WO2022061934A1 - 图像处理方法、装置、系统、平台及计算机可读存储介质 - Google Patents

图像处理方法、装置、系统、平台及计算机可读存储介质 Download PDF

Info

Publication number
WO2022061934A1
WO2022061934A1 PCT/CN2020/118577 CN2020118577W WO2022061934A1 WO 2022061934 A1 WO2022061934 A1 WO 2022061934A1 CN 2020118577 W CN2020118577 W CN 2020118577W WO 2022061934 A1 WO2022061934 A1 WO 2022061934A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
light intensity
image processing
preset
brightness
Prior art date
Application number
PCT/CN2020/118577
Other languages
English (en)
French (fr)
Inventor
徐斌
李志强
李静
Original Assignee
深圳市大疆创新科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳市大疆创新科技有限公司 filed Critical 深圳市大疆创新科技有限公司
Priority to PCT/CN2020/118577 priority Critical patent/WO2022061934A1/zh
Priority to CN202080015619.3A priority patent/CN113491099A/zh
Publication of WO2022061934A1 publication Critical patent/WO2022061934A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/95—Computational photography systems, e.g. light-field imaging systems
    • H04N23/951—Computational photography systems, e.g. light-field imaging systems by using two or more images to influence resolution, frame rate or aspect ratio
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/60—Control of cameras or camera modules
    • H04N23/695—Control of camera direction for changing a field of view, e.g. pan, tilt or based on tracking of objects
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/70—Circuitry for compensating brightness variation in the scene
    • H04N23/73—Circuitry for compensating brightness variation in the scene by influencing the exposure time
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/80—Camera processing pipelines; Components thereof
    • H04N23/81—Camera processing pipelines; Components thereof for suppressing or minimising disturbance in the image signal generation

