A multi-camera intelligent adaptive shooting method and system for embedded devices

Abstract

This application discloses a multi-camera intelligent adaptive shooting method and system for embedded devices, relating to the field of embedded vision acquisition technology. The method includes: scanning camera devices connected to the embedded device, acquiring device feature information, and determining camera type information; determining at least one target camera device in response to a target shooting task; acquiring preview images and determining environmental feature information characterizing the current shooting environment's illumination state; determining a target shooting strategy, including camera control parameters and image processing methods, based on the environmental feature information; acquiring target image data according to the camera control parameters and generating shooting results according to the image processing methods; and, when preset low-light conditions are met, increasing exposure correlation parameters and performing multi-frame fusion noise reduction and local contrast enhancement. This enables automatic camera device matching, adaptive shooting strategy determination, and low-light image quality improvement, enhancing imaging stability in complex environments.

Description

Technical Field

[0001] This application relates to the field of embedded vision acquisition technology, and in particular to a multi-camera intelligent adaptive shooting method and system for embedded devices. Background Technology

[0002] With the development of the Internet of Things, smart manufacturing, mobile robots, and intelligent sensing technologies, an increasing number of embedded devices are being equipped with camera modules to achieve visual functions such as image acquisition, video recording, environmental observation, and target recognition. These embedded devices typically operate in complex indoor and outdoor environments, and their application scenarios may involve tasks such as inspection, security, access control, navigation, and detection. Therefore, high requirements are placed on the stability of image acquisition, device compatibility, and system operational reliability.

[0003] Existing embedded camera solutions typically perform basic photo or video recording functions. However, in practical use, the image acquisition effect and device management efficiency remain insufficient due to differences in camera hardware specifications, operating environment, system resources, and application tasks. For example, when external lighting conditions vary significantly, acquired images are prone to uneven brightness, loss of detail, or excessive noise, affecting the accuracy of subsequent image recognition and visual analysis. When multiple different types of cameras are connected to an embedded device simultaneously, differences in interface types, device information, and operating status between the cameras can increase the complexity of system configuration, device maintenance, and troubleshooting. Furthermore, some existing shooting processes are relatively fixed, making it difficult to adapt the image acquisition process to different shooting tasks and operating environments, thus limiting the effectiveness of embedded vision systems in complex scenarios.

[0004] Therefore, in the visual acquisition process of embedded devices, the insufficiency of image quality stability in complex environments, the complexity of managing multiple camera devices, and the poor adaptability of image acquisition processes have become urgent problems to be solved. Summary of the Invention

[0005] This application provides a multi-camera intelligent adaptive shooting method and system for embedded devices, aiming to solve the problems of insufficient image quality stability in complex environments, complex management of multi-camera devices, and poor adaptability of image acquisition processes in the visual acquisition process of embedded devices in the prior art.

[0006] Firstly, a multi-camera intelligent adaptive shooting method for embedded devices is provided, applied to an embedded device connected to multiple camera devices, the method comprising: Scan the camera devices connected to the embedded device, obtain the device feature information corresponding to each camera device, and determine the camera type information corresponding to each camera device based on the device feature information; wherein, the device feature information includes device node information, device capability information and / or device name information; In response to a target shooting task, at least one target camera device is determined from a plurality of camera devices based on the target shooting task and the camera type information; The target camera device acquires preview images, and environmental feature information that characterizes the current shooting environment's lighting conditions is determined based on the preview images; Based on the environmental feature information, a target shooting strategy corresponding to the target shooting task is determined, and the target shooting strategy includes camera control parameters and image processing methods; The target camera device is controlled to acquire target image data according to the camera control parameters. The target image data is processed according to the image processing method described above to generate the shooting result corresponding to the target shooting task; Wherein, when the environmental feature information indicates that the current shooting environment meets the preset low light conditions, the target shooting strategy includes improving the exposure correlation parameters of the target camera device, and performing multi-frame fusion noise reduction processing and local contrast enhancement processing on the target image data.

