S105, which are described in detail below.

[0027] S101 acquires pool images in real time and inputs them into the target detection model, outputting the observation position and detection confidence of the human target.

[0028] In one embodiment, the pool image is any frame of video data acquired in real time by a camera installed above or to the side of the counter-current pool. The observation position refers to the specific coordinates of the swimmer in the pool's physical spatial coordinate system, obtained from the pool image. The detection confidence score is a rating given by the target detection model to the credibility of the identified human target. It objectively reflects the clarity of the current water surface and the presence of interference such as splashes. The detection confidence score can be directly output by the target detection model, with one detection confidence score corresponding to one frame of the pool image.

[0029] To accurately obtain the observation position and detection confidence of human targets, pool images are acquired in real time and input into a target detection model, which outputs the observation position of human targets. This includes: using the target detection model to identify human bounding boxes in the pool images and extracting the center point of the human bounding boxes; using a preset perspective transformation matrix to map the center point from the image pixel coordinate system to the pool physical coordinate system to obtain the observation position of human targets.

[0030] It should be noted that since the position and orientation of the camera remain unchanged, camera calibration can be performed to obtain a preset perspective transformation matrix. By using the calibrated perspective transformation matrix for coordinate mapping, the center point of the human body bounding box in the pool image can be mapped to the pool's physical coordinate system.

[0031] Specifically, the target detection model employs a lightweight model. For example, the YOLOv8-Nano model is used. This model has low computational cost and fast inference speed, and can be deployed on embedded control boards with limited computing power, ensuring real-time processing capabilities for pool images and avoiding control delays.

[0032] In this way, the physical position of the human body in the pool was obtained, and the reliability of visual observation was quantified by detecting confidence, providing a data foundation for subsequent adaptive control of the pool water speed.

[0033] S102, the optimal state of the previous moment is input into the motion model to predict the predicted position at the current moment, and the difference between the observed position and the predicted position is calculated to obtain the observation deviation. The optimal state includes the optimal position and the optimal velocity.

[0034] In one embodiment, the optimal state is the best estimate obtained by fusing the predicted and observed positions using a Kalman filter algorithm. The optimal state includes both the optimal position and the optimal velocity. The predicted position is the theoretical position at the current moment predicted by the motion model based on the optimal state at the previous moment.

[0035] To achieve this prediction, inputting the optimal state from the previous moment into the motion model to predict the current position includes: extracting the optimal position and optimal velocity from the optimal state from the previous moment; calculating the time interval between the current moment and the previous moment; and adding the product of the optimal position, the optimal velocity, and the time interval to obtain the predicted position at the current moment.

[0036] Specifically, if we assume that the human body moves at a constant velocity in a straight line within adjacent moments, then at the current moment... Predicted location The calculation formula is:

[0037] ;in, For the previous moment The optimal position, For the previous moment The optimal speed, This represents the time interval between the current moment and the previous moment.

[0038] Then, the difference between the observed position and the predicted position is calculated to obtain the observation bias. Current time observation bias It characterizes the degree of deviation between the observed position determined by the image at the current moment and the predicted position, and its calculation formula is:

[0039] ;in, For the current moment The observation location For the current moment The predicted position; the observation bias includes measurement noise and abrupt changes in the actual human motion state.

[0040] Understandably, in the Kalman filter algorithm, the acquisition of the optimal state depends on both the predicted position and the observed position. However, the predicted position cannot be obtained at the initial moment. Therefore, the observed position can be used as the optimal state at the initial moment to calculate the optimal state at the next moment. The optimal state can truly reflect the human body's operating status.

[0041] Thus, by comparing the predicted position with the observed position, the observation bias, which includes changes in movement trends and environmental noise, is accurately extracted. This observation bias is used to determine the true movement intention of the human body.

[0042] S103: Obtain the observation deviation sequence within the preset time window; use the detection confidence level to perform weighted statistics on the observation deviation sequence to obtain the trend factor characterizing the true movement intention of the human body; determine the observation noise covariance based on the detection confidence level; and determine the process noise covariance based on the trend factor.