Definitions

  • the present application relates to the technical field of image processing, and in particular, to an image processing method, apparatus, system, platform, and computer-readable storage medium.
  • dark-light imaging mainly performs white balance, mosaic, noise reduction, and Gamma correction on images captured in a dark-light environment.
  • the brightness of the image captured in the light environment although the problem of dark light imaging can be solved to a certain extent through the above method, but the processed image is relatively blurred, the image quality and clarity of dark light imaging cannot be guaranteed, and the user experience is not good. .
  • the embodiments of the present application provide an image processing method, device, system, platform, and computer-readable storage medium, which aim to improve the quality and clarity of images captured in a dark light environment.
  • an embodiment of the present application provides an image processing method, including:
  • Control the photographing device to take pictures according to the short exposure time of the single frame and the target number of shots, to obtain at least one first image
  • the at least one first image is processed according to the preset dark-light imaging model and the brightness gear multiple corresponding to the illumination intensity, to obtain a second image, wherein the preset dark-light imaging model is based on the training image and
  • the labeled reference images are obtained by training the neural network model.
  • an embodiment of the present application further provides an image processing apparatus, where the image processing apparatus is used to control a photographing apparatus, and the image processing apparatus includes a memory and a processor;
  • the memory is used to store computer programs
  • the processor is configured to execute the computer program and implement the following steps when executing the computer program:
  • Control the photographing device to take pictures according to the short exposure time of the single frame and the target number of shots, to obtain at least one first image
  • the at least one first image is processed according to the preset dark-light imaging model and the brightness gear multiple corresponding to the illumination intensity, to obtain a second image, wherein the preset dark-light imaging model is based on the training image and
  • the labeled reference images are obtained by training the neural network model.
  • an embodiment of the present application further provides a photographing device, where the photographing device includes a memory and a processor;
  • the memory is used to store computer programs
  • the processor is configured to execute the computer program and implement the following steps when executing the computer program:
  • Control the photographing device to take pictures according to the short exposure time of the single frame and the target number of shots, to obtain at least one first image
  • the at least one first image is processed according to the preset dark-light imaging model and the brightness gear multiple corresponding to the illumination intensity, to obtain a second image, wherein the preset dark-light imaging model is based on the training image and
  • the labeled reference images are obtained by training the neural network model.
  • an embodiment of the present application further provides a movable platform, where the movable platform includes:
  • the power system is arranged on the platform body, and the power system is used to provide moving power for the movable platform;
  • a gimbal the gimbal is mounted on the platform body, and the gimbal is used for carrying a photographing device;
  • the image processing device is provided on the platform body, and the image processing device is further used to control the movement of the movable platform.
  • an embodiment of the present application further provides a photographing system, where the photographing system includes a gimbal, a photographing device mounted on the gimbal, and the image processing device as described above.
  • an embodiment of the present application further provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the above-mentioned The steps of an image processing method.
  • the embodiments of the present application provide an image processing method, device, system, platform, and computer-readable storage medium, by determining a single-frame short exposure of a shooting device according to the illumination intensity when the illumination intensity of the current environment is less than a preset illumination intensity time and target shooting times, and then control the shooting device to take pictures according to the short exposure time of a single frame and the target shooting times to obtain at least one first image, and finally the preset obtained by training the neural network model based on the training image and the labeled reference image
  • the dark-light imaging model and the brightness gear multiple corresponding to the light intensity process at least one first image to obtain a second image, which can improve the quality and clarity of images captured in a dark-light environment.
  • FIG. 1 is a schematic diagram of a scene for implementing the image processing method provided by the embodiment of the present application
  • FIG. 2 is a schematic diagram of another scenario for implementing the image processing method provided by the embodiment of the present application.
  • FIG. 3 is a schematic flowchart of steps of an image processing method provided by an embodiment of the present application.
  • Fig. 4 is the sub-step schematic flow chart of the image processing method in Fig. 3;
  • FIG. 5 is a schematic block diagram of the structure of an image processing apparatus provided by an embodiment of the present application.
  • FIG. 6 is a schematic block diagram of the structure of a photographing apparatus provided by an embodiment of the present application.
  • FIG. 7 is a schematic structural block diagram of a movable platform provided by an embodiment of the present application.
  • FIG. 8 is a schematic structural block diagram of a photographing system provided by an embodiment of the present application.
  • dark-light imaging mainly performs white balance, mosaic, noise reduction, and Gamma correction on the images captured in the dark environment in sequence.
  • the brightness of the image captured in the light environment although the problem of dark light imaging can be solved to a certain extent through the above method, but the processed image is relatively blurred, the image quality and clarity of dark light imaging cannot be guaranteed, and the user experience is not good. .
  • the embodiments of the present application provide an image processing method, device, system, platform, and computer-readable storage medium.
  • Single-frame short exposure time and target shooting times and then control the shooting device to take pictures according to the single-frame short exposure time and target shooting times to obtain at least one first image, and finally train the neural network model based on the training image and the labeled reference image.
  • the obtained preset dark-light imaging model and the brightness gear multiple corresponding to the light intensity process at least one first image to obtain a second image, which can improve the quality and clarity of an image captured in a dark-light environment.
  • FIG. 1 is a schematic diagram of a scene for implementing the image processing method provided by the embodiment of the present application.
  • the scene includes a handheld gimbal 100 and a photographing device 200 mounted on the handheld gimbal 100 .
  • the handheld gimbal 100 includes a handle portion 101 and a gimbal 102 disposed on the handle portion 101 .
  • the gimbal 102 uses When the photographing device 200 is mounted, the photographing device 200 may be integrated with the pan/tilt 102 , or may be externally connected to the pan/tilt 102 .
  • the photographing device 200 may be a smartphone, or a camera, such as a single-lens reflex camera, or a camera.
  • the handheld gimbal 100 can carry the photographing device 200 for fixing the photographing device 200 and changing the height, inclination and/or direction of the photographing device 200 , or for stably keeping the photographing device 200 in a certain posture and controlling the photographing device 200 to shoot.
  • the pan/tilt head 102 includes three-axis motors, and the three-axis motors are a pitch (pitch) axis motor 1021, a yaw (yaw) axis motor 1022, and a roll (roll) axis motor (not shown in FIG. 1 ), respectively.
  • the three-axis motor is used to adjust the balance posture of the photographing device 200 mounted on the gimbal 102 so as to photograph a stable and smooth picture.
  • the PTZ 102 is also provided with an inertial measurement unit (Inertial measurement unit, IMU), which can be, for example, at least one of an accelerometer or a gyroscope, which can be used to measure the attitude and acceleration of the PTZ 102, etc. Adjust the posture of the gimbal 102 .
  • the handle portion 101 is also provided with an inertial measurement unit (Inertial measurement unit, IMU), for example including at least one of an accelerometer or a gyroscope, which can be used to measure the attitude and acceleration of the handle portion 101, etc., In order to adjust the posture of the pan/tilt head 102 according to the posture of the handle part 101 and the posture of the pan/tilt head 102 .
  • the handheld gimbal 100 is communicatively connected to the photographing device 200 , and the handheld gimbal 100 may be connected to the photographing device 200 through a control line, such as a shutter cable.
  • a control line such as a shutter cable.
  • the type of the shutter release cable is not limited here, for example, the shutter release cable may be a Universal Serial Bus (Universal Serial Bus, USB).
  • the handheld gimbal 100 can also be connected to the photographing device 200 in a wireless manner. communication connection between.
  • the handheld cloud platform 100 further includes an image processing device (not shown in the figure), and the image processing device is arranged inside the handle portion 101.
  • the image processing device obtains the light intensity of the current environment. , and when the light intensity of the current environment is less than the preset light intensity, determine the single-frame short exposure time and the target shooting times of the shooting device 200 according to the light intensity; control the shooting device 200 to take pictures according to the single-frame short exposure time and the target shooting times to obtain at least one first image; at least one first image is processed by the preset dark-light imaging model obtained by training the neural network model based on the training image and the labeled reference image and the brightness gear multiple corresponding to the illumination intensity , to obtain a second image, which can improve the quality and clarity of the image captured in a dark light environment.
  • the image processing device can obtain the light intensity of the current environment by means of sensors such as cameras, and can also obtain the light intensity of the current environment by means of weather data downloaded by the server.
  • the handle portion 101 is further provided with a control key, so that the user can operate the control key to control the pan/tilt head 102 or the photographing device 200 .
  • the control key may be, for example, a key, a trigger, a knob or a joystick, etc., of course, other forms of physical keys or virtual keys are also included.
  • the virtual keys may be virtual buttons provided on the touch screen for interacting with the user.
  • the joystick can be used to control the movement of at least one rotating shaft, and then control the movement of the photographing device 200 . It will be appreciated that the joystick can also be used for other functions. It can be understood that the number of control keys may be one or more.
  • control keys When the number of control keys is one, different operation modes can be used to generate different control instructions for the control key, for example, the number of times of pressing is different; when the number of control keys is multiple, for example, the first control key, second control key, third control key, etc., different control keys are used to generate different control instructions.
  • FIG. 2 is a schematic diagram of another scenario for implementing the image processing method provided by the embodiment of the present application.
  • the scenario includes a control terminal 300 and a movable platform 400 , and the control terminal 300 and the movable platform 400
  • the control terminal 300 includes a display device 310
  • the display device 310 is used for displaying the image sent by the movable platform 400 .
  • the display device 310 includes a display screen disposed on the control terminal 300 or a display independent of the control terminal 300, and the display independent of the control terminal 300 may include a mobile phone, a tablet computer, a personal computer, etc., or may also be a Other electronic equipment with a display screen.
  • the display screen includes an LED display screen, an OLED display screen, an LCD display screen, and the like.
  • the movable platform 400 includes a platform body 410, a gimbal 420 mounted on the platform body, and a power system 430.
  • the gimbal 420 is used to carry the photographing device 500
  • the power system 430 includes a motor 431 and a propeller 432. 431 is used to drive the propeller 432 to rotate, so as to provide moving power for the movable platform.
  • the pan/tilt 420 includes three-axis motors, which are a translation axis motor 421, a pitch axis motor 422, and a roll axis motor 423, which are used to adjust the balance posture of the photographing device 500 mounted on the pan/tilt 420, so as to capture images anytime, anywhere. High-precision stable picture.
  • the movable platform 400 further includes an image processing device (not shown in the figure), and the image processing device is arranged inside the platform body 410.
  • the image processing device obtains the light intensity of the current environment, And when the light intensity of the current environment is less than the preset light intensity, determine the single-frame short exposure time and the target shooting times of the shooting device 500 according to the light intensity; control the shooting device 500 to take pictures according to the single-frame short exposure time and the target shooting times, At least one first image is obtained; at least one first image is processed by a preset dark-light imaging model obtained by training the neural network model based on the training image and the labeled reference image and the brightness gear multiple corresponding to the illumination intensity, Obtaining the second image can improve the quality and definition of the image captured in a dark light environment.
  • the image processing device can obtain the light intensity of the current environment by means of sensors such as cameras, and can also obtain the light intensity of the current environment by means of weather data downloaded by the server.
  • the movable platform includes movable robots, unmanned aerial vehicles and unmanned vehicles, etc.
  • the movable platform 400 is an unmanned aerial vehicle, and the power system 430 can make the unmanned aerial vehicle take off vertically from the ground, or land vertically on the ground, Without the need for any horizontal movement of the drone (eg no taxiing on the runway).
  • power system 430 may allow the drone to pre-set positions and/or steering in the air.
  • One or more of the power systems 430 may be controlled independently of the other power systems 430 .
  • one or more power systems 430 may be controlled simultaneously.
  • the drone may have multiple horizontally oriented power systems 430 to track the lift and/or push of the target.
  • the horizontally oriented power system 430 can be actuated to provide the drone with the ability to take off vertically, land vertically, and hover.
  • one or more of the horizontally oriented power systems 430 may rotate in a clockwise direction, while one or more of the other horizontally oriented power systems may rotate in a counter-clockwise direction.
  • the rotational rate of each power system 430 in the horizontal direction can be varied independently to achieve lift and/or push operations caused by each power system 430 to adjust the spatial orientation, speed and/or acceleration of the drone (eg, relative to multiple rotation and translation up to three degrees of freedom).
  • the drone may also include a sensing system, which may include one or more sensors to sense the spatial orientation, velocity, and/or acceleration of the drone (eg, relative to up to three Degree of freedom rotation and translation), angular acceleration, attitude, position (absolute position or relative position), etc.
  • the one or more sensors include GPS sensors, motion sensors, inertial sensors, proximity sensors, or image sensors.
  • the sensing system can also be used to collect data on the environment in which the UAV is located, such as climatic conditions, potential obstacles to be approached, locations of geographic features, locations of man-made structures, and the like.
  • the drone may include a tripod
  • the tripod is a contact piece between the drone and the ground when the drone lands
  • the tripod can be received by the unmanned aerial vehicle in a flying state (for example, when the unmanned aerial vehicle is cruising). It can only be put down when landing; it can also be fixedly installed on the drone and kept in the state of being put down all the time.
  • the movable platform 400 can communicate with the control terminal 300, and can realize data interaction between the control terminal 300 and the movable platform 400, such as the movement control of the movable platform 400, the control of the load (when When the payload is the photographing device 500, the control terminal 300 can control the photographing device 500), wherein the control terminal 300 can communicate with the movable platform 400 and/or the payload, and the communication between the movable platform 400 and the control terminal 300 can be Wireless communication can provide direct communication between the movable platform 400 and the control terminal 300 . This direct communication can occur without any intermediary devices or networks.
  • indirect communication may be provided between the movable platform 400 and the control terminal 300 .
  • Such indirect communication may take place by means of one or more intermediaries or networks.
  • indirect communication may utilize a telecommunications network.
  • Indirect communication may take place by means of one or more routers, communication towers, satellites, or any other intermediary device or network.
  • Examples of types of communication may include, but are not limited to, communication via the Internet, Local Area Network (LAN), Wide Area Network (WAN), Bluetooth, Near Field Communication (NFC) technology, based on technologies such as General Packet Radio Service (GPRS), GSM Enhanced Data GSM Environment (EDGE), 3G, 4G, or Long Term Evolution (LTE) protocols for mobile data protocols, infrared (IR) communication technology, and/or Wi-Fi, and may be wireless, wired, or its combination.
  • GPRS General Packet Radio Service
  • EDGE GSM Enhanced Data GSM Environment
  • 3G Third Generation
  • 4G Long Term Evolution
  • LTE Long Term Evolution
  • control terminal 300 may include but not limited to: smart phone/mobile phone, tablet computer, personal digital assistant (PDA), desktop computer, media content player, video game station/system, virtual reality system, augmented reality system, wearable devices (eg, watches, glasses, gloves, headwear (eg, hats, helmets, virtual reality headsets, augmented reality headsets, head mounted devices (HMDs), headbands), pendants, armbands, leg loops, shoes, vest), gesture recognition device, microphone, any electronic device capable of providing or rendering image data, or any other type of device.
  • the control terminal 300 may be a handheld terminal, and the control terminal 300 may be portable.
  • the control terminal 300 may be carried by a human user. In some cases, the control terminal 300 may be remote from the human user, and the user may control the control terminal 300 using wireless and/or wired communication.
  • FIG. 1 or FIG. 2 the scene in FIG. 1 or FIG. 2 is only used to explain the image processing method provided by the embodiment of the present application, but does not constitute a limitation on the application scene of the image processing method provided by the embodiment of the present application.
  • FIG. 3 is a schematic flowchart of steps of an image processing method provided by an embodiment of the present application.
  • the image processing method may include steps S101 to S103.
  • Step S101 when the light intensity of the current environment is less than the preset light intensity, determine the short exposure time of a single frame and the target shooting times of the shooting device according to the light intensity.
  • the illumination intensity of the current environment where the photographing device is located is obtained, and it is determined whether the illumination intensity of the current environment is less than the preset illumination intensity, and when the illumination intensity of the current environment is smaller than the preset illumination intensity, according to The light intensity determines the single-frame short exposure time of the photographing device and the number of target shots.
  • the preset light sensor can be set in the shooting device, and the preset light sensor can also be set outside the shooting device, which is not specifically limited in this embodiment, and the short exposure time of a single frame is used to indicate the shooting device
  • the exposure time for shooting one frame of image and the preset illumination intensity can also be set based on the actual situation, which is not specifically limited in this embodiment, for example, the preset illumination intensity is 1 lux.
  • the method of determining the short exposure time of a single frame and the number of target shooting times of the photographing device according to the light intensity may be: obtaining the long exposure time and the brightness level multiple corresponding to the light intensity; The bit multiple is used to determine the single-frame short exposure time of the shooting device; the target shooting times of the shooting device are determined according to the mapping relationship between the preset light intensity and the target shooting times and the light intensity.
  • the single-frame short exposure time and target shooting times of the photographing device may be determined according to the pre-stored light intensity, the single-frame short exposure time, the mapping relationship between the brightness gear multiples, and the light intensity of the current environment where the photographing device is located.
  • the mapping relationship between the light intensity and the target shooting times, and the mapping relationship between the pre-stored light intensity, short exposure time of a single frame, and brightness gear multiples can be set based on the actual situation, which is not specifically limited in this embodiment.
  • the method of determining the short exposure time of a single frame of the photographing device may be: when the ISO (sensitivity) of the photographing device is fixed, determine the long exposure time and the brightness.
  • the ratio of the gear multiples is determined, and the ratio of the long exposure time to the brightness gear multiple is determined as the single-frame short exposure time of the photographing device. For example, if the long exposure time is 4 seconds and the brightness level multiple is 32, the ratio of the long exposure time of 4 seconds to the brightness level multiple of 32 is 0.125, so the single-frame short exposure time of the photographing device is 0.125 seconds.
  • Step S102 controlling the photographing device to take pictures according to the short exposure time of the single frame and the target number of shots to obtain at least one first image.
  • the exposure time of the shooting device is controlled so that the exposure time of the first image obtained by shooting is the short exposure time of the single frame, and when the actual shooting times reach the set target shooting times. , and control the photographing device to stop photographing to obtain at least one first image. For example, if the short exposure time of a single frame is 0.25 seconds, and the number of shots of the target is 8, then when capturing an image, the exposure time of each control of the shooting device is 0.25 seconds, and 8 shots are required, so that the short exposure time of a single frame is 0.25 8 first images in seconds.
  • Step S103 process the at least one first image according to the preset dark light imaging model and the luminance gear multiple corresponding to the illumination intensity to obtain a second image.
  • the preset dark-light imaging model is obtained by training a neural network model according to a training image and a labeled reference image, and the training image includes a single-frame short exposure time corresponding to different light intensities less than the preset light intensity and the images captured by the number of shots, the labeled reference images include images captured according to long exposure times corresponding to different light intensities less than the preset light intensity, and the neural network model is trained through a large number of training images and corresponding labeled reference images.
  • a preset dark light imaging model can be obtained, and the specific hierarchical structure of the neural network model can be set according to the actual situation, which is not specifically limited in this implementation.
  • the process of establishing the dark light imaging model may be as follows: obtaining a training image and a reference image corresponding to the training image, and adjusting the brightness of the training image according to the brightness level multiple corresponding to the illumination intensity; The images are fused to obtain a fused training image, and the neural network model is iteratively trained according to the fused training image and the corresponding reference image until the trained neural network model converges, thereby obtaining a dark-light imaging model.
  • the process of establishing the dark-light imaging model may also be: acquiring a training image and a reference image corresponding to the training image, and adjusting the brightness of the training image according to the brightness gear multiple corresponding to the illumination intensity;