[0007] Optionally, in the above scheme, scanning the camera devices connected to the embedded device, obtaining device feature information corresponding to each camera device, and determining the camera type information corresponding to each camera device based on the device feature information includes: Scan the camera device nodes of the embedded device to obtain a set of candidate device nodes; Perform a device capability query on the candidate device nodes in the candidate device node set to obtain the device capability information and device name information corresponding to each candidate device node; The device capability information and device name information are matched with the preset device configuration relationship to obtain the camera type information corresponding to each candidate device node; A list of camera devices is generated based on each candidate device node and the camera type information corresponding to each candidate device node.

[0008] Optionally, in the above scheme, after generating the camera device list, the method further includes: Based on the camera device list, determine the camera devices whose metadata needs to be obtained; Invoke the device information acquisition instruction corresponding to the camera device whose metadata is to be acquired, and obtain the device metadata corresponding to the camera device whose metadata is to be acquired; The device metadata is structured to obtain device management information; The camera device list is updated based on the device management information so that the updated camera device list includes camera type information and device metadata.

[0009] Optionally, in response to a target shooting task, determining at least one target camera device from a plurality of camera devices based on the target shooting task and the camera type information includes: The target shooting task is analyzed to obtain the task type information and task requirement information corresponding to the target shooting task; wherein, the task type information includes image shooting task, video recording task and / or video compositing task; Based on the task type information and the task requirement information, determine the target camera type corresponding to the target shooting task; The target camera type is matched with the camera type information corresponding to each camera device to obtain at least one candidate camera device; Based on the operating status information of the candidate camera devices, at least one target camera device is determined from the at least one candidate camera device.

[0010] Optionally, in the above scheme, a preview image is acquired through the target camera device, and environmental feature information characterizing the current shooting environment's lighting state is determined based on the preview image, including: Multiple preview images are continuously captured by the target camera device to obtain a preview image sequence; Brightness statistics are performed on each frame of the preview image sequence to obtain the image brightness information corresponding to each frame of the preview image; Based on the image brightness information corresponding to each frame of the preview image, determine the ambient brightness information corresponding to the current shooting environment; The environmental feature information is determined based on the ambient brightness information.

[0011] Optionally, in the above scheme, determining the target shooting strategy corresponding to the target shooting task based on the environmental feature information includes: The environmental feature information is compared with the preset illumination judgment conditions to obtain the illumination judgment result; If the illumination judgment result indicates that the current shooting environment does not meet the preset low illumination conditions, the target shooting strategy corresponding to the target shooting task is determined to be the normal light shooting strategy. If the illumination judgment result indicates that the current shooting environment meets the preset low illumination conditions, the target shooting strategy corresponding to the target shooting task is determined to be the low light enhancement shooting strategy. Based on the normal light shooting strategy or the low light enhancement shooting strategy, determine the camera control parameters and image processing method corresponding to the target shooting task.

[0012] Optionally, in the above scheme, the camera control parameters and image processing method corresponding to the target shooting task are determined according to the normal light shooting strategy or the low light enhancement shooting strategy, including: When the target shooting strategy is a normal light shooting strategy, the camera control parameters are determined to include automatic exposure parameters and automatic white balance parameters, and the image processing method is determined to include conventional image output processing method. When the target shooting strategy is a low-light enhancement shooting strategy, the camera control parameters are determined to include manual exposure parameters, gain parameters, and basic image adjustment parameters, and the image processing methods are determined to include multi-frame fusion noise reduction processing and local contrast enhancement processing. Based on at least some of the parameters selected from the automatic exposure parameters, the automatic white balance parameters, the manual exposure parameters, the gain parameters, and the image basic adjustment parameters, a parameter configuration instruction for controlling the target camera device is generated.

[0013] Optionally, in the above scheme, the target image data is processed according to the image processing method to generate the shooting result corresponding to the target shooting task, including: When the image processing method includes a multi-frame fusion noise reduction processing method, the continuously acquired multi-frame images to be fused are determined from the target image data; The multiple frames of images to be fused are subjected to temporal fusion processing to obtain a denoised image; Generate intermediate image results corresponding to the target shooting task based on the denoised image; The intermediate image results are encoded according to the output format corresponding to the target shooting task to generate the shooting result corresponding to the target shooting task.