[0043] In one embodiment, the observation bias includes both changes in motion trend and environmental noise. Considering the periodic strokes and splashing interference unique to swimming, the detection confidence can indirectly reflect the environmental noise of the timestamp corresponding to the pool image. The greater the environmental noise, the lower the detection confidence. Therefore, in order to extract the true motion intention from the noisy observation bias, it is necessary to use the detection confidence to perform weighted statistics on the observation bias sequence.

[0044] The observation deviation sequence is weighted and statistically analyzed using the detection confidence level to obtain a trend factor that represents the true movement intention of the human body. This includes: within a preset time window, calculating the product of the observation deviation of each frame of pool image and the normalized detection confidence level of that frame of pool image to obtain the weighted deviation; calculating the average value of all weighted deviations within the preset time window, and using the absolute value of the average value as the trend factor.

[0045] Specifically, at the current moment Trend factors The calculation formula is:

[0046] ;

[0047] in, For the current moment The corresponding preset time window length, For the first time within the preset time window Observational bias in frame pool images For the first Normalized detection confidence corresponding to the frame pool image.

[0048] Among them, the The calculation process for the normalized detection confidence corresponding to the frame of the pool image is as follows: A preset time window includes multiple frames of pool images; the maximum detection confidence value among these multiple frames is obtained; the [value of the first frame is then set to the next frame]. The ratio of the detection confidence score to the maximum value of the frame pool image is used as the first... Normalized detection confidence corresponding to the frame pool image.

[0049] Understandably, observation bias can be positive, negative, or zero. The arm strokes during swimming cause periodic back-and-forth swaying of the body, resulting in alternating positive and negative observation biases within a preset time window. Summation can cancel out these invalid periodic fluctuations, bringing the result close to zero. However, when the user actually accelerates, the positional bias is unidirectional and continuous; summation accumulates this bias, causing... The impact is significantly increased. Simultaneously, detection confidence is introduced as a weight. When visual interference from water splashes leads to low confidence, the weight of the frame deviation is reduced, preventing environmental noise from interfering with the calculation of the trend factor.

[0050] When calculating the trend factor, in order to further adapt to the stroke frequency under different swimming speeds, the length of the preset time window is negatively correlated with the human swimming speed; the control method also includes: obtaining the optimal speed in the optimal state at the previous moment, and calculating the window length at the current moment based on the inverse relationship.

[0051] Specifically, the window length at the current moment. satisfy:

[0052] ;in, As a reference constant, The optimal velocity is the velocity in the optimal state at the previous moment. To prevent the default positive value with a denominator of zero, This represents the floor function.

[0053] The optimal speed in the optimal state at the previous moment When it increases, Automatically decreases and accelerates the refresh rate of trend factors, improving sensitivity to the acceleration process; when the optimal speed in the previous optimal state... When decreasing, It automatically increases in size to cover a longer stroke cycle, preventing body undulations during slow strokes from being misinterpreted as a speed change signal.

[0054] Secondly, in order for the Kalman filter to adapt to changes in environmental clarity, the observation noise covariance is determined based on the detection confidence level by: calculating the ratio of the baseline observation noise to the detection confidence level at the current time, and using the calculation result as the observation noise covariance at the current time.

[0055] Specifically, at the current moment Observation noise covariance satisfy:

[0056] ;in, The preset reference observation noise, For the current moment The detection confidence level, To prevent the default positive value with a denominator of zero, its value can be set to 1.

[0057] When the water splashes are large, the environmental noise is also large, which in turn affects the current moment. Detection confidence When it decreases, The sharp increase in position makes the Kalman filter realize that the current observation position is unreliable and that it should rely more on the predicted position, thereby shielding visual noise.

[0058] Finally, in order to solve the control lag problem, the process noise covariance is determined based on the trend factor, which includes: calculating the product of the sensitivity coefficient and the trend factor at the current time, and adding the product to 1 to obtain the adjustment coefficient; and using the product of the reference process noise and the adjustment coefficient as the process noise covariance at the current time.

[0059] Specifically, at the current moment Process noise covariance satisfy:

[0060] ;in, The preset reference process noise, The sensitivity coefficient, For the current moment Trend factors.