  • the training image and the reference image corresponding to the training image are used to iteratively train the neural network model until the trained neural network model converges, thereby obtaining a dark-light imaging model.
  • step S103 may include sub-steps S1031 to S1033.
  • Sub-step S1031 adjusting the brightness of the at least one first image according to the brightness gear multiple corresponding to the light intensity
  • Sub-step S1032 fuse the at least one first image after adjusting the brightness to obtain a third image
  • Sub-step S1033 Process the third image according to the preset dark-light imaging model to obtain a second image.
  • the preset dark light imaging model can split the third image into images of different frequency bands, and the second image is synthesized according to the images of different frequency bands.
  • the images of different frequency bands include at least one image of the first frequency band and at least one image of the second frequency band, the frequency band difference between the first frequency band and the second frequency band is greater than the preset frequency band difference, and the image of the first frequency band includes the first frequency band difference.
  • the low-frequency components in the three images, and the images in the second frequency band include the high-frequency components in the third image.
  • the format of the first image is RAW format
  • the format of images in different frequency bands is RGB format or YUV format
  • the format of the second image is RGB format or YUV format. It can be understood that, the preset frequency band difference may be set based on the actual situation, which is not specifically limited in this embodiment.
  • the preset dark-light imaging model includes an image splitting layer and an image combining layer, the image splitting layer is used to split the third image into images of different frequency bands, and the image combining layer is For synthesizing images of different frequency bands to output a second image.
  • the image splitting layer includes at least one low-frequency component splitting layer and at least one high-frequency component splitting layer, the low-frequency component splitting layer is used for splitting low-frequency components from the third image, and the high-frequency component splitting layer. Layering is used to separate out high frequency components from the third image.
  • the loss functions of the low-frequency component splitting layer and the high-frequency component splitting layer are different.
  • the image processing effect of the model can be improved, and the quality and clarity of images captured in a dark environment can be improved.
  • the process of establishing a preset dark-light imaging model that only includes an image splitting layer and an image synthesizing layer may be: acquiring a training image and a reference image corresponding to the training image, and adjusting the training image according to the brightness gear multiple corresponding to the light intensity The brightness of the image; fuse the training image after adjusting the brightness to obtain the fused training image, and iteratively train the neural network model according to the fused training image and the corresponding reference image, until the trained neural network model converges, Thus, the dark-light imaging model is obtained.
  • the loss function of the low-frequency component splitting layer is determined according to the loss of the image block to which each pixel in the low-frequency component belongs. Or the loss function of the low-frequency component splitting layer is determined according to the loss of the image block to which each pixel in the low-frequency component belongs and the weight coefficient of the loss of the image block, and the weight coefficient of the loss of the image block is determined according to the labeling.
  • the low-frequency components in the reference image are determined.
  • the loss function of the high-frequency component splitting layer is based on the single-point loss and pixel loss of each pixel in the high-frequency component splitting.
  • the weight coefficient of the single point loss of the point is determined, and the weight coefficient of the single point loss of the pixel point is determined according to the high-frequency components in the marked reference image.
  • the training of the high-frequency component split layer can be divided into two stages.
  • the first stage is to use the single-point loss of the pixels in the high-frequency component to train the high-frequency component split layer, so that the training
  • the high-frequency components output by the subsequent high-frequency component splitting layer have the same shape as the high-frequency components in the annotated reference image
  • the second stage is to generate a weight coefficient table according to the high-frequency components in the annotated reference image
  • the high-frequency component splitting layer is retrained according to the generated weight coefficient table, so that the high-frequency component splitting layer can better restore the edge and texture information of the image.
  • the weight coefficient table the weight coefficients corresponding to the pixels in the flat area in the high-frequency components are relatively small, and the weight coefficients corresponding to the pixels of the texture and the edge in the high-frequency components are relatively large.
  • the number of low-frequency component splitting layers and high-frequency component splitting layers in the preset dark-light imaging model can be set based on actual conditions.
  • the preset dark-light imaging model includes one low-frequency component splitting layer and three high-frequency component splitting layers.
  • frequency component splitting layer, and the three high-frequency component splitting layers include the first high-frequency component splitting layer, the second high-frequency component splitting layer, and the third high-frequency component splitting layer
  • the dark-light imaging model is preset
  • the processing process for the third image may be: splitting the low-frequency components from the third image through the low-frequency component splitting layer, splitting the first high-frequency components from the third image through the first high-frequency component splitting layer, The second high-frequency component is split from the third image by the second high-frequency component splitting layer, the third high-frequency component is split from the third image by the third high-frequency component splitting layer, and finally the image is synthesized
  • the layer synthesizes the low-frequency component, the first high-frequency component, the second high-frequency component, and the third high-frequency
  • the at least one first image is processed according to the preset dark light imaging model and the brightness gear multiple corresponding to the light intensity, and the second image is obtained in the following manner:
  • the brightness gear multiple adjusts the brightness of the at least one first image;
  • the at least one first image after the brightness is adjusted is processed according to a preset dark-light imaging model to obtain a second image.
  • the preset dark light imaging model includes an image fusion layer, an image splitting layer and an image synthesis layer, and the image fusion layer is used to fuse at least one first image after the brightness is adjusted.
  • the process of establishing the preset dark-light imaging model including the image fusion layer, the image splitting layer and the image synthesis layer may be as follows: obtaining a training image and a reference image corresponding to the training image, and determining the brightness level corresponding to the light intensity according to the The brightness of the training image is adjusted by multiples; the neural network model is iteratively trained according to the brightness-adjusted training image and the reference image corresponding to the training image, until the trained neural network model converges, thereby obtaining a dark-light imaging model.
  • the at least one first image after the brightness adjustment is processed according to the preset dark light imaging model, and the method for obtaining the second image may be as follows: using the image fusion layer to process the at least one first image after adjusting the brightness of the at least one first image.
  • the images are fused to obtain a third image; the third image is split into images of different frequency bands through an image splitting layer, and the images of different frequency bands are synthesized through an image synthesis layer to obtain a second image.
  • the preset dark-light imaging model including image fusion layer, image split layer and image synthesis layer can quickly output images with better image quality and definition.
  • the photographing device when the illumination intensity of the current environment is less than the preset illumination intensity, it is determined whether the photographing device is in a motion state; if the photographing device is in a motion state, the single-frame short exposure time and target shooting of the photographing device are determined according to the illumination intensity. controlling the shooting device to take pictures according to the short exposure time of a single frame and the target shooting times, to obtain at least one first image; according to the preset dark light imaging model and the brightness gear multiple corresponding to the light intensity, the at least one first image is Processing is performed to obtain a second image.
  • an inertial measurement unit (Inertial Measurement Unit, IMU) is set in the photographing device, and whether the photographing device is in a motion state can be determined by the inertial measurement unit. It is also possible to determine whether the photographing device is in a moving state by using the inertial measurement unit in the handheld gimbal or the inertial measurement unit in the movable platform.
  • IMU Inertial Measurement Unit
  • the shooting device is controlled to adjust the position of the infrared filter so that the shooting device Infrared light in the environment can enter the shooting device; control the shooting device to take pictures according to the short exposure time of a single frame and the number of target shots, and obtain at least one first image; according to the preset dark light imaging model and the brightness level corresponding to the light intensity
  • the multiple processes at least one first image to obtain a second image.
  • the photographing device includes an infrared filter, and the infrared filter is used for filtering infrared light.
  • the infrared light can enter the shooting device when the shooting device takes pictures based on the short exposure time and shooting times of a single frame, so as to improve the brightness of the image obtained by taking pictures, and then pass the dark light imaging model and light intensity.
  • the corresponding brightness gear multiples are used to process the image, which can further improve the quality and clarity of the image captured in the dark light environment.
  • the single-frame short exposure time and the target shooting times of the photographing device are determined according to the light intensity, and then the single-frame short exposure time and the target shooting times are determined according to the light intensity.
  • the number of shots is controlled by the shooting device to take pictures, and at least one first image is obtained.
  • the preset dark-light imaging model obtained by training the neural network model based on the training image and the labeled reference image is paired with the brightness gear multiple corresponding to the light intensity.
  • At least one first image is processed to obtain a second image, which can improve the quality and definition of the image captured in a dark light environment.
  • FIG. 5 is a schematic structural block diagram of an image processing apparatus provided by an embodiment of the present application.
  • the image processing apparatus 600 includes a processor 601 and a memory 602, and the processor 601 and the memory 602 are connected through a bus 603, such as an I2C (Inter-integrated Circuit) bus.
  • the image processing device 600 is used to control the photographing device.
  • the image processing device may be, for example, a chip or a processing device or the like.
  • the processor 601 may be a micro-controller unit (Micro-controller Unit, MCU), a central processing unit (Central Processing Unit, CPU), or a digital signal processor (Digital Signal Processor, DSP) or the like.
  • MCU Micro-controller Unit
  • CPU Central Processing Unit
  • DSP Digital Signal Processor
  • the memory 602 may be a Flash chip, a read-only memory (ROM, Read-Only Memory) magnetic disk, an optical disk, a U disk, a mobile hard disk, and the like.
  • ROM Read-Only Memory
  • the memory 602 may be a Flash chip, a read-only memory (ROM, Read-Only Memory) magnetic disk, an optical disk, a U disk, a mobile hard disk, and the like.
  • the processor 601 is used for running the computer program stored in the memory 602, and implements the following steps when executing the computer program:
  • Control the photographing device to take pictures according to the short exposure time of the single frame and the target number of shots, to obtain at least one first image
  • the at least one first image is processed according to the preset dark-light imaging model and the brightness gear multiple corresponding to the illumination intensity, to obtain a second image, wherein the preset dark-light imaging model is based on the training image and
  • the labeled reference images are obtained by training the neural network model.
  • the processor when the processor implements processing the at least one first image according to a preset dark-light imaging model and a brightness gear multiple corresponding to the illumination intensity, the processor is configured to implement:
  • the third image is processed according to the preset dark light imaging model to obtain a second image.
  • the preset dark-light imaging model can split the third image into images of different frequency bands, and the second image is synthesized according to the images of different frequency bands.
  • the preset dark light imaging model includes an image splitting layer and an image combining layer
  • the image splitting layer is used to split the third image into images of different frequency bands
  • the image combining The layer is used to synthesize the images of the different frequency bands to output the second image.
  • the image splitting layer includes at least one low-frequency component splitting layer and at least one high-frequency component splitting layer, and the low-frequency component splitting layer is used for splitting low-frequency components from the third image. components, and the high-frequency component splitting layer is used for splitting high-frequency components from the third image.
  • the loss function of the low-frequency component splitting layer is determined according to the loss of the image block to which each pixel in the low-frequency component belongs.
  • the loss function of the low-frequency component splitting layer is determined according to the loss of the image block to which each pixel point in the low-frequency component belongs and the weight coefficient of the loss of the image block.
  • the loss function of the high-frequency component splitting layer is determined according to the single-point loss of each pixel point in the high-frequency component splitting and the weight coefficient of the single-point loss of the pixel point.
  • the weight coefficient of the single-point loss of the pixel point is determined according to the high-frequency components in the annotated reference image.
  • the training image includes images captured according to a single frame of short exposure time and the number of shots corresponding to different light intensities less than a preset light intensity
  • the reference image includes images obtained according to different light levels less than a preset light intensity. The intensity corresponding to the long exposure time of the obtained image.
  • the processing of the at least one first image according to the preset dark light imaging model and the brightness gear multiple corresponding to the illumination intensity to obtain the second image includes:
  • the at least one first image whose brightness has been adjusted is processed according to the preset dark-light imaging model to obtain a second image.
  • the preset dark-light imaging model includes an image fusion layer, an image splitting layer, and an image synthesis layer, and the image fusion layer is used to fuse the at least one first image after brightness adjustment.
  • the at least one first image after the brightness adjustment is processed according to the preset dark-light imaging model to obtain a second image, including:
  • the at least one first image after the brightness adjustment is fused by the image fusion layer to obtain a third image
  • the third image is split into images of different frequency bands through the image splitting layer, and the images of different frequency bands are synthesized through the image synthesis layer to obtain a second image.
  • the photographing device includes an infrared filter
  • the processor controls the photographing device to take pictures according to the short exposure time of a single frame and the target number of shots, and before obtaining at least one first image, further Used to implement:
  • the photographing device is controlled to adjust the position of the infrared filter, so that the infrared light in the environment where the photographing device is located can enter the photographing device.
  • the processor determines the short exposure time of a single frame and the number of target shooting times of the shooting device according to the light intensity, the processor is configured to:
  • the target shooting times of the shooting device is determined according to the preset mapping relationship between the light intensity and the target shooting times and the light intensity.
  • the processor is further configured to:
  • the short exposure time of a single frame and the target shooting times of the photographing device are determined according to the light intensity.
  • FIG. 6 is a schematic structural block diagram of a photographing apparatus provided by an embodiment of the present application.
  • the photographing apparatus 700 includes a processor 701 and a memory 702, and the processor 701 and the memory 702 are connected through a bus 703, such as an I2C (Inter-integrated Circuit) bus.
  • a bus 703 such as an I2C (Inter-integrated Circuit) bus.
  • the processor 701 may be a micro-controller unit (Micro-controller Unit, MCU), a central processing unit (Central Processing Unit, CPU) or a digital signal processor (Digital Signal Processor, DSP) or the like.
  • MCU Micro-controller Unit
  • CPU Central Processing Unit
  • DSP Digital Signal Processor
  • the memory 702 may be a Flash chip, a read-only memory (ROM, Read-Only Memory) magnetic disk, an optical disk, a U disk, a mobile hard disk, and the like.
  • ROM Read-Only Memory
  • the memory 702 may be a Flash chip, a read-only memory (ROM, Read-Only Memory) magnetic disk, an optical disk, a U disk, a mobile hard disk, and the like.
  • the processor 701 is used for running the computer program stored in the memory 702, and implements the following steps when executing the computer program:
  • Control the photographing device to take pictures according to the short exposure time of the single frame and the target number of shots, to obtain at least one first image
  • the at least one first image is processed according to the preset dark-light imaging model and the brightness gear multiple corresponding to the illumination intensity, to obtain a second image, wherein the preset dark-light imaging model is based on the training image and
  • the labeled reference images are obtained by training the neural network model.
  • the processor when the processor implements processing the at least one first image according to a preset dark-light imaging model and a brightness gear multiple corresponding to the illumination intensity, the processor is configured to implement:
  • the third image is processed according to the preset dark light imaging model to obtain a second image.
  • the preset dark-light imaging model can split the third image into images of different frequency bands, and the second image is synthesized according to the images of different frequency bands.
  • the preset dark light imaging model includes an image splitting layer and an image combining layer
  • the image splitting layer is used to split the third image into images of different frequency bands
  • the image combining The layer is used to synthesize the images of the different frequency bands to output the second image.
  • the image splitting layer includes at least one low-frequency component splitting layer and at least one high-frequency component splitting layer, and the low-frequency component splitting layer is used for splitting low-frequency components from the third image. components, and the high-frequency component splitting layer is used for splitting high-frequency components from the third image.
  • the loss function of the low-frequency component splitting layer is determined according to the loss of the image block to which each pixel in the low-frequency component belongs.
  • the loss function of the low-frequency component splitting layer is determined according to the loss of the image block to which each pixel point in the low-frequency component belongs and the weight coefficient of the loss of the image block.
  • the loss function of the high-frequency component splitting layer is determined according to the single-point loss of each pixel point in the high-frequency component splitting and the weight coefficient of the single-point loss of the pixel point.
  • the weight coefficient of the single-point loss of the pixel point is determined according to the high-frequency components in the annotated reference image.
  • the training image includes images captured according to a single frame of short exposure time and the number of shots corresponding to different light intensities less than a preset light intensity
  • the reference image includes images obtained according to different light levels less than a preset light intensity. The intensity corresponds to the image taken with the long exposure time.
  • the processing of the at least one first image according to the preset dark light imaging model and the brightness gear multiple corresponding to the illumination intensity to obtain the second image includes:
  • the at least one first image whose brightness has been adjusted is processed according to the preset dark-light imaging model to obtain a second image.
  • the preset dark-light imaging model includes an image fusion layer, an image splitting layer, and an image synthesis layer, and the image fusion layer is used to fuse the at least one first image after brightness adjustment.
  • the at least one first image after the brightness adjustment is processed according to the preset dark-light imaging model to obtain a second image, including:
  • the at least one first image after the brightness adjustment is fused by the image fusion layer to obtain a third image
  • the third image is split into images of different frequency bands through the image splitting layer, and the images of different frequency bands are synthesized through the image synthesis layer to obtain a second image.
  • the photographing device includes an infrared filter
  • the processor controls the photographing device to take pictures according to the short exposure time of a single frame and the target number of shots, and before obtaining at least one first image, further Used to implement:
  • the photographing device is controlled to adjust the position of the infrared filter, so that the infrared light in the environment where the photographing device is located can enter the photographing device.
  • the processor determines the short exposure time of a single frame and the number of target shooting times of the shooting device according to the light intensity, the processor is configured to:
  • the target shooting times of the shooting device is determined according to the preset mapping relationship between the light intensity and the target shooting times and the light intensity.
  • the processor is further configured to:
  • the short exposure time of a single frame and the target shooting times of the photographing device are determined according to the light intensity.
  • FIG. 7 is a schematic structural block diagram of a movable platform provided by an embodiment of the present application.
  • the movable platform 800 includes:
  • the power system 810 is arranged on the platform body, and the power system 810 is used to provide moving power for the movable platform;
  • a pan-tilt 820, the pan-tilt 820 is mounted on the platform body, and the pan-tilt 820 is used for mounting a photographing device;
  • An image processing device 830, the image processing device 830 is provided on the platform body, and the image processing device 830 is also used for controlling the movable platform 800 to move.
  • FIG. 8 is a schematic structural block diagram of a photographing system provided by an embodiment of the present application.
  • the imaging system 900 includes a pan/tilt 910 , an imaging device 920 mounted on the pan/tilt 910 , and an image processing device 930 .
  • the pan-tilt 910 is connected to the handle part, and the image processing device 930 is provided on the handle part, or the pan-tilt 910 and the image processing device 930 are provided on the movable platform, and the image processing device 930 is also used to control the movement of the movable platform.
  • Embodiments of the present application further provide a computer-readable storage medium, where a computer program is stored in the computer-readable storage medium, and the computer program includes program instructions, and the processor executes the program instructions to realize the provision of the above embodiments.
  • the steps of the image processing method are described in detail below.
  • the computer-readable storage medium may be the internal storage unit of the movable platform, the handheld pan/tilt, or the photographing device described in any of the foregoing embodiments, such as a hard disk of the movable platform, the handheld pan/tilt, or the photographing device or RAM.
  • the computer-readable storage medium may also be an external storage device of the movable platform, the handheld pan/tilt, or the photographing device, such as a plug-in hard disk equipped on the movable platform, the handheld pan/tilt, or the photographing device, and intelligent storage.
  • Card Smart Media Card, SMC
  • Secure Digital Secure Digital
  • SD flash memory card
  • Flash Card flash memory card