[0014] Optionally, in the above scheme, generating intermediate image results corresponding to the target shooting task based on the denoised image includes: The luminance channel of the denoised image is extracted to obtain the luminance channel to be enhanced. The brightness channel to be enhanced is subjected to local contrast enhancement processing to obtain the enhanced brightness channel; An enhanced image is generated based on the enhanced luminance channel; The enhanced image is determined as the intermediate image result corresponding to the target shooting task.

[0015] Secondly, an intelligent adaptive shooting system for multiple cameras in an embedded device is provided, applied to an embedded device connected to multiple camera devices, the system comprising: The device identification and management module is used to scan the camera devices connected to the embedded device, obtain the device feature information corresponding to each camera device, and determine the camera type information corresponding to each camera device based on the device feature information; wherein, the device feature information includes device node information, device capability information and / or device name information; The task parsing module is used to, in response to a target shooting task, determine at least one target camera device from a plurality of camera devices based on the target shooting task and the camera type information; An environmental perception module is used to acquire preview images through the target camera device and determine environmental feature information that characterizes the current shooting environment's lighting status based on the preview images. The strategy determination module is used to determine the target shooting strategy corresponding to the target shooting task based on the environmental feature information. The target shooting strategy includes camera control parameters and image processing methods. The image acquisition module is used to control the target camera device to acquire target image data according to the camera control parameters; The image processing module is used to process the target image data according to the image processing method to generate the shooting result corresponding to the target shooting task; Wherein, when the environmental feature information indicates that the current shooting environment meets the preset low light conditions, the target shooting strategy includes improving the exposure correlation parameters of the target camera device, and performing multi-frame fusion noise reduction processing and local contrast enhancement processing on the target image data.

[0016] Compared with the prior art, this application has at least the following beneficial effects: Based on further analysis and research of existing technical problems, this application recognizes that in the visual acquisition process of embedded devices, there are issues such as insufficient image quality stability in complex environments, complex management of multiple camera devices, and poor adaptability of image acquisition processes. By scanning the camera devices connected to the embedded device and determining the camera type information corresponding to each camera device based on device feature information, the system can clearly identify the type and available objects of different camera devices before executing the shooting task. This allows the system to select at least one matching target camera device for the target shooting task, reducing the management complexity caused by manual configuration of camera devices. After determining the target camera device, a preview image is acquired through that target camera device, and a table is determined based on the preview image. By acquiring environmental feature information about the current shooting environment's lighting conditions, the system can obtain the current lighting status before formally acquiring target image data. Furthermore, based on this environmental feature information, it determines a target shooting strategy, including camera control parameters and image processing methods, so that the shooting process no longer executes the same acquisition procedure but can adaptively adjust to the current shooting environment. When the current shooting environment meets preset low-light conditions, it enhances the basic imaging brightness of the target image data by increasing the exposure correlation parameters of the target camera device, reduces random noise easily generated under low light through multi-frame fusion noise reduction processing, and improves the local detail of the image through local contrast enhancement processing, thereby generating the shooting result corresponding to the target shooting task. Therefore, this application can achieve automatic matching of target camera devices, adaptive determination of shooting strategies, and synergistic improvement of low-light image quality in multi-camera embedded devices, thereby solving the problems of insufficient image quality stability in complex environments, complex management of multi-camera devices, and poor adaptability of image acquisition processes. Attached Figure Description

[0017] Figure 1 A flowchart illustrating a multi-camera intelligent adaptive shooting method for an embedded device provided in one embodiment of this application; Figure 2 This is a flowchart of the main program for a multi-camera intelligent adaptive shooting method for an embedded device provided in one embodiment of this application; Figure 3 This is a flowchart illustrating an embodiment of the intelligent photo-taking process provided in this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0019] like Figure 2As shown, this method can run on an embedded Linux system of an embedded device, which can be a robot, drone, smart access control device, industrial inspection equipment, or other intelligent terminal device that requires image acquisition, video recording, and environmental observation. The main program can receive target shooting tasks and enter the corresponding task flow based on these tasks. Target shooting tasks can include image capture, video recording, and video synthesis tasks, as well as auxiliary tasks such as device information query and device diagnostics. Through this main program flow, camera device identification, camera device management, shooting task execution, image processing, and result output can be integrated into a unified control flow.