[0061] When trend factors When the process noise covariance in the Kalman filter is increased, the process noise covariance in the filter is actively increased. This allows the filter to recognize that the previous uniform velocity model has failed, meaning the predicted position is unreliable, thereby improving the response speed to the observed position and eliminating prediction lag.

[0062] In this way, adaptive adjustment of parameters in the Kalman filter is achieved, visual noise is shielded when there is a lot of water splash interference and environmental noise, and hysteresis is eliminated when the user changes speed. This makes the output of the Kalman filter more dependent on the observation position.

[0063] S104. Calculate the Kalman gain based on the observation noise covariance and the process noise covariance, and use the Kalman gain to fuse the observation bias into the predicted position to obtain the optimal state at the current time.

[0064] In one embodiment, the Kalman gain, a key parameter in the Kalman filter, determines whether to place more trust in the predicted location or the observed location when outputting the optimal state. This is based on the observation noise covariance. Process noise covariance It can adjust the Kalman gain in real time.

[0065] Specifically, at the current moment Kalman gain Satisfying the relation:

[0066] ;

[0067] in, Observation matrix transpose, The prediction error covariance matrix is ​​calculated using the process noise covariance. Update the prediction error covariance matrix. State transition matrix ,in It is the sampling time interval; The observation matrix is ​​used to observe the state vector. Extracting position components Therefore, the observation matrix for ; For the velocity component.

[0068] Subsequently, the state is updated using Kalman gain to obtain the optimal state at the current time step. :

[0069] ;in, To predict the position vector, Due to observational bias, For the current moment Kalman gain.

[0070] Please see Figure 2 This is a comparison chart of the errors of traditional Kalman filtering and improved Kalman filtering according to embodiments of this application. In the experimental scenario, the user carries a portable device such as a wristband with positioning function. The portable device is used to collect the user's real location, and the error is the difference between the optimal position in the optimal state and the real position. Figure 2 It can be seen that, in the absence of water splash interference and user acceleration, the errors of the traditional Kalman filter and the improved Kalman filter are basically the same. During the acceleration phase of user acceleration, the error of the improved Kalman filter is smaller than that of the traditional Kalman filter. In the water splash interference phase, the error of the improved Kalman filter is significantly smaller than that of the traditional Kalman filter. Therefore, the improved Kalman filter can accurately output the motion state of the human body.

[0071] Thus, we obtain the optimal state that accurately reflects the human body's motion state, which includes the optimal position and the optimal speed.

[0072] S105, adaptively adjusts the output power of the pool propeller based on the optimal state at the current moment.

[0073] In this embodiment, to achieve the effect of the water moving faster than the person and slower than the person moving, the calculated motion state needs to be converted into a specific motor control signal. Adaptively adjusting the output power of the pool propeller based on the optimal state at the current moment includes: inputting the optimal state at the current moment into a PID controller to obtain a target control value; and converting the target control value into the drive frequency of the frequency converter to drive the motor of the pool propeller.

[0074] Specifically, the optimal speed in the optimal state is extracted as a feedforward signal, which is used to achieve the adjustment effect of "faster person, faster water; slower person, slower water"; the deviation between the optimal position and the center line of the pool is extracted as a feedback signal, which is used to fine-tune the position of pulling the person back to the center line of the pool; and a target control value is calculated using a PID algorithm. Satisfying the relation:

[0075] ;

[0076] in, This is the speed feedforward coefficient, used to determine the ratio between motor frequency and water outlet velocity. For feedforward signal, This is the feedforward component corresponding to the feedforward signal; As a feedback signal, , and These are the proportional coefficient, integral coefficient, and derivative coefficient of the PID controller. The target control value is the sum of the feedforward component and the feedback component, corresponding to the feedback signal. This target control value is mapped to the frequency command of the frequency converter, adjusting the water outlet velocity of the laminar flow propeller so that the water flow velocity dynamically matches the swimmer's swimming speed.

[0077] In this way, when a swimmer accelerates, the system can detect the water speed in advance through trend factors and increase the water speed accordingly, achieving the effect of faster water speed when the swimmer is fast and slower water speed when the swimmer is slow. Furthermore, when the swimmer's position deviates, the system can fine-tune the system to pull the swimmer back to the center line of the pool. Thus, adaptive control of the pool water speed is achieved.