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Theoretical Computer Science (AREA)
  • Computing Systems (AREA)
  • Physics & Mathematics (AREA)
  • Biophysics (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Studio Devices (AREA)

Abstract

一种图像处理方法、装置、系统、平台及计算机可读存储介质，该方法包括：在当前环境的光照强度小于预设光照强度时，根据所述光照强度确定拍摄装置的单帧短曝光时间和目标拍摄次数(S101)；根据所述单帧短曝光时间和目标拍摄次数控制所述拍摄装置拍照，得到至少一张第一图像(S102)；根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理，得到第二图像 (S103)。通过上述方法提高了图像的清晰度。

Description

图像处理方法、装置、系统、平台及计算机可读存储介质 技术领域
本申请涉及图像处理技术领域，尤其涉及一种图像处理方法、装置、系统、平台及计算机可读存储介质。
背景技术
目前，暗光成像主要是对在暗光环境下拍摄到的图像依次进行解白平衡、马赛克、降噪和Gamma校正等处理，也可以通过提高感光度和曝光时间等曝光参数的方式来增加暗光环境下拍摄到的图像的亮度，通过上述方式虽然可以在一定程度上解决暗光成像的问题，但处理得到的图像较为模糊，无法保证暗光成像的图像质量和清晰度，用户体验不好。
发明内容
基于此，本申请实施例提供了一种图像处理方法、装置、系统、平台及计算机可读存储介质，旨在提高暗光环境下拍摄到的图像的质量和清晰度。
第一方面，本申请实施例提供了一种图像处理方法，包括：
在当前环境的光照强度小于预设光照强度时，根据所述光照强度确定拍摄装置的单帧短曝光时间和目标拍摄次数；
根据所述单帧短曝光时间和目标拍摄次数控制所述拍摄装置拍照，得到至少一张第一图像；
根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理，得到第二图像，其中，所述预设暗光成像模型是根据训练图像和标注的参考图像对神经网络模型进行训练得到的。
第二方面，本申请实施例还提供了一种图像处理装置，所述图像处理装置用于控制拍摄装置，所述图像处理装置包括存储器和处理器；
所述存储器用于存储计算机程序；
所述处理器，用于执行所述计算机程序并在执行所述计算机程序时，实现如下步骤：
在当前环境的光照强度小于预设光照强度时，根据所述光照强度确定拍摄装置的单帧短曝光时间和目标拍摄次数；
根据所述单帧短曝光时间和目标拍摄次数控制所述拍摄装置拍照，得到至少一张第一图像；
根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理，得到第二图像，其中，所述预设暗光成像模型是根据训练图像和标注的参考图像对神经网络模型进行训练得到的。
第三方面，本申请实施例还提供了一种拍摄装置，所述拍摄装置包括存储器和处理器；
所述存储器用于存储计算机程序；
所述处理器，用于执行所述计算机程序并在执行所述计算机程序时，实现如下步骤：
在当前环境的光照强度小于预设光照强度时，根据所述光照强度确定拍摄装置的单帧短曝光时间和目标拍摄次数；
根据所述单帧短曝光时间和目标拍摄次数控制所述拍摄装置拍照，得到至少一张第一图像；
根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理，得到第二图像，其中，所述预设暗光成像模型是根据训练图像和标注的参考图像对神经网络模型进行训练得到的。
第四方面，本申请实施例还提供了一种可移动平台，所述可移动平台包括：
平台本体；
动力系统，所述动力系统设于所述平台本体上，所述动力系统用于为所述可移动平台提供移动动力；
云台，所述云台搭载于所述平台本体，所述云台用于搭载拍摄装置；
如上所述的图像处理装置，所述图像处理装置设于所述平台本体上，所述图像处理装置还用于控制所述可移动平台移动。
第五方面，本申请实施例还提供了一种拍摄系统，所述拍摄系统包括云台、搭载于所述云台的拍摄装置和如上所述的图像处理装置。
第六方面，本申请实施例还提供了一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，所述计算机程序被处理器执行时使所述处理器实现如上所述的图像处理方法的步骤。
本申请实施例提供了一种图像处理方法、装置、系统、平台及计算机可读存储介质，通过在当前环境的光照强度小于预设光照强度时，根据该光照强度确定拍摄装置的单帧短曝光时间和目标拍摄次数，然后根据单帧短曝光时间和 目标拍摄次数控制拍摄装置拍照，得到至少一张第一图像，最后通过基于训练图像和标注的参考图像对神经网络模型进行训练得到的预设暗光成像模型和该光照强度对应的亮度档位倍数对至少一张第一图像进行处理，得到第二图像，能够提高暗光环境下拍摄到的图像的质量和清晰度。
应当理解的是，以上的一般描述和后文的细节描述仅是示例性和解释性的，并不能限制本申请。
附图说明
为了更清楚地说明本申请实施例技术方案，下面将对实施例描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1是实施本申请实施例提供的图像处理方法的一场景示意图；
图2是实施本申请实施例提供的图像处理方法的另一场景示意图；
图3是本申请实施例提供的一种图像处理方法的步骤示意流程图；
图4是图3中的图像处理方法的子步骤示意流程图；
图5是本申请实施例提供的一种图像处理装置的结构示意性框图；
图6是本申请实施例提供的一种拍摄装置的结构示意性框图；
图7是本申请实施例提供的一种可移动平台的结构示意性框图；
图8是本申请实施例提供的一种拍摄系统的结构示意性框图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
附图中所示的流程图仅是示例说明，不是必须包括所有的内容和操作/步骤，也不是必须按所描述的顺序执行。例如，有的操作/步骤还可以分解、组合或部分合并，因此实际执行的顺序有可能根据实际情况改变。
下面结合附图，对本申请的一些实施方式作详细说明。在不冲突的情况下，下述的实施例及实施例中的特征可以相互组合。
目前，暗光成像主要是对在暗光环境下拍摄到的图像依次进行解白平衡、 马赛克、降噪和Gamma校正等处理，也可以通过提高感光度和曝光时间等曝光参数的方式来增加暗光环境下拍摄到的图像的亮度，通过上述方式虽然可以在一定程度上解决暗光成像的问题，但处理得到的图像较为模糊，无法保证暗光成像的图像质量和清晰度，用户体验不好。
为解决上述问题，本申请实施例提供一种图像处理方法、装置、系统、平台及计算机可读存储介质，通过在当前环境的光照强度小于预设光照强度时，根据该光照强度确定拍摄装置的单帧短曝光时间和目标拍摄次数，然后根据单帧短曝光时间和目标拍摄次数控制拍摄装置拍照，得到至少一张第一图像，最后通过基于训练图像和标注的参考图像对神经网络模型进行训练得到的预设暗光成像模型和该光照强度对应的亮度档位倍数对至少一张第一图像进行处理，得到第二图像，能够提高暗光环境下拍摄到的图像的质量和清晰度。
请参阅图1，图1是实施本申请实施例提供的图像处理方法的一场景示意图。如图1所示，该场景包括手持云台100和搭载于手持云台100上的拍摄装置200，手持云台100包括手柄部101和设于手柄部101上的云台102，云台102用于搭载拍摄装置200，拍摄装置200可以与云台102一体设置，也可以外接于云台102。示例性的，拍摄装置200可以为智能手机，也可以为相机，例如为单反相机，还可以为摄像头。手持云台100可以承载拍摄装置200，用于固定拍摄装置200以及改变拍摄装置200的高度、倾角和/或方向，或者用于将拍摄装置200稳定地保持在某一姿态上，并控制拍摄装置200进行拍摄。
在一实施例中，云台102包括三轴电机，三轴电机分别为俯仰(pitch)轴电机1021、平移(yaw)轴电机1022和横滚(roll)轴电机(图1中未示出)，所述三轴电机用于调整搭载于云台102上的拍摄装置200的平衡姿态，以便拍摄稳定流畅的画面。其中，云台102上还设置有惯性测量单元(Inertial measurement unit，IMU)，可例如为加速度计或陀螺仪中的至少一种，可以用于测量云台102的姿态和加速度等，以便根据姿态调整云台102的姿态。在一实施例中，手柄部101上也设置有惯性测量单元(Inertial measurement unit，IMU)，例如包括加速度计或陀螺仪中的至少一种，可以用于测量手柄部101的姿态和加速度等，以便根据手柄部101的姿态和云台102的姿态调整云台102的姿态。
在一实施例中，手持云台100与拍摄装置200通信连接，手持云台100可以通过控制线与拍摄装置200连接，该控制线例如为快门线。此处不限定快门线的种类，例如，该快门线可以是通用串行总线(Universal Serial Bus，USB)。手持云台100也可以通过无线的方式与拍摄装置200连接，例如，通过手持云 台100内置的第一蓝牙模块与拍摄装置200内置的第二蓝牙模块，建立手持云台100与拍摄装置200之间的通信连接。
在一实施例中，手持云台100还包括图像处理装置(图中未示出)，该图像处理装置设置于手柄部101的内部，在需要拍照时，该图像处理装置获取当前环境的光照强度，并在当前环境的光照强度小于预设光照强度时，根据该光照强度确定拍摄装置200的单帧短曝光时间和目标拍摄次数；根据该单帧短曝光时间和目标拍摄次数控制拍摄装置200拍照，得到至少一张第一图像；通过基于训练图像和标注的参考图像对神经网络模型进行训练得到的预设暗光成像模型和光照强度对应的亮度档位倍数对至少一张第一图像进行处理，得到第二图像，能够提高暗光环境下拍摄到的图像的质量和清晰度。其中，该图像处理装置可以借助摄像头等传感器来获取当前环境的光照强度，也可以借助服务器下载的天气数据来获取当前环境的光照强度。
在一实施例中，手柄部101上还设置有控制键，以便用户操作该控制键以控制云台102或拍摄装置200。该控制键可例如为按键、扳机、旋钮或者摇杆等，当然也包括其他形式的物理按键或者虚拟按键。其中，虚拟按键可以为设置于触摸屏上的用于和用户交互的虚拟按钮。摇杆可以用于控制至少一个转轴的运动，进而控制拍摄装置200的运动。可以理解的是，遥杆也可以用于其他功能。可以理解的是，控制键的数量可以为一个或多个。当控制键的数量为一个时，可以针对该控制键采用不同的操作方式产生不同的控制指令，不同的操作方式比如为按压次数不同；当控制键的数量为多个时，比如包括第一控制键、第二控制键和第三控制键等，不同控制键用于产生不同的控制指令。
请参阅图2，图2是实施本申请实施例提供的图像处理方法的另一场景示意图，如图2所示，该场景包括控制终端300和可移动平台400，控制终端300与可移动平台400通信连接，控制终端300包括显示装置310，显示装置310用于显示可移动平台400发送的图像。需要说明的是，显示装置310包括设置在控制终端300上的显示屏或者独立于控制终端300的显示器，独立于控制终端300的显示器可以包括手机、平板电脑或者个人电脑等，或者也可以是带有显示屏的其他电子设备。其中，该显示屏包括LED显示屏、OLED显示屏、LCD显示屏等等。
在一实施例中，可移动平台400包括平台本体410、搭载于平台本体上的云台420和动力系统430，云台420用于搭载拍摄装置500，动力系统430包括电机431和螺旋桨432，电机431用于驱动螺旋桨432旋转，从而为可移动平 台提供移动动力。其中，云台420包括三轴电机，分别为平移轴电机421、俯仰轴电机422和横滚轴电机423，用于调整搭载于云台420上的拍摄装置500的平衡姿态，以便随时随地拍摄出高精度的稳定画面。
在一实施例中，可移动平台400还包括图像处理装置(图中未示出)，该图像处理装置设置于平台本体410内部，在需要拍照时，该图像处理装置获取当前环境的光照强度，并在当前环境的光照强度小于预设光照强度时，根据该光照强度确定拍摄装置500的单帧短曝光时间和目标拍摄次数；根据该单帧短曝光时间和目标拍摄次数控制拍摄装置500拍照，得到至少一张第一图像；通过基于训练图像和标注的参考图像对神经网络模型进行训练得到的预设暗光成像模型和光照强度对应的亮度档位倍数对至少一张第一图像进行处理，得到第二图像，能够提高暗光环境下拍摄到的图像的质量和清晰度。其中，该图像处理装置可以借助摄像头等传感器来获取当前环境的光照强度，也可以借助服务器下载的天气数据来获取当前环境的光照强度。
其中，可移动平台包括可移动机器人、无人机和无人车等，可移动平台400为无人机，动力系统430能够使无人机垂直地从地面起飞，或者垂直地降落在地面上，而不需要无人机任何水平运动(如不需要在跑道上滑行)。可选的，动力系统430可以允许无人机在空中预设位置和/或方向盘旋。一个或者多个动力系统430在受到控制时可以独立于其它的动力系统430。可选的，一个或者多个动力系统430可以同时受到控制。例如，无人机可以有多个水平方向的动力系统430，以追踪目标的提升及/或推动。水平方向的动力系统430可以被致动以提供无人机垂直起飞、垂直降落、盘旋的能力。
在一实施例中，水平方向的动力系统430中的一个或者多个可以顺时针方向旋转，而水平方向的动力系统中的其它一个或者多个可以逆时针方向旋转。例如，顺时针旋转的动力系统430与逆时针旋转的动力系统430的数量一样。每一个水平方向的动力系统430的旋转速率可以独立变化，以实现每个动力系统430导致的提升及/或推动操作，从而调整无人机的空间方位、速度及/或加速度(如相对于多达三个自由度的旋转及平移)。
