A first-view image stabilization method, device and software program product for a drone

Abstract

The application provides a UAV first-view image stabilizing method, device and software program product, relates to the UAV field, and solves the problem that the existing UAV first-view image stabilizing scheme cannot adapt to actual UAV business scenarios. The method comprises the following steps: acquiring and correcting time information of a UAV first-view image at a current time to obtain a target time; acquiring an original attitude angle of the UAV at the target time, combining the original attitude angle at the target time with an original attitude angle at a previous time of the target time and a filter output attitude angle to generate an adaptive filter coefficient at the target time; determining a filter output attitude angle at the target time according to the adaptive filter coefficient at the target time, and performing image stabilization control on the UAV first-view image at the current time by using the filter output attitude angle at the target time.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicles (UAVs), and more particularly to a method, apparatus, and software program for stabilizing first-view images of UAVs. Background Technology

[0002] Currently, first-person view (FPS) footage from drones is widely used in manual operation, aerial surveying, assisted sensing, virtual reality overlay displays, and subsequent intelligent analysis. To improve the control experience and visual stability, it is usually necessary to compensate for changes in the drone's attitude in real time, so that the main line of sight of the final displayed image remains as stable as possible.

[0003] In existing solutions, a common approach is to directly compensate and control the current image based on the attitude information acquired at the current moment, or to perform simple low-pass filtering on the attitude angle with fixed parameters before participating in the compensation. While these solutions can achieve a certain level of image stabilization in laboratory environments, they still have significant shortcomings in real-world UAV operational scenarios and cannot meet the actual needs of complex operating conditions.

[0004] Therefore, there is an urgent need for a drone first-person view image stabilization solution that can be combined with actual drone business scenarios. Summary of the Invention

[0005] Based on the above analysis, the embodiments of the present invention aim to solve the problem that existing first-person view stabilization solutions for drones cannot adapt to actual drone business scenarios.

[0006] In a first aspect, embodiments of the present invention provide a method for stabilizing images from a first-view perspective of a UAV, comprising the following steps: Acquire and correct the time information of the UAV's first-person view image at the current moment to obtain the target time; The original attitude angle of the UAV at the target time is obtained, and combined with the original attitude angle of the previous time and the filtered output attitude angle, the adaptive filtering coefficients at the target time are generated. Based on the adaptive filtering coefficients at the target time, the filtered output attitude angle at the target time is determined, and the filtered output attitude angle at the target time is used to perform image stabilization control on the first-view image of the UAV at the current time.

[0007] Based on a further improvement to the above method, the adaptive filtering coefficients for the target time are generated according to the original attitude angle at the target time, the original attitude angle relative to the previous time, and the filtered output attitude angle relative to the previous time. The first filter adjustment amount is determined based on the error magnitude between the original attitude angle at the target time and the filtered output attitude angle relative to the previous time. The second filter adjustment amount is determined based on the original attitude angle at the target time and the change amplitude of the original attitude angle relative to the previous time. The adaptive filter coefficients for the target time are generated based on the first filter adjustment amount and the second filter adjustment amount.

[0008] Based on a further improvement to the above method, the first filter adjustment amount is determined according to the error magnitude between the original attitude angle at the target time and the filtered output attitude angle relative to the previous time. The first filter adjustment amount C1 is obtained in the following way:

[0009] Among them, E l The first error threshold is... To find the minimum value function, E is the error magnitude between the original attitude angle at the target time and the filtered output attitude angle relative to the previous time. E is obtained in the following way:

[0010] in, The original attitude angle at the target time. The filtered output attitude angle is relative to the previous time relative to the target time.

[0011] Based on a further improvement to the above method, the first filter adjustment amount is determined according to the error magnitude between the original attitude angle at the target time and the filtered output attitude angle relative to the previous time. The first filter adjustment amount C1 is obtained in the following way:

[0012] Among them, E l E is the first error threshold. s C is the second error threshold. 1l C is the upper limit of the first filter adjustment value corresponding to the first error threshold. 1s The first filter adjustment value is the base value corresponding to the second error threshold. E is the error magnitude between the original attitude angle at the target time and the filtered output attitude angle relative to the previous time. E is obtained in the following way:

[0013] in, The original attitude angle at the target time. The filtered output attitude angle is relative to the previous time relative to the target time.