[0020] In this embodiment, as Figure 1 and Figure 3 As shown, a multi-camera intelligent adaptive shooting method for embedded devices is provided, in which the embedded device connects to multiple camera devices. These multiple camera devices can include one or more of the following: ordinary RGB cameras, wide-angle cameras, depth cameras, 3D cameras, and thermal imaging cameras. Before performing a target shooting task, the embedded device can first scan the connected camera devices, obtain the device characteristic information corresponding to each camera device, and determine the camera type information corresponding to each camera device based on the device characteristic information. The device characteristic information includes device node information, device capability information, and / or device name information. Device node information can be used to characterize the access path of the camera device in the embedded Linux system; device capability information can be used to characterize the image acquisition format, resolution, frame rate, input / output capabilities, etc., supported by the camera device; and device name information can be used to characterize the hardware name or driver identification name of the camera device. Camera type information can be used to characterize whether the camera device is a wide-angle camera, depth camera, 3D camera, thermal imaging camera, or other type of camera.

[0021] After determining the camera type information, the embedded device can respond to the target shooting task and, based on the target shooting task and camera type information, determine at least one target camera device from multiple camera devices. For example, when the target shooting task is a general image shooting task, an RGB camera or a wide-angle camera can be selected as the target camera device; when the target shooting task requires acquiring scene depth information, a depth camera or a 3D camera can be selected as the target camera device; when the target shooting task requires acquiring temperature distribution information, a thermal imaging camera can be selected as the target camera device; when the target shooting task is a video compositing task, multiple camera devices can be selected simultaneously as target camera devices.

[0022] After identifying the target camera device, the embedded device can acquire preview images using the target camera device and determine environmental feature information to characterize the current shooting environment's lighting conditions based on the preview images. Specifically, before formally acquiring target image data, one or more preview images can be acquired using the target camera device, and the brightness information of the preview images can be statistically analyzed to obtain the current shooting environment's lighting conditions. Environmental feature information can include ambient brightness values, normalized light levels, low-light judgment indicators, or other information that can characterize the current shooting environment's lighting conditions. For example, the preview image can be converted to a grayscale image or its brightness channel can be extracted, and the ambient brightness information can be determined based on the average pixel brightness. This ambient brightness information can then be mapped to a lighting level within a preset range.

[0023] After obtaining environmental feature information, the embedded device can determine the target shooting strategy corresponding to the target shooting task based on the environmental feature information. The target shooting strategy includes camera control parameters and image processing methods. Camera control parameters may include one or more of the following: exposure-related parameters, gain parameters, white balance parameters, brightness parameters, contrast parameters, saturation parameters, resolution parameters, and frame rate parameters. Image processing methods may include one or more of the following: conventional image output processing methods, multi-frame fusion noise reduction processing methods, local contrast enhancement processing methods, and encoding compression processing methods. By using both camera control parameters and image processing methods as the target shooting strategy, the embedded device can coordinate between the acquisition and processing ends, rather than relying solely on the camera's default parameters to complete the shooting.

[0024] After determining the target shooting strategy, the embedded device can control the target camera device to acquire target image data according to the camera control parameters. For image capture tasks, the target image data can be single-frame image data or continuously acquired multi-frame image data; for video recording tasks, the target image data can be a sequence of video frames continuously acquired at a preset frame rate; for video compositing tasks, the target image data can be a sequence of image frames from one or more target camera devices. After acquisition, the embedded device can process the target image data according to image processing methods to generate the shooting result corresponding to the target shooting task. The shooting result can be an image file, a video file, a composite video file, or a structured file containing device information.