在一实施例中，无人机还可以包括传感系统，传感系统可以包括一个或者多个传感器，以感测无人机的空间方位、速度及/或加速度(如相对于多达三个自由度的旋转及平移)、角加速度、姿态、位置(绝对位置或者相对位置)等。所述一个或者多个传感器包括GPS传感器、运动传感器、惯性传感器、近程传感器或者影像传感器。可选的，传感系统还可以用于采集无人飞行器所处的环 境数据，如气候条件、要接近的潜在的障碍、地理特征的位置、人造结构的位置等。另外，无人机可以包括脚架，所述脚架是无人机降落时，无人机与地面的接触件，脚架可以是无人飞行器在飞行状态(例如无人飞行器在巡航时)收起，在降落时才放下；也可以固定安装在无人机上，一直处于放下的状态。
在一实施例中，可移动平台400能够与控制终端300进行通信，可以实现控制终端300与可移动平台400之间的数据交互，例如对可移动平台400的移动控制、对负载的控制(当负载为拍摄装置500时，控制终端300可以控制该拍摄装置500)，其中，控制终端300可以与可移动平台400和/或负载进行通信，可移动平台400与控制终端300之间的通信可以是无线通信，可以在可移动平台400和控制终端300之间提供直接通信。这种直接通信可以无需任何中间装置或网络地发生的。
在一实施例中，可以在可移动平台400与控制终端300之间提供间接通信。这种间接通信可以借助于一个或多个中间装置或网络来发生。例如，间接通信可以利用电信网络。间接通信可以借助于一个或多个路由器、通信塔、卫星、或任何其他的中间装置或网络来进行。通信类型的实例可以包括但不限于经由以下方式的通信：因特网，局域网(LAN)，广域网(WAN)，蓝牙，近场通信(NFC)技术，基于诸如通用分组无线电服务(GPRS)、GSM增强型数据GSM环境(EDGE)、3G、4G、或长期演进(LTE)协议的移动数据协议的网络，红外线(IR)通信技术，和/或Wi-Fi，并且可以是无线式、有线式、或其组合。
其中，控制终端300可以包括但不限于：智能电话/手机、平板电脑、个人数字助理(PDA)、台式计算机、媒体内容播放器、视频游戏站/系统、虚拟现实系统、增强现实系统、可穿戴式装置(例如，手表、眼镜、手套、头饰(例如，帽子、头盔、虚拟现实头戴耳机、增强现实头戴耳机、头装式装置(HMD)、头带)、挂件、臂章、腿环、鞋子、马甲)、手势识别装置、麦克风、能够提供或渲染图像数据的任意电子装置、或者任何其他类型的装置。该控制终端300可以是手持终端，控制终端300可以是便携式的。该控制终端300可以由人类用户携带。在一些情况下，控制终端300可以远离人类用户，并且用户可以使用无线和/或有线通信来控制控制终端300。
以下，将结合图1或图2中的场景对本申请的实施例提供的图像处理方法进行详细介绍。需知，图1或图2中的场景仅用于解释本申请实施例提供的图像处理方法，但并不构成对本申请实施例提供的图像处理方法应用场景的限定。
请参阅图3，图3是本申请实施例提供的一种图像处理方法的步骤示意流 程图。
如图3所示，该图像处理方法可以包括步骤S101至步骤S103。
步骤S101、在当前环境的光照强度小于预设光照强度时，根据所述光照强度确定拍摄装置的单帧短曝光时间和目标拍摄次数。
示例性的，在获取到拍摄指令时，获取拍摄装置所处当前环境的光照强度，并确定当前环境的光照强度是否小于预设光照强度，在当前环境的光照强度小于预设光照强度时，根据该光照强度确定拍摄装置的单帧短曝光时间和目标拍摄次数。其中，在当前环境的光照强度小于预设光照强度时，可以确定拍摄装置处于暗光环境，当检测到用户对拍摄按键的按压操作时，生成拍摄指令，拍摄装置所处当前环境的光照强度可以根据预置光照度传感器确定，预置光照传感器可以设置在拍摄装置中，预置光照传感器也可以设置在拍摄装置外，本实施例对此不做具体限定，单帧短曝光时间用于指示拍摄装置拍摄一帧图像的曝光时间，预设光照强度也可以基于实际情况进行设置，本实施例对此不做具体限定，例如，预设光照强度为1lux。
在一实施例中，根据该光照强度确定拍摄装置的单帧短曝光时间和目标拍摄次数的方式可以为：获取与光照强度对应的长曝光时间和亮度档位倍数；根据长曝光时间和亮度档位倍数，确定拍摄装置的单帧短曝光时间；根据预设的光照强度与目标拍摄次数之间的映射关系和该光照强度，确定拍摄装置的目标拍摄次数。通过确定暗光环境下的拍摄装置的目标拍摄次数、长曝光时间和亮度档位倍数，再通过长曝光时间和亮度档位倍数，可以确定单帧短曝光时间，便于后续拍摄装置在该单帧短曝光时间下拍摄到多张图像，使得通过对多张图像进行处理后的图像能够媲美长曝光下拍摄到的图像，以提高暗光环境下拍摄到的图像的质量和清晰度。
其中，拍摄装置的单帧短曝光时间和目标拍摄次数可以根据预存的光照强度、单帧短曝光时间、亮度档位倍数之间的映射关系和拍摄装置所处当前环境的光照强度确定，预设的光照强度与目标拍摄次数之间的映射关系以及预存的光照强度、单帧短曝光时间、亮度档位倍数之间的映射关系可以基于实际情况进行设置，本实施例对此不做具体限定。
在一实施例中，根据长曝光时间和亮度档位倍数，确定拍摄装置的单帧短曝光时间的方式可以为：在拍摄装置的ISO(感光度)固定的情况下，确定长曝光时间与亮度档位倍数的比值，并将长曝光时间与亮度档位倍数的比值确定为拍摄装置的单帧短曝光时间。例如，长曝光时间为4秒，亮度档位倍数为32， 则长曝光时间4秒与亮度档位倍数32的比值为0.125，因此拍摄装置的单帧短曝光时间为0.125秒。
步骤S102、根据所述单帧短曝光时间和目标拍摄次数控制所述拍摄装置拍照，得到至少一张第一图像。
示例性的，在控制拍摄装置拍摄时，控制拍摄装置的曝光时间，使得拍摄得到的第一图像的曝光时间为该单帧短短曝光时间，并在实际拍摄次数达到设定的目标拍摄次数时，控制拍摄装置停止拍摄，得到至少一张第一图像。例如，单帧短曝光时间为0.25秒，目标拍摄次数为8次，则拍摄图像时，每次控制拍摄装置的曝光时间为0.25秒，且需要拍摄8次，从而得到单帧短曝光时间为0.25秒的8张第一图像。
步骤S103、根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理，得到第二图像。
其中，所述预设暗光成像模型是根据训练图像和标注的参考图像对神经网络模型进行训练得到的，所述训练图像包括根据小于预设光照强度的不同光照强度对应的单帧短曝光时间和拍摄次数拍摄得到的图像，标注的参考图像包括根据小于预设光照强度的不同光照强度对应的长曝光时间拍摄得到的图像，通过大量的训练图像和对应标注的参考图像对神经网络模型进行训练，可以得到预设暗光成像模型，神经网络模型的具体层级结构可以根据实际情况进行设置，本实施对此不做具体限定。
在一实施例中，暗光成像模型的建立过程可以为：获取训练图像和该训练图像对应的参考图像，并根据光照强度对应的亮度档位倍数调整训练图像的亮度；对调整亮度后的训练图像进行融合，得到融合后的训练图像，并根据融合后的训练图像和对应的参考图像对神经网络模型进行迭代训练，直到训练后的神经网络模型收敛，从而得到暗光成像模型。
在一实施例中，暗光成像模型的建立过程还可以为：获取训练图像和该训练图像对应的参考图像，并根据光照强度对应的亮度档位倍数调整训练图像的亮度；根据调整亮度后的训练图像和该训练图像对应的参考图像，对神经网络模型进行迭代训练，直到训练后的神经网络模型收敛，从而得到暗光成像模型。
在一实施例中，如图4所示，步骤S103可以包括子步骤S1031至S1033。
子步骤S1031、根据所述光照强度对应的亮度档位倍数调整所述至少一张第一图像的亮度；
子步骤S1032、对调整亮度后的所述至少一张第一图像进行融合，得到第 三图像；
子步骤S1033、根据所述预设暗光成像模型对所述第三图像进行处理，得到第二图像。
其中，所述预设暗光成像模型能够将第三图像拆分为不同频段的图像，所述第二图像是根据不同频段的图像合成的。所述不同频段的图像包括至少一个第一频段的图像和至少一个第二频段的图像，第一频段与第二频段之间的频段差值大于预设频段差值，第一频段的图像包括第三图像中的低频成分，第二频段的图像包括第三图像中的高频成分。其中，第一图像的格式为RAW格式，不同频段的图像的格式为RGB格式或YUV格式，第二图像的格式为RGB格式或YUV格式。可以理解的是，预设频段差值可基于实际情况进行设置，本实施例对此不做具体限定。
在一实施例中，所述预设暗光成像模型包括图像拆分层和图像合成层，所述图像拆分层用于将第三图像拆分为不同频段的图像，所述图像合成层用于对不同频段的图像进行合成，以输出第二图像。所述图像拆分层包括至少一个低频成分拆分层和至少一个高频成分拆分层，所述低频成分拆分层用于从第三图像中拆分出低频成分，所述高频成分拆分层用于从第三图像中拆分出高频成分。
在一实施例中，在对神经网络模型进行训练时，低频成分拆分层和高频成分拆分层的损失函数不同。通过给低频成分拆分层和高频成分拆分层设置不同的损失函数来训练，可以提高模型对图像的处理效果，提高暗光环境下拍摄到的图像的质量和清晰度。其中，对于仅包含图像拆分层和图像合成层的预设暗光成像模型的建立过程可以为：获取训练图像和该训练图像对应的参考图像，并根据光照强度对应的亮度档位倍数调整训练图像的亮度；对调整亮度后的训练图像进行融合，得到融合后的训练图像，并根据融合后的训练图像和对应的参考图像对神经网络模型进行迭代训练，直到训练后的神经网络模型收敛，从而得到暗光成像模型。
示例性的，由于低频成分侧重于恢复图像的色彩和粗略的结构纹理等信息，因此，所述低频成分拆分层的损失函数是根据低频成分中各像素点所属的图像块的损失确定的。或者所述低频成分拆分层的损失函数是根据低频成分中各像素点所属的图像块的损失和所述图像块的损失的权重系数确定的，所述图像块的损失的权重系数是根据标注的参考图像中的低频成分确定的。通过设置与低频成分适配的损失函数，便于训练出效果好的低频成分拆分层，使得通过暗光成像模型处理图像时，可以提高图像中的低频成分的质量和清晰度。
示例性的，由于高频成分侧重于恢复图像的边缘和纹理等信息，因此，所述高频成分拆分层的损失函数是根据高频成分拆分中的各像素点的单点损失和像素点的单点损失的权重系数确定的，所述像素点的单点损失的权重系数是根据标注的参考图像中的高频成分确定的。通过设置与高频成分适配的损失函数，便于训练出效果好的高频成分拆分层，使得通过暗光成像模型处理图像时，可以提高图像中的低频成分的质量和清晰度。
由于图像的边缘、纹理等细节的高频成分占的像素比例的极少的，如果边缘、纹理等细节处于平坦区使用同样比重的损失函数，将导致训练出的高频成分拆分层的效果不好，因此，对于高频成分拆分层的训练可以分为两个阶段，第一个阶段是利用高频成分中的像素点的单点损失对高频成分拆分层进行训练，使得训练后的高频成分拆分层输出的高频成分具有与标注的参考图像中的高频成分一致的形态；第二阶段是根据标注的参考图像中的高频成分生成一张权重系数表，并根据生成的权重系数表重新训练高频成分拆分层，使得高频成分拆分层能够较好的恢复图像的边缘和纹理等信息。其中，该权重系数表中，高频成分中处于平坦区的像素点对应的权重系数较小，高频成分中的纹理和边缘的像素点对应的权重系数较大。
其中，预设暗光成像模型中的低频成分拆分层和高频成分拆分层的数量可基于实际情况进行设置，例如，预设暗光成像模型包括一个低频成分拆分层和三个高频成分拆分层，且三个高频成分拆分层包括第一高频成分拆分层、第二高频成分拆分层和第三高频成分拆分层，则预设暗光成像模型对第三图像的处理过程可以为：通过低频成分拆分层从第三图像中拆分出低频成分，通过第一高频成分拆分层从第三图像中拆分出第一高频成分，通过第二高频成分拆分层从第三图像中拆分出第二高频成分，通过第三高频成分拆分层从第三图像中拆分出第三高频成分，最后通过图像合成层对从第三图像中拆分出的低频成分、第一高频成分、第二高频成分和第三高频成分进行合成，得到第二图像。
在一实施例中，根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理，得到第二图像的方式可以为：根据光照强度对应的亮度档位倍数调整至少一张第一图像的亮度；根据预设暗光成像模型对调整亮度后的至少一张第一图像进行处理，得到第二图像。其中，预设暗光成像模型包括图像融合层、图像拆分层和图像合成层，该图像融合层用于对调整亮度后的至少一张第一图像进行融合。
其中，对于包含图像融合层、图像拆分层和图像合成层的预设暗光成像模 型的建立过程可以为：获取训练图像和该训练图像对应的参考图像，并根据光照强度对应的亮度档位倍数调整训练图像的亮度；根据调整亮度后的训练图像和该训练图像对应的参考图像，对神经网络模型进行迭代训练，直到训练后的神经网络模型收敛，从而得到暗光成像模型。
在一实施例中，根据预设暗光成像模型对调整亮度后的至少一张第一图像进行处理，得到第二图像的方式可以为：通过图像融合层对调整亮度后的至少一张第一图像进行融合，得到第三图像；通过图像拆分层将第三图像拆分为不同频段的图像，并通过图像合成层对所述不同频段的图像进行合成，得到第二图像。通过包含图像融合层、图像拆分层和图像合成层的预设暗光成像模型可以快速的输出图像质量和清晰度较好的图像。
在一实施例中，在当前环境的光照强度小于预设光照强度时，确定拍摄装置是否处于运动状态；若拍摄装置处于运动状态，则根据光照强度确定拍摄装置的单帧短曝光时间和目标拍摄次数；根据单帧短曝光时间和目标拍摄次数控制拍摄装置拍照，得到至少一张第一图像；根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对至少一张第一图像进行处理，得到第二图像。其中，拍摄装置内设置有惯性测量单元(Inertial Measurement Unit，IMU)，通过该惯性测量单元可以确定拍摄装置是否处于运动状态，另外，对于拍摄装置搭载于手持云台或可移动平台的云台上的场景，也可以通过手持云台内的惯性测量单元或可移动平台内的惯性测量单元来确定拍摄装置是否处于运动状态。在拍摄装置处于运动状态，且当前环境的光照强度小于预设光照强度时，通过上述方案可以提高暗光环境下拍摄到的图像的质量和清晰度。