[0014] Based on a further improvement to the above method, the second filter adjustment amount is determined according to the original attitude angle at the target time and the magnitude of the change in the original attitude angle relative to the previous time. The second filter adjustment value C2 is obtained in the following way:

[0015] in, For motion enhancement threshold, To find the minimum value function, The original attitude angle at the target time and the magnitude of the change in the original attitude angle relative to the previous time are given. Obtained in the following ways:

[0016] in, The original attitude angle at the target time. The original attitude angle is relative to the previous time relative to the target time.

[0017] A further improvement to the above method, based on the first filter adjustment amount and the second filter adjustment amount, generates the adaptive filter coefficients for the target time as follows: The adaptive filtering coefficient α at the target time is obtained in the following way: or

[0018] in, and These are the lower and upper limits of the adaptive filter coefficients, respectively. These are the base values ​​for the adaptive filter coefficients. and C1 and C2 are the weighting coefficients corresponding to the first filter adjustment amount and the second filter adjustment amount, respectively. To find the minimum value function, This is the function for finding the maximum value.

[0019] Based on a further improvement to the above method, determining the filtered output attitude angle at the target time according to the adaptive filtering coefficients at the target time includes: The filtered output attitude angle at the target time Obtained in the following ways:

[0020] in, The original attitude angle at the target time. The filtered output attitude angle is relative to the previous time relative to the target time.

[0021] Based on a further improvement of the above method, the acquisition and correction of the time information of the UAV's first-view image at the current moment includes: Based on the link delay of the UAV first-view image, the time information of the UAV first-view image at the current moment is corrected to obtain the target time, including; Wherein, the target time Obtained in the following ways:

[0022] in, The time information of the first-view image of the UAV at the current moment. This is the image latency compensation amount determined based on the link latency of the UAV's first-view image.

[0023] Secondly, embodiments of the present invention provide a first-view image stabilization device for unmanned aerial vehicles, comprising: The correction module is used to acquire and correct the time information of the UAV's first-person view image at the current moment to obtain the target time; A generation module is used to acquire the original attitude angles of the UAV at the target time, and combine them with the original attitude angles of the previous time step and the filtered output attitude angles to generate adaptive filtering coefficients for the target time; and The image stabilization control module is used to determine the filtered output attitude angle at the target time based on the adaptive filtering coefficient at the target time, and to perform image stabilization control on the first-view image of the UAV at the current time using the filtered output attitude angle at the target time.

[0024] Thirdly, embodiments of the present invention provide a software program product including program code instructions, wherein when the program product is executed by a computing device, the program code instructions cause the computing device to perform the method described in any of the first aspects of the present invention.

[0025] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects: 1. In this embodiment of the invention, the "image timestamp advance + historical posture lookup" mechanism is introduced, so the posture information involved in the compensation can more accurately correspond to the actual imaging time of the image, which can significantly reduce the compensation deviation caused by the time misalignment between the image and the posture.

[0026] 2. Because the embodiments of the present invention introduce a two-factor adaptive filtering mechanism of "attitude error amplitude + attitude change amplitude", a smoother filtering strategy can be used to suppress jitter in scenarios with small attitude disturbances; and a more sensitive filtering strategy can be used to reduce compensation lag in scenarios with rapid maneuvering.

[0027] 3. Since the output compensation angle is used to update the mapping parameters in the first-view output link of the UAV, this invention does not stop at the level of attitude estimation or theoretical compensation, but can directly improve the stability, continuity and visual consistency of the final output image.

[0028] 4. Compared with schemes that rely solely on fixed compensation parameters, the embodiments of the present invention can better adapt to different flight states and different image link delay conditions, reducing trailing, overshoot, and back-and-forth oscillation.

[0029] 5. Compared to solutions that rely solely on current attitude data, this invention is more suitable for engineering implementation in real-time video links, and is particularly applicable to scenarios with high requirements for image stability, such as UAV first-person perspective, aerial survey preview, assisted perception, and virtual reality display.