[0078] According to a second aspect of this application, this application also provides a swimming pool water speed adaptive control system based on human body detection. Figure 3 This is a structural block diagram of a swimming pool water velocity adaptive control system based on human body detection, according to an embodiment of this application. Figure 3 As shown, the system 50 includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the adaptive control method for pool water speed based on human body detection according to the first aspect of this application. The system also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface. Their configuration and functions are known in the art and will not be described further here.

[0079] It should be noted that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of this application, and these all fall within the scope of protection of this application.

Claims

1. A method for pool water speed adaptive control based on human detection, characterized in that, The control method includes: real-time acquisition of pool images and inputting them into a target detection model, outputting the observed position and detection confidence of the human target; inputting the optimal state of the previous moment into the motion model to predict the predicted position of the current moment, and calculating the difference between the observed position and the predicted position to obtain the observation deviation; the optimal state includes the optimal position and the optimal speed; the observed position refers to the specific coordinates of the swimmer in the physical space coordinate system of the pool obtained from the pool images; the detection confidence refers to the confidence score of the target detection model for the identified human target; Obtain the observation deviation sequence within a preset time window, and perform weighted statistics on the observation deviation sequence using detection confidence to obtain trend factors characterizing the true movement intention of the human body, including: Within a preset time window, the observation bias of each frame of pool image is calculated as the product of the normalized detection confidence of that frame of pool image to obtain the weighted bias; the average value of all weighted biases within the preset time window is calculated, and the absolute value of the average value is used as the trend factor. The observation noise covariance is determined based on the detection confidence level, and the process noise covariance is determined based on the trend factor. The observation noise covariance is negatively correlated with the detection confidence level, and the process noise covariance is positively correlated with the trend factor. Determining the observation noise covariance based on the detection confidence level includes: calculating the ratio of the baseline observation noise to the detection confidence level at the current time, and using the calculation result as the observation noise covariance at the current time. Determining the process noise covariance based on the trend factor includes: calculating the product of the sensitivity coefficient and the trend factor at the current time, and adding 1 to the product to obtain the adjustment coefficient; and using the product of the baseline process noise and the adjustment coefficient as the process noise covariance at the current time. The Kalman gain is calculated based on the observation noise covariance and the process noise covariance. The observation bias is then fused into the predicted position using the Kalman gain to obtain the optimal state at the current moment. The output power of the pool propeller is adaptively adjusted based on the optimal state at the current moment.

2. The adaptive control method for pool water velocity based on human body detection according to claim 1, characterized in that, The length of the preset time window is negatively correlated with the human swimming speed; the control method further includes: obtaining the optimal speed in the optimal state of the previous moment, and calculating the window length at the current moment based on the inverse relationship.

3. The adaptive control method for pool water velocity based on human body detection according to claim 1, characterized in that, The real-time acquisition of pool images and input into the target detection model, outputting the observation position of the human target, includes: using the target detection model to identify the human bounding box in the pool image and extracting the center point of the human bounding box; using a preset perspective transformation matrix to map the center point from the image pixel coordinate system to the pool physical coordinate system to obtain the observation position of the human target.

4. The adaptive control method for pool water velocity based on human body detection according to claim 1, characterized in that, The target detection model described is a lightweight target detection model.

5. The swimming pool water velocity adaptive control method based on human body detection according to claim 1, characterized in that, The step of inputting the optimal state of the previous moment into the motion model to predict the predicted position at the current moment includes: extracting the optimal position and optimal velocity from the optimal state of the previous moment; calculating the time interval between the current moment and the previous moment; and adding the product of the optimal position, the optimal velocity, and the time interval to obtain the predicted position at the current moment.

6. The adaptive control method for pool water velocity based on human body detection according to claim 1, characterized in that, The adaptive adjustment of the pool propeller's output power based on the optimal state at the current moment includes: inputting the optimal state at the current moment into the PID controller to obtain the target control value; converting the target control value into the drive frequency of the frequency converter to drive the motor of the pool propeller.

7. A swimming pool water velocity adaptive control system based on human body detection, characterized in that, It includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement the swimming pool water speed adaptive control method based on human body detection according to any one of claims 1 to 6.

Images