在一实施例中，在当前环境的光照强度小于预设光照强度时，根据光照强度确定拍摄装置的单帧短曝光时间和目标拍摄次数；控制拍摄装置调整红外线滤光片的位置，使得拍摄装置所处环境中的红外光能够进入拍摄装置；根据单帧短曝光时间和目标拍摄次数控制拍摄装置拍照，得到至少一张第一图像；根据预设暗光成像模型和光照强度对应的亮度档位倍数对至少一张第一图像进行处理，得到第二图像。其中，拍摄装置包括红外滤光片，该红外滤光片用于过滤红外光。通过调整红外滤光片的位置，使得拍摄装置在基于单帧短曝光时间和拍摄次数拍照时，红外光能够进入拍摄装置，以提高拍照得到的图像的亮度，之后通过暗光成像模型和光照强度对应的亮度档位倍数对图像进行处理，可以进一步地提高暗光环境下拍摄到的图像的质量和清晰度。
上述实施例提供的图像处理方法，通过在当前环境的光照强度小于预设光 照强度时，根据该光照强度确定拍摄装置的单帧短曝光时间和目标拍摄次数，然后根据单帧短曝光时间和目标拍摄次数控制拍摄装置拍照，得到至少一张第一图像，最后通过基于训练图像和标注的参考图像对神经网络模型进行训练得到的预设暗光成像模型和该光照强度对应的亮度档位倍数对至少一张第一图像进行处理，得到第二图像，能够提高暗光环境下拍摄到的图像的质量和清晰度。
请参阅图5，图5是本申请实施例提供的一种图像处理装置的结构示意性框图。
如图5所示，图像处理装置600包括处理器601和存储器602，处理器601和存储器602通过总线603连接，该总线603比如为I2C(Inter-integrated Circuit)总线。图像处理装置600用于控制拍摄装置。图像处理装置例如可以是芯片或者处理装置等。
具体地，处理器601可以是微控制单元(Micro-controller Unit，MCU)、中央处理单元(Central Processing Unit，CPU)或数字信号处理器(Digital Signal Processor，DSP)等。
具体地，存储器602可以是Flash芯片、只读存储器(ROM，Read-Only Memory)磁盘、光盘、U盘或移动硬盘等。
其中，所述处理器601用于运行存储在存储器602中的计算机程序，并在执行所述计算机程序时实现如下步骤：
在当前环境的光照强度小于预设光照强度时，根据所述光照强度确定拍摄装置的单帧短曝光时间和目标拍摄次数；
根据所述单帧短曝光时间和目标拍摄次数控制所述拍摄装置拍照，得到至少一张第一图像；
根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理，得到第二图像，其中，所述预设暗光成像模型是根据训练图像和标注的参考图像对神经网络模型进行训练得到的。
在一实施例中，所述处理器实现根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理时，用于实现：
根据所述光照强度对应的亮度档位倍数调整所述至少一张第一图像的亮度；
对调整亮度后的所述至少一张第一图像进行融合，得到第三图像；
根据所述预设暗光成像模型对所述第三图像进行处理，得到第二图像。
在一实施例中，所述预设暗光成像模型能够将所述第三图像拆分为不同频段的图像，所述第二图像是根据所述不同频段的图像合成的。
在一实施例中，所述预设暗光成像模型包括图像拆分层和图像合成层，所述图像拆分层用于将所述第三图像拆分为不同频段的图像，所述图像合成层用于对所述不同频段的图像进行合成，以输出所述第二图像。
在一实施例中，所述图像拆分层包括至少一个低频成分拆分层和至少一个高频成分拆分层，所述低频成分拆分层用于从所述第三图像中拆分出低频成分，所述高频成分拆分层用于从所述第三图像中拆分出高频成分。
在一实施例中，所述低频成分拆分层的损失函数是根据所述低频成分中各像素点所属的图像块的损失确定的。
在一实施例中，所述低频成分拆分层的损失函数是根据所述低频成分中各像素点所属的图像块的损失和所述图像块的损失的权重系数确定的。
在一实施例中，所述高频成分拆分层的损失函数是根据所述高频成分拆分中的各像素点的单点损失和所述像素点的单点损失的权重系数确定的。
在一实施例中，所述像素点的单点损失的权重系数是根据标注的参考图像中的高频成分确定的。
在一实施例中，所述训练图像包括根据小于预设光照强度的不同光照强度对应的单帧短曝光时间和拍摄次数拍摄得到的图像，所述参考图像包括根据小于预设光照强度的不同光照强度对应的长曝光时间拍摄得到的图像。
在一实施例中，所述根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理，得到第二图像，包括：
根据所述光照强度对应的亮度档位倍数调整所述至少一张第一图像的亮度；
根据所述预设暗光成像模型对调整亮度后的所述至少一张第一图像进行处理，得到第二图像。
在一实施例中，所述预设暗光成像模型包括图像融合层、图像拆分层和图像合成层，所述图像融合层用于对调整亮度后的所述至少一张第一图像进行融合，所述根据所述预设暗光成像模型对调整亮度后的所述至少一张第一图像进行处理，得到第二图像，包括：
通过所述图像融合层对调整亮度后的所述至少一张第一图像进行融合，得到第三图像；
通过所述图像拆分层将所述第三图像拆分为不同频段的图像，并通过所述图像合成层对所述不同频段的图像进行合成，得到第二图像。
在一实施例中，所述拍摄装置包括红外线滤光片，所述处理器实现根据所述单帧短曝光时间和目标拍摄次数控制所述拍摄装置拍照，得到至少一张第一 图像之前，还用于实现：
控制所述拍摄装置调整所述红外线滤光片的位置，使得所述拍摄装置所处环境中的红外光能够进入所述拍摄装置。
在一实施例中，所述处理器实现根据所述光照强度确定所述拍摄装置的单帧短曝光时间和目标拍摄次数时，用于实现：
获取与所述光照强度对应的长曝光时间和亮度档位倍数；
根据所述长曝光时间和所述亮度档位倍数，确定所述拍摄装置的单帧短曝光时间；
根据预设的光照强度与目标拍摄次数之间的映射关系和所述光照强度，确定所述拍摄装置的目标拍摄次数。
在一实施例中，所述处理器实现根据所述光照强度确定所述拍摄装置的单帧短曝光时间和目标拍摄次数之前，还用于实现：
在当前环境的光照强度小于预设光照强度时，确定所述拍摄装置是否处于运动状态；
若所述拍摄装置处于运动状态，则根据所述光照强度确定所述拍摄装置的单帧短曝光时间和目标拍摄次数。
需要说明的是，所属领域的技术人员可以清楚地了解到，为了描述的方便和简洁，上述描述的图像处理装置的具体工作过程，可以参考前述图像处理方法实施例中的对应过程，在此不再赘述。
请参阅图6，图6是本申请实施例提供的一种拍摄装置的结构示意性框图。
如图6所示，该拍摄装置700包括处理器701和存储器702，处理器701和存储器702通过总线703连接，该总线703比如为I2C(Inter-integrated Circuit)总线。
具体地，处理器701可以是微控制单元(Micro-controller Unit，MCU)、中央处理单元(Central Processing Unit，CPU)或数字信号处理器(Digital Signal Processor，DSP)等。
具体地，存储器702可以是Flash芯片、只读存储器(ROM，Read-Only Memory)磁盘、光盘、U盘或移动硬盘等。
其中，所述处理器701用于运行存储在存储器702中的计算机程序，并在执行所述计算机程序时实现如下步骤：
在当前环境的光照强度小于预设光照强度时，根据所述光照强度确定拍摄装置的单帧短曝光时间和目标拍摄次数；
根据所述单帧短曝光时间和目标拍摄次数控制所述拍摄装置拍照，得到至少一张第一图像；
根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理，得到第二图像，其中，所述预设暗光成像模型是根据训练图像和标注的参考图像对神经网络模型进行训练得到的。
在一实施例中，所述处理器实现根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理时，用于实现：
根据所述光照强度对应的亮度档位倍数调整所述至少一张第一图像的亮度；
对调整亮度后的所述至少一张第一图像进行融合，得到第三图像；
根据所述预设暗光成像模型对所述第三图像进行处理，得到第二图像。
在一实施例中，所述预设暗光成像模型能够将所述第三图像拆分为不同频段的图像，所述第二图像是根据所述不同频段的图像合成的。
在一实施例中，所述预设暗光成像模型包括图像拆分层和图像合成层，所述图像拆分层用于将所述第三图像拆分为不同频段的图像，所述图像合成层用于对所述不同频段的图像进行合成，以输出所述第二图像。
在一实施例中，所述图像拆分层包括至少一个低频成分拆分层和至少一个高频成分拆分层，所述低频成分拆分层用于从所述第三图像中拆分出低频成分，所述高频成分拆分层用于从所述第三图像中拆分出高频成分。
在一实施例中，所述低频成分拆分层的损失函数是根据所述低频成分中各像素点所属的图像块的损失确定的。
在一实施例中，所述低频成分拆分层的损失函数是根据所述低频成分中各像素点所属的图像块的损失和所述图像块的损失的权重系数确定的。
在一实施例中，所述高频成分拆分层的损失函数是根据所述高频成分拆分中的各像素点的单点损失和所述像素点的单点损失的权重系数确定的。
在一实施例中，所述像素点的单点损失的权重系数是根据标注的参考图像中的高频成分确定的。
在一实施例中，所述训练图像包括根据小于预设光照强度的不同光照强度对应的单帧短曝光时间和拍摄次数拍摄得到的图像，所述参考图像包括根据小于预设光照强度的不同光照强度对应的长曝光时间拍摄得到的图像。
在一实施例中，所述根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理，得到第二图像，包括：
根据所述光照强度对应的亮度档位倍数调整所述至少一张第一图像的亮度；
根据所述预设暗光成像模型对调整亮度后的所述至少一张第一图像进行处理，得到第二图像。
在一实施例中，所述预设暗光成像模型包括图像融合层、图像拆分层和图像合成层，所述图像融合层用于对调整亮度后的所述至少一张第一图像进行融合，所述根据所述预设暗光成像模型对调整亮度后的所述至少一张第一图像进行处理，得到第二图像，包括：
通过所述图像融合层对调整亮度后的所述至少一张第一图像进行融合，得到第三图像；
通过所述图像拆分层将所述第三图像拆分为不同频段的图像，并通过所述图像合成层对所述不同频段的图像进行合成，得到第二图像。
在一实施例中，所述拍摄装置包括红外线滤光片，所述处理器实现根据所述单帧短曝光时间和目标拍摄次数控制所述拍摄装置拍照，得到至少一张第一图像之前，还用于实现：
控制所述拍摄装置调整所述红外线滤光片的位置，使得所述拍摄装置所处环境中的红外光能够进入所述拍摄装置。
在一实施例中，所述处理器实现根据所述光照强度确定所述拍摄装置的单帧短曝光时间和目标拍摄次数时，用于实现：
获取与所述光照强度对应的长曝光时间和亮度档位倍数；
根据所述长曝光时间和所述亮度档位倍数，确定所述拍摄装置的单帧短曝光时间；
根据预设的光照强度与目标拍摄次数之间的映射关系和所述光照强度，确定所述拍摄装置的目标拍摄次数。
在一实施例中，所述处理器实现根据所述光照强度确定所述拍摄装置的单帧短曝光时间和目标拍摄次数之前，还用于实现：
在当前环境的光照强度小于预设光照强度时，确定所述拍摄装置是否处于运动状态；
若所述拍摄装置处于运动状态，则根据所述光照强度确定所述拍摄装置的单帧短曝光时间和目标拍摄次数。
需要说明的是，所属领域的技术人员可以清楚地了解到，为了描述的方便和简洁，上述描述的拍摄装置的具体工作过程，可以参考前述图像处理方法实施例中的对应过程，在此不再赘述。
请参阅图7，图7是本申请实施例提供的一种可移动平台的结构示意性框 图。
如图7所示，可移动平台800包括：
平台本体；
动力系统810，所述动力系统810设于所述平台本体上，所述动力系统810用于为所述可移动平台提供移动动力；
云台820，所述云台820搭载于所述平台本体，所述云台820用于搭载拍摄装置；
图像处理装置830，所述图像处理装置830设于所述平台本体上，所述图像处理装置830还用于控制所述可移动平台800移动。
需要说明的是，所属领域的技术人员可以清楚地了解到，为了描述的方便和简洁，上述描述的可移动平台的具体工作过程，可以参考前述图像处理方法实施例中的对应过程，在此不再赘述。
请参阅图8，图8是本申请实施例提供的一种拍摄系统的结构示意性框图。
如图8所示，拍摄系统900包括云台910、搭载于云台910的拍摄装置920、图像处理装置930。其中，云台910连接于手柄部，图像处理装置930设置在手柄部上，或者云台910和图像处理装置930设置在可移动平台上，图像处理装置930还用于控制可移动平台移动。
需要说明的是，所属领域的技术人员可以清楚地了解到，为了描述的方便和简洁，上述描述的拍摄系统的具体工作过程，可以参考前述图像处理方法实施例中的对应过程，在此不再赘述。
本申请实施例还提供一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，所述计算机程序中包括程序指令，所述处理器执行所述程序指令，实现上述实施例提供的图像处理方法的步骤。
其中，所述计算机可读存储介质可以是前述任一实施例所述的可移动平台、手持云台或拍摄装置的内部存储单元，例如所述可移动平台、手持云台或拍摄装置的硬盘或内存。所述计算机可读存储介质也可以是所述可移动平台、手持云台或拍摄装置的外部存储设备，例如所述可移动平台、手持云台或拍摄装置上配备的插接式硬盘，智能存储卡(Smart Media Card，SMC)，安全数字(Secure Digital，SD)卡，闪存卡(Flash Card)等。
应当理解，在此本申请说明书中所使用的术语仅仅是出于描述特定实施例的目的而并不意在限制本申请。如在本申请说明书和所附权利要求书中所使用的那样，除非上下文清楚地指明其它情况，否则单数形式的“一”、“一个”及“该” 意在包括复数形式。
还应当理解，在本申请说明书和所附权利要求书中使用的术语“和/或”是指相关联列出的项中的一个或多个的任何组合以及所有可能组合，并且包括这些组合。
以上所述，仅为本申请的具体实施方式，但本申请的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到各种等效的修改或替换，这些修改或替换都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应以权利要求的保护范围为准。