[0030] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0031] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts. Figure 1 A flowchart illustrating a first-view image stabilization method for unmanned aerial vehicles (UAVs) according to an embodiment of the present invention is shown. Detailed Implementation

[0032] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0033] This invention provides a method for stabilizing first-view images from unmanned aerial vehicles (UAVs). Figure 1 A flowchart illustrating a first-view image stabilization method for unmanned aerial vehicles (UAVs) according to an embodiment of the present invention is shown. Figure 1 As shown, the first-view image stabilization method for this UAV includes: Step S100: Obtain and correct the time information of the UAV's first-person view image at the current moment to obtain the target time.

[0034] In this embodiment, the image timestamp of the drone's first-person view image at the current moment can be obtained.

[0035] In this embodiment, the timestamp of the UAV's first-view image at the current moment can be shifted forward based on the latency compensation amount of the UAV's first-view image to obtain the query timestamp of the target attitude. The query timestamp of the target attitude is the target time in this embodiment.

[0036] In some embodiments, the target time It can be obtained in the following ways:

[0037] in, This provides the time information for the first-person view image from the drone at the current moment. This refers to the image latency compensation amount determined based on the link latency of the UAV's first-person view image. Image latency compensation amount It can be a fixed value, a calibration value, an empirical value, an online estimate, or a combination thereof.

[0038] In some embodiments, image delay compensation amount The settings can be configured based on the drone model, image resolution, image processing pipeline, cache depth, and operating load. Preferably, the image latency compensation amount... The delay ranges from 20ms to 120ms. Preferably, the image delay compensation amount... The delay ranges from 40ms to 80ms. In this embodiment, the image delay compensation amount... The optimal time is approximately 60ms.

[0039] In existing solutions, image acquisition, caching, transmission, and processing all involve latency. By the time the algorithm actually obtains the image, the actual imaging time is usually earlier than the current processing time. If the current inertial navigation data or attitude data is directly used for compensation, a time misalignment occurs where "attitude data is ahead and image content is behind," leading to delayed image compensation, overshoot, or jitter. Therefore, this invention proposes an image timestamp advance mechanism. The attitude information used in the compensation more accurately corresponds to the actual image imaging time, significantly reducing the compensation deviation caused by the time misalignment between the image and attitude.

[0040] Step S200: Obtain the original attitude angle of the UAV at the target time, and combine it with the original attitude angle of the previous time and the filtered output attitude angle to generate the adaptive filtering coefficients for the target time.

[0041] In this embodiment, the original attitude data of the UAV corresponding to the target time can be queried from the UAV's attitude history cache. The original attitude data of the UAV corresponding to the target time is the target attitude. Then, the original attitude angles used for image stabilization are extracted from the target attitude. The original attitude angles here can be the pitch angle or a compensation angle derived from the pitch angle.

[0042] In this embodiment, the UAV's attitude history cache stores multiple historical attitude samples indexed by timestamps. The method for querying the original UAV attitude data corresponding to the target time in the UAV's attitude history cache can employ at least one of the following: 1. Based on the query timestamp of the target pose, select the historical pose sample with the smallest time difference.

[0043] 2. If there are two adjacent historical pose samples that surround the query timestamp of the target pose, then perform linear interpolation on these two adjacent historical pose samples to obtain the target pose.

[0044] 3. If a historical pose sample is not found, the most recent valid target pose will be used instead.

[0045] Step S400: Generate the adaptive filtering coefficients for the target time based on the original attitude angle at the target time, the original attitude angle relative to the previous time, and the filtered output attitude angle relative to the previous time.

[0046] This invention proposes a method for stabilizing first-view images of UAVs based on two-factor adaptive attitude filtering. Specifically, adaptive filtering coefficients for the target time are generated based on the attitude change between the original attitude angle at the target time and the original attitude angle relative to the previous time (first factor) and the attitude error between the original attitude angle at the target time and the filtered output attitude angle relative to the previous time (second factor).

[0047] In some embodiments, the adaptive filter coefficients at the target time can be generated according to the following steps: Step 401: Determine the first filter adjustment amount based on the error magnitude between the original attitude angle at the target time and the filtered output attitude angle relative to the previous time.

[0048] Step 402: Determine the second filter adjustment amount based on the original attitude angle at the target time and the change amplitude of the original attitude angle relative to the previous time.