[0025] Given that the environmental characteristics indicate the current shooting environment meets preset low-light conditions, the target shooting strategy includes improving the exposure-related parameters of the target camera device and performing multi-frame fusion noise reduction and local contrast enhancement on the target image data. Exposure-related parameters may include exposure time, exposure mode, gain parameters, or other camera acquisition parameters related to image brightness enhancement. Multi-frame fusion noise reduction reduces random noise by continuously acquiring multiple frames and performing temporal fusion; local contrast enhancement enhances the brightness information of the image to improve shadow details. After processing, the image or video can be encoded and compressed according to the output requirements of the target shooting task, for example, generating JPEG image or video files, and saved to a specified storage path.

[0026] Through the above implementation method, multiple camera devices are first identified so that the target shooting task can be matched with a suitable target camera device; then, the current lighting conditions of the shooting environment are determined by previewing the image, enabling the shooting process to adaptively determine the target shooting strategy according to environmental changes; finally, in low-light conditions, the target image data is improved through exposure-related parameter adjustment, multi-frame fusion noise reduction processing, and local contrast enhancement processing. Therefore, the method can reduce the complexity of managing multiple camera devices, improve the adaptability of the image acquisition process to complex environments, and improve the stability of imaging quality in low-light scenes.

[0027] In some embodiments, scanning the camera devices connected to the embedded device, obtaining device feature information corresponding to each camera device, and determining the camera type information corresponding to each camera device based on the device feature information may include the following steps.

[0028] First, the embedded device can scan for camera device nodes to obtain a set of candidate device nodes. In an embedded Linux system, camera device nodes can be represented as / dev / video0, / dev / video1, / dev / video2, etc. The embedded device can traverse the device nodes in the / dev directory that meet the preset naming rules and add the scanned device nodes to the candidate device node set. The candidate device node set represents the device nodes in the current system that may be related to the camera device.

[0029] Then, the embedded device can perform device capability queries on the candidate device nodes in the candidate device node set to obtain the device capability information and device name information corresponding to each candidate device node. Specifically, the device capability information and device name information of the candidate device nodes can be obtained through the video device capability query interface. For example, the V4L2 capability interface can be used to query the video capture capabilities, device driver information, device name field, and other information supported by the candidate device node. In a specific example, the device capability structure can be obtained through the VIDIOC_QUERYCAP interface, and the device name field can be extracted from the device capability structure to obtain the device name information.

[0030] Subsequently, the embedded device can match the device capability information and device name information with the preset device configuration relationship to obtain the camera type information corresponding to each candidate device node. The preset device configuration relationship can be stored in the form of a configuration file, such as a JSON configuration file recording the correspondence between different device names, device capability characteristics, supplier identifiers, product identifiers, and camera types. By matching the queried device capability information and device name information with the preset device configuration relationship, the camera type information corresponding to the candidate device node can be automatically identified.

[0031] Finally, the embedded device can generate a camera device list based on each candidate device node and the corresponding camera type information. The camera device list may include one or more of the following: device node information, camera type information, device name information, device capability information, and device status information. This camera device list can serve as the data basis for selecting target camera devices in subsequent target acquisition tasks.

[0032] Through the above implementation methods, this approach can automatically discover multiple camera devices connected to an embedded device and determine the camera type information based on device capability information and device name information, eliminating the need for manual configuration of each device node and camera type. Therefore, this method can improve the recognition efficiency of heterogeneous camera devices and reduce the complexity of deploying and maintaining multi-camera systems.

[0033] In some embodiments, after generating the camera device list, the embedded device can also obtain the device metadata of the camera devices based on the camera device list and update the camera device list using the device metadata.

[0034] Specifically, the embedded device can determine the camera devices whose metadata needs to be acquired based on the camera device list. The camera devices whose metadata needs to be acquired can be all camera devices in the list, the target camera device to be used in the target shooting task, or a specific camera device requiring device diagnostics or maintenance. After determining the camera devices whose metadata needs to be acquired, the corresponding device information acquisition method can be determined based on the camera type information and device node information.

[0035] Then, the embedded device can invoke the