Claims (36)

  1. 一种图像处理方法，其特征在于，包括：
    在当前环境的光照强度小于预设光照强度时，根据所述光照强度确定拍摄装置的单帧短曝光时间和目标拍摄次数；
    根据所述单帧短曝光时间和目标拍摄次数控制所述拍摄装置拍照，得到至少一张第一图像；
    根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理，得到第二图像，其中，所述预设暗光成像模型是根据训练图像和标注的参考图像对神经网络模型进行训练得到的。
  2. 根据权利要求1所述的图像处理方法，其特征在于，所述根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理，包括：
    根据所述光照强度对应的亮度档位倍数调整所述至少一张第一图像的亮度；
    对调整亮度后的所述至少一张第一图像进行融合，得到第三图像；
    根据所述预设暗光成像模型对所述第三图像进行处理，得到第二图像。
  3. 根据权利要求2所述的图像处理方法，其特征在于，所述预设暗光成像模型能够将所述第三图像拆分为不同频段的图像，所述第二图像是根据所述不同频段的图像合成的。
  4. 根据权利要求2所述的图像处理方法，其特征在于，所述预设暗光成像模型包括图像拆分层和图像合成层，所述图像拆分层用于将所述第三图像拆分为不同频段的图像，所述图像合成层用于对所述不同频段的图像进行合成，以输出所述第二图像。
  5. 根据权利要求4所述的图像处理方法，其特征在于，所述图像拆分层包括至少一个低频成分拆分层和至少一个高频成分拆分层，所述低频成分拆分层用于从所述第三图像中拆分出低频成分，所述高频成分拆分层用于从所述第三图像中拆分出高频成分。
  6. 根据权利要求5所述的图像处理方法，其特征在于，所述低频成分拆分层的损失函数是根据所述低频成分中各像素点所属的图像块的损失确定的。
  7. 根据权利要求5所述的图像处理方法，其特征在于，所述低频成分拆分层的损失函数是根据所述低频成分中各像素点所属的图像块的损失和所述图像 块的损失的权重系数确定的。
  8. 根据权利要求5所述的图像处理方法，其特征在于，所述高频成分拆分层的损失函数是根据所述高频成分拆分中的各像素点的单点损失和所述像素点的单点损失的权重系数确定的。
  9. 根据权利要求8所述的图像处理方法，其特征在于，所述像素点的单点损失的权重系数是根据标注的参考图像中的高频成分确定的。
  10. 根据权利要求1-9中任一项所述的图像处理方法，其特征在于，所述训练图像包括根据小于预设光照强度的不同光照强度对应的单帧短曝光时间和拍摄次数拍摄得到的图像，所述参考图像包括根据小于预设光照强度的不同光照强度对应的长曝光时间拍摄得到的图像。
  11. 根据权利要求1-9中任一项所述的图像处理方法，其特征在于，所述根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理，得到第二图像，包括：
    根据所述光照强度对应的亮度档位倍数调整所述至少一张第一图像的亮度；
    根据所述预设暗光成像模型对调整亮度后的所述至少一张第一图像进行处理，得到第二图像。
  12. 根据权利要求11所述的图像处理方法，其特征在于，所述预设暗光成像模型包括图像融合层、图像拆分层和图像合成层，所述图像融合层用于对调整亮度后的所述至少一张第一图像进行融合，所述根据所述预设暗光成像模型对调整亮度后的所述至少一张第一图像进行处理，得到第二图像，包括：
    通过所述图像融合层对调整亮度后的所述至少一张第一图像进行融合，得到第三图像；
    通过所述图像拆分层将所述第三图像拆分为不同频段的图像，并通过所述图像合成层对所述不同频段的图像进行合成，得到第二图像。
  13. 根据权利要求1-9中任一项所述的图像处理方法，其特征在于，所述拍摄装置包括红外线滤光片，所述根据所述单帧短曝光时间和目标拍摄次数控制所述拍摄装置拍照，得到至少一张第一图像之前，还包括：
    控制所述拍摄装置调整所述红外线滤光片的位置，使得所述拍摄装置所处环境中的红外光能够进入所述拍摄装置。
  14. 根据权利要求1-9中任一项所述的图像处理方法，其特征在于，所述根据所述光照强度确定所述拍摄装置的单帧短曝光时间和目标拍摄次数，包括：
    获取与所述光照强度对应的长曝光时间和亮度档位倍数；
    根据所述长曝光时间和所述亮度档位倍数，确定所述拍摄装置的单帧短曝光时间；
    根据预设的光照强度与目标拍摄次数之间的映射关系和所述光照强度，确定所述拍摄装置的目标拍摄次数。
  15. 根据权利要求1-9中任一项所述的图像处理方法，其特征在于，所述根据所述光照强度确定所述拍摄装置的单帧短曝光时间和目标拍摄次数之前，还包括：
    在当前环境的光照强度小于预设光照强度时，确定所述拍摄装置是否处于运动状态；
    若所述拍摄装置处于运动状态，则根据所述光照强度确定所述拍摄装置的单帧短曝光时间和目标拍摄次数。
  16. 一种图像处理装置，其特征在于，所述图像处理装置用于控制拍摄装置，所述图像处理装置包括存储器和处理器；
    所述存储器用于存储计算机程序；
    所述处理器，用于执行所述计算机程序并在执行所述计算机程序时，实现如下步骤：
    在当前环境的光照强度小于预设光照强度时，根据所述光照强度确定拍摄装置的单帧短曝光时间和目标拍摄次数；
    根据所述单帧短曝光时间和目标拍摄次数控制所述拍摄装置拍照，得到至少一张第一图像；
    根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理，得到第二图像，其中，所述预设暗光成像模型是根据训练图像和标注的参考图像对神经网络模型进行训练得到的。
  17. 根据权利要求16所述的图像处理装置，其特征在于，所述处理器实现根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理时，用于实现：
    根据所述光照强度对应的亮度档位倍数调整所述至少一张第一图像的亮度；
    对调整亮度后的所述至少一张第一图像进行融合，得到第三图像；
    根据所述预设暗光成像模型对所述第三图像进行处理，得到第二图像。
  18. 根据权利要求17所述的图像处理装置，其特征在于，所述预设暗光成像模型能够将所述第三图像拆分为不同频段的图像，所述第二图像是根据所述不同频段的图像合成的。
  19. 根据权利要求17所述的图像处理装置，其特征在于，所述预设暗光成像模型包括图像拆分层和图像合成层，所述图像拆分层用于将所述第三图像拆分为不同频段的图像，所述图像合成层用于对所述不同频段的图像进行合成，以输出所述第二图像。
  20. 根据权利要求19所述的图像处理装置，其特征在于，所述图像拆分层包括至少一个低频成分拆分层和至少一个高频成分拆分层，所述低频成分拆分层用于从所述第三图像中拆分出低频成分，所述高频成分拆分层用于从所述第三图像中拆分出高频成分。
  21. 根据权利要求20所述的图像处理装置，其特征在于，所述低频成分拆分层的损失函数是根据所述低频成分中各像素点所属的图像块的损失确定的。
  22. 根据权利要求20所述的图像处理装置，其特征在于，所述低频成分拆分层的损失函数是根据所述低频成分中各像素点所属的图像块的损失和所述图像块的损失的权重系数确定的。
  23. 根据权利要求20所述的图像处理装置，其特征在于，所述高频成分拆分层的损失函数是根据所述高频成分拆分中的各像素点的单点损失和所述像素点的单点损失的权重系数确定的。
  24. 根据权利要求23所述的图像处理装置，其特征在于，所述像素点的单点损失的权重系数是根据标注的参考图像中的高频成分确定的。
  25. 根据权利要求16-24中任一项所述的图像处理装置，其特征在于，所述训练图像包括根据小于预设光照强度的不同光照强度对应的单帧短曝光时间和拍摄次数拍摄得到的图像，所述参考图像包括根据小于预设光照强度的不同光照强度对应的长曝光时间拍摄得到的图像。
  26. 根据权利要求16-24中任一项所述的图像处理装置，其特征在于，所述根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理，得到第二图像，包括：
    根据所述光照强度对应的亮度档位倍数调整所述至少一张第一图像的亮度；
    根据所述预设暗光成像模型对调整亮度后的所述至少一张第一图像进行处理，得到第二图像。
  27. 根据权利要求26所述的图像处理装置，其特征在于，所述预设暗光成像模型包括图像融合层、图像拆分层和图像合成层，所述图像融合层用于对调整亮度后的所述至少一张第一图像进行融合，所述根据所述预设暗光成像模型对调整亮度后的所述至少一张第一图像进行处理，得到第二图像，包括：
    通过所述图像融合层对调整亮度后的所述至少一张第一图像进行融合，得到第三图像；
    通过所述图像拆分层将所述第三图像拆分为不同频段的图像，并通过所述图像合成层对所述不同频段的图像进行合成，得到第二图像。
  28. 根据权利要求16-24中任一项所述的图像处理装置，其特征在于，所述拍摄装置包括红外线滤光片，所述处理器实现根据所述单帧短曝光时间和目标拍摄次数控制所述拍摄装置拍照，得到至少一张第一图像之前，还用于实现：
    控制所述拍摄装置调整所述红外线滤光片的位置，使得所述拍摄装置所处环境中的红外光能够进入所述拍摄装置。
  29. 根据权利要求16-24中任一项所述的图像处理装置，其特征在于，所述处理器实现根据所述光照强度确定所述拍摄装置的单帧短曝光时间和目标拍摄次数时，用于实现：
    获取与所述光照强度对应的长曝光时间和亮度档位倍数；
    根据所述长曝光时间和所述亮度档位倍数，确定所述拍摄装置的单帧短曝光时间；
    根据预设的光照强度与目标拍摄次数之间的映射关系和所述光照强度，确定所述拍摄装置的目标拍摄次数。
  30. 根据权利要求16-24中任一项所述的图像处理装置，其特征在于，所述处理器实现根据所述光照强度确定所述拍摄装置的单帧短曝光时间和目标拍摄次数之前，还用于实现：
    在当前环境的光照强度小于预设光照强度时，确定所述拍摄装置是否处于运动状态；
    若所述拍摄装置处于运动状态，则根据所述光照强度确定所述拍摄装置的单帧短曝光时间和目标拍摄次数。
  31. 一种拍摄装置，其特征在于，所述拍摄装置包括存储器和处理器；
    所述存储器用于存储计算机程序；
    所述处理器，用于执行所述计算机程序并在执行所述计算机程序时，实现如下步骤：
    在当前环境的光照强度小于预设光照强度时，根据所述光照强度确定拍摄装置的单帧短曝光时间和目标拍摄次数；
    根据所述单帧短曝光时间和目标拍摄次数控制所述拍摄装置拍照，得到至少一张第一图像；
    根据预设暗光成像模型和所述光照强度对应的亮度档位倍数对所述至少一张第一图像进行处理，得到第二图像，其中，所述预设暗光成像模型是根据训练图像和标注的参考图像对神经网络模型进行训练得到的。
  32. 一种可移动平台，其特征在于，所述可移动平台包括：
    平台本体；
    动力系统，所述动力系统设于所述平台本体上，所述动力系统用于为所述可移动平台提供移动动力；
    云台，所述云台搭载于所述平台本体，所述云台用于搭载拍摄装置；
    如权利要求16-30中任一项所述的图像处理装置，所述图像处理装置设于所述平台本体上，所述图像处理装置还用于控制所述可移动平台移动。
  33. 一种拍摄系统，其特征在于，所述拍摄系统包括云台、搭载于所述云台的拍摄装置和如权利要求16-30中任一项所述的图像处理装置。
  34. 根据权利要求33所述的拍摄系统，其特征在于，所述云台连接于手柄部，所述图像处理装置设置在所述手柄部上。
  35. 根据权利要求33所述的拍摄系统，其特征在于，所述云台和所述图像处理装置设置在可移动平台上，所述图像处理装置还用于控制所述可移动平台移动。
  36. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质存储有计算机程序，所述计算机程序被处理器执行时使所述处理器实现如权利要求1-15中任一项所述的图像处理方法的步骤。
PCT/CN2020/118577 2020-09-28 2020-09-28 图像处理方法、装置、系统、平台及计算机可读存储介质 WO2022061934A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2020/118577 WO2022061934A1 (zh) 2020-09-28 2020-09-28 图像处理方法、装置、系统、平台及计算机可读存储介质
CN202080015619.3A CN113491099A (zh) 2020-09-28 2020-09-28 图像处理方法、装置、系统、平台及计算机可读存储介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/118577 WO2022061934A1 (zh) 2020-09-28 2020-09-28 图像处理方法、装置、系统、平台及计算机可读存储介质