[0049] Step 403: Generate the adaptive filter coefficients for the target time based on the first filter adjustment amount and the second filter adjustment amount.

[0050] Steps 401 to 403 will be described below with reference to some specific embodiments.

[0051] In some embodiments, step 401 includes: The first filter adjustment value C1 is obtained in the following way:

[0052] Among them, E l The first error threshold is... To find the minimum value function, E is the error magnitude between the original attitude angle at the target time and the filtered output attitude angle relative to the previous time. E is obtained as follows:

[0053] in, The original attitude angle at the target time. This is the filtered output attitude angle relative to the previous time step before the target time.

[0054] In some embodiments, step 401 includes: The first filter adjustment amount C1 is obtained in the following way:

[0055] Among them, E l E is the first error threshold. s C is the second error threshold. 1l C is the upper limit of the first filter adjustment value corresponding to the first error threshold. 1s Here, E is the base value of the first filter adjustment corresponding to the second error threshold, and E is the error magnitude between the original attitude angle at the target time and the filtered output attitude angle relative to the previous time. E is obtained in the following way:

[0056] in, The original attitude angle at the target time. This is the filtered output attitude angle relative to the previous time step before the target time. In this embodiment, it is preferable to satisfy... Among them, E l E s C 1l C 1s These are all empirical values ​​and can be adjusted based on actual flight performance.

[0057] In some embodiments, step 402 includes: The second filter adjustment value C2 is obtained in the following way:

[0058] in, For motion enhancement threshold, To find the minimum value function, Let be the original attitude angle at the target time and the magnitude of the change in the original attitude angle relative to the previous time. Obtained in the following ways:

[0059] in, The original attitude angle at the target time. Let be the original attitude angle relative to the previous time step before the target time. From the above equation, it can be seen that when... When it increases, Increase accordingly, when At that time, the second filter adjustment reaches the upper limit of motion enhancement. This is the upper limit for motion enhancement, which is an empirical value and can be adjusted based on actual flight performance.

[0060] In some embodiments, step 403 includes: The adaptive filter coefficient α at the target time is obtained in the following way: or

[0061] in, and These are the lower and upper limits of the adaptive filter coefficients, respectively. These are the base values ​​for the adaptive filter coefficients. and C1 and C2 are the weighting coefficients corresponding to the first and second filter adjustment values, respectively. To find the minimum value function, This is the function for finding the maximum value.

[0062] Existing solutions use fixed compensation parameters to suppress jitter. However, if jitter is suppressed solely by fixed compensation parameters, it is prone to being insufficiently smooth under small attitude disturbances, and prone to slow response under large attitude changes or rapid maneuvers, making it difficult to simultaneously achieve stability and tracking performance. In addition, under different flight states of the UAV, the amplitude and rate of change of attitude angle and image latency are not constant, and a single fixed compensation parameter is difficult to adapt to complex operating conditions.

[0063] In this embodiment, the adaptive filtering coefficient α adopts a two-factor adaptive filtering mechanism of "attitude error amplitude + attitude change amplitude". Therefore, in scenarios with small attitude disturbances, a smoother filtering strategy can be used to suppress jitter; in scenarios with rapid maneuvering, a more sensitive filtering strategy can be used to reduce compensation lag.

[0064] In some embodiments, the second error threshold E s The range is from 0.01 rad to 0.05 rad, and the first error threshold E l The range is from 0.05 rad to 0.15 rad. Motion enhancement threshold. The range is from 0.02 rad to 0.08 rad. The first error threshold E l Larger, to enhance response speed; second error threshold E s Smaller size to enhance stability.

[0065] Step S300: Determine the filtered output attitude angle at the target time based on the adaptive filtering coefficients at the target time, and use the filtered output attitude angle at the target time to perform image stabilization control on the UAV first-view image at the current time.

[0066] In this embodiment, the target attitude angle can be updated by using adaptive filtering coefficients to obtain the filtered output compensation angle.

[0067] In some embodiments, step S500 includes: Filtered output attitude angle at the target time Obtained in the following ways:

[0068] in, The original attitude angle at the target time. This is the filtered output attitude angle relative to the previous time step before the target time.

[0069] As can be seen from the above formula, the first filter adjustment C1 and the second filter adjustment C2 are not directly used as the filter output. Instead, they first jointly determine the adaptive filter coefficient α of the current frame, and then the adaptive filter coefficient controls the current target attitude angle. Upward filter output The approximation speed. Among them, the larger the adaptive filter coefficient α, the more the current output is biased towards the latest target attitude angle, and the stronger the responsiveness; the smaller the adaptive filter coefficient α, the more the current output is biased towards the historical filtering results, and the stronger the smoothness.

[0070] In this embodiment, image stabilization control includes, but is not limited to, any one or more of the following: 1. Select the target mapping table based on the filter output compensation angle.

[0071] 2. Update the image remapping parameters based on the filter output compensation angle.

[0072] 3. Drive the image distortion correction or remapping link based on the filter output compensation angle.

[0073] 4. Update the virtual line of sight direction based on the filter output compensation angle.

[0074] 5. Output the first-person perspective image after the filter output compensation angle is stabilized.

[0075] Image stabilization control based on the filter output compensation angle can be achieved using existing technologies, all of which can achieve the desired effect / purpose of the invention, and will not be elaborated here. The innovation of this invention lies in providing a more accurate method for generating the filter output compensation angle.

[0076] Secondly, embodiments of the present invention provide a first-view image stabilization device for unmanned aerial vehicles, comprising: The correction module is used to acquire and correct the time information of the UAV's first-person view image at the current moment to obtain the target time; A generation module is used to acquire the original attitude angles of the UAV at the target time, and combine them with the original attitude angles of the previous time step and the filtered output attitude angles to generate adaptive filtering coefficients for the target time; and The image stabilization control module is used to determine the filtered output attitude angle at the target time based on the adaptive filtering coefficient at the target time, and to perform image stabilization control on the first-view image of the UAV at the current time using the filtered output attitude angle at the target time.

[0077] Thirdly, embodiments of the present invention provide a software program product including program code instructions, wherein when the program product is executed by a computing device, the program code instructions cause the computing device to perform the method described in any of the first aspects of the present invention.

[0078] Compared with the prior art, the embodiments of the present invention can achieve at least one of the following beneficial effects: 1. In this embodiment of the invention, the "image timestamp advance + historical posture lookup" mechanism is introduced, so the posture information involved in the compensation can more accurately correspond to the actual imaging time of the image, which can significantly reduce the compensation deviation caused by the time misalignment between the image and the posture.

[0079] 2. Because the embodiments of the present invention introduce a two-factor adaptive filtering mechanism of "attitude error amplitude + attitude change amplitude", a smoother filtering strategy can be used to suppress jitter in scenarios with small attitude disturbances; and a more sensitive filtering strategy can be used to reduce compensation lag in scenarios with rapid maneuvering.

[0080] 3. Since the output compensation angle is used to update the mapping parameters in the first-view output link of the UAV, this invention does not stop at the level of attitude estimation or theoretical compensation, but can directly improve the stability, continuity and visual consistency of the final output image.

[0081] 4. Compared with schemes that rely solely on fixed compensation parameters, the embodiments of the present invention can better adapt to different flight states and different image link delay conditions, reducing trailing, overshoot, and back-and-forth oscillation.

[0082] 5. Compared to solutions that rely solely on current attitude data, this invention is more suitable for engineering implementation in real-time video links, and is particularly applicable to scenarios with high requirements for image stability, such as UAV first-person perspective, aerial survey preview, assisted perception, and virtual reality display.

[0083] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for stabilizing first-view images from a UAV, characterized in that, Includes the following steps: Acquire and correct the time information of the UAV's first-person view image at the current moment to obtain the target time; Obtain the original attitude angles of the UAV at the target time. The adaptive filtering coefficients for the target time are generated by combining the original attitude angle and the filtered output attitude angle of the previous time step. Based on the adaptive filtering coefficients at the target time, the filtered output attitude angle at the target time is determined, and the filtered output attitude angle at the target time is used to perform image stabilization control on the first-view image of the UAV at the current time.