Publications (1)

Publication Number Publication Date
WO2022061934A1 true WO2022061934A1 (zh) 2022-03-31

Family

ID=77933698

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/118577 WO2022061934A1 (zh) 2020-09-28 2020-09-28 图像处理方法、装置、系统、平台及计算机可读存储介质

Country Status (2)

Country Link
CN (1) CN113491099A (zh)
WO (1) WO2022061934A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117475498B (zh) * 2023-12-28 2024-03-15 苏州元脑智能科技有限公司 自适应目标检测方法及装置

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110050937A1 (en) * 2009-08-26 2011-03-03 Altek Corporation Backlight photographing method
US20150015740A1 (en) * 2013-07-10 2015-01-15 Samsung Electronics Co., Ltd. Image processing method for improving image quality and image processing device therewith
CN109218628A (zh) * 2018-09-20 2019-01-15 Oppo广东移动通信有限公司 图像处理方法、装置、电子设备及存储介质
CN109218627A (zh) * 2018-09-18 2019-01-15 Oppo广东移动通信有限公司 图像处理方法、装置、电子设备及存储介质
CN110072051A (zh) * 2019-04-09 2019-07-30 Oppo广东移动通信有限公司 基于多帧图像的图像处理方法和装置

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108965731A (zh) * 2018-08-22 2018-12-07 Oppo广东移动通信有限公司 一种暗光图像处理方法及装置、终端、存储介质
CN111064904A (zh) * 2019-12-26 2020-04-24 深圳深知未来智能有限公司 一种暗光图像增强方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110050937A1 (en) * 2009-08-26 2011-03-03 Altek Corporation Backlight photographing method
US20150015740A1 (en) * 2013-07-10 2015-01-15 Samsung Electronics Co., Ltd. Image processing method for improving image quality and image processing device therewith
CN109218627A (zh) * 2018-09-18 2019-01-15 Oppo广东移动通信有限公司 图像处理方法、装置、电子设备及存储介质
CN109218628A (zh) * 2018-09-20 2019-01-15 Oppo广东移动通信有限公司 图像处理方法、装置、电子设备及存储介质
CN110072051A (zh) * 2019-04-09 2019-07-30 Oppo广东移动通信有限公司 基于多帧图像的图像处理方法和装置

Also Published As

Publication number Publication date
CN113491099A (zh) 2021-10-08

Similar Documents

Publication Publication Date Title
US10936894B2 (en) Systems and methods for processing image data based on region-of-interest (ROI) of a user
US11385645B2 (en) Remote control method and terminal
US10979615B2 (en) System and method for providing autonomous photography and videography
WO2019242553A1 (zh) 控制拍摄装置的拍摄角度的方法、控制装置及可穿戴设备
US20190291864A1 (en) Transformable apparatus
WO2018072155A1 (zh) 一种用于控制无人机的穿戴式设备及无人机系统
WO2018205104A1 (zh) 无人机拍摄控制方法、无人机拍摄方法、控制终端、无人机控制装置和无人机
WO2019227333A1 (zh) 集体照拍摄方法和装置
CN110291777B (zh) 图像采集方法、设备及机器可读存储介质
CN105847682A (zh) 全景图像的拍摄方法、装置及系统
US20220350330A1 (en) Remote control method and terminal
WO2020014953A1 (zh) 一种图像处理方法及设备
WO2022061934A1 (zh) 图像处理方法、装置、系统、平台及计算机可读存储介质
WO2020168519A1 (zh) 拍摄参数的调整方法、拍摄设备以及可移动平台
TWI436270B (zh) 虛擬望遠方法及其裝置
WO2022109860A1 (zh) 跟踪目标对象的方法和云台
WO2022188151A1 (zh) 影像拍摄方法、控制装置、可移动平台和计算机存储介质
US20210092306A1 (en) Movable body, image generation method, program, and recording medium
WO2022056683A1 (zh) 视场确定方法、视场确定装置、视场确定系统和介质
WO2021232424A1 (zh) 飞行辅助方法及装置、无人飞行器、遥控器、显示器、无人飞行器系统和存储介质
WO2018010472A1 (zh) 控制无人机云台转动的智能显示设备及其控制系统
WO2018010473A1 (zh) 基于智能显示设备的无人机云台转动控制方法
TWI682878B (zh) 使用無人機的旅遊系統與方法
CN111045209A (zh) 使用无人机的旅游系统与方法
WO2022061615A1 (zh) 待跟随目标的确定方法、装置、系统、设备及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20954781

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20954781

Country of ref document: EP

Kind code of ref document: A1