2. The method according to claim 1, characterized in that, The process of obtaining the original attitude angle of the UAV at the target time, and combining it with the original attitude angle of the previous time and the filtered output attitude angle, generates the adaptive filtering coefficients for the target time, including: The first filter adjustment amount is determined based on the error magnitude between the original attitude angle at the target time and the filtered output attitude angle relative to the previous time. The second filter adjustment amount is determined based on the original attitude angle at the target time and the change amplitude of the original attitude angle relative to the previous time. The adaptive filter coefficients for the target time are generated based on the first filter adjustment amount and the second filter adjustment amount.

3. The method according to claim 2, characterized in that, Based on the error magnitude between the original attitude angle at the target time and the filtered output attitude angle relative to the previous time, the first filter adjustment amount is determined as follows: The first filter adjustment amount C1 is obtained in the following way: Among them, E l The first error threshold is... To find the minimum value function, E is the error magnitude between the original attitude angle at the target time and the filtered output attitude angle relative to the previous time. E is obtained in the following way: in, The original attitude angle at the target time. The filtered output attitude angle is relative to the previous time step before the target time step.

4. The method according to claim 2, characterized in that, Based on the error magnitude between the original attitude angle at the target time and the filtered output attitude angle relative to the previous time, the first filter adjustment amount is determined as follows: The first filter adjustment amount C1 is obtained in the following way: Among them, E l E is the first error threshold. s C is the second error threshold. 1l C is the upper limit of the first filter adjustment value corresponding to the first error threshold. 1s The first filter adjustment value is the base value corresponding to the second error threshold. E is the error magnitude between the original attitude angle at the target time and the filtered output attitude angle relative to the previous time. E is obtained in the following way: in, The original attitude angle at the target time. The filtered output attitude angle is relative to the previous time step before the target time step.

5. The method according to claim 2, characterized in that, Based on the original attitude angle at the target time and the magnitude of the change in the original attitude angle relative to the previous time, the second filter adjustment amount is determined as follows: The second filter adjustment value C2 is obtained in the following way: in, For motion enhancement threshold, To find the minimum value function, The original attitude angle at the target time and the magnitude of the change in the original attitude angle relative to the previous time are given. Obtained in the following ways: in, The original attitude angle at the target time. The original attitude angle is relative to the previous time relative to the target time.

6. The method according to claim 2, characterized in that, The adaptive filter coefficients for the target time are generated based on the first filter adjustment amount and the second filter adjustment amount, including: The adaptive filtering coefficient α at the target time is obtained in the following way: or in, and These are the lower and upper limits of the adaptive filter coefficients, respectively. These are the base values ​​for the adaptive filter coefficients. and C1 and C2 are the weighting coefficients corresponding to the first filter adjustment amount and the second filter adjustment amount, respectively. To find the minimum value function, This is the function for finding the maximum value.

7. The method according to claim 6, characterized in that, Determining the filtered output attitude angle at the target time based on the adaptive filtering coefficients at the target time includes: The filtered output attitude angle at the target time Obtained in the following ways: in, The original attitude angle at the target time. The filtered output attitude angle is relative to the previous time step before the target time step.

8. The method according to any one of claims 1-7, characterized in that, The acquisition and correction of the time information of the UAV first-view image at the current moment includes: Based on the link delay of the UAV first-view image, the time information of the UAV first-view image at the current moment is corrected to obtain the target time; Wherein, the target time Obtained in the following ways: in, The time information of the first-view image of the UAV at the current moment. This is the image latency compensation amount determined based on the link latency of the UAV's first-view image.

9. A first-view image stabilization device for unmanned aerial vehicles, characterized in that, include: The correction module is used to acquire and correct the time information of the UAV's first-person view image at the current moment to obtain the target time; The generation module is used to obtain the original attitude angle of the UAV at the target time, and combine it with the original attitude angle of the previous time and the filtered output attitude angle to generate the adaptive filtering coefficients at the target time. as well as The image stabilization control module is used to determine the filtered output attitude angle at the target time based on the adaptive filtering coefficient at the target time, and to perform image stabilization control on the first-view image of the UAV at the current time using the filtered output attitude angle at the target time.

10. A software program product comprising program code instructions that, when executed by a computing device, cause the computing device to perform the method of any one of claims 1 to 8.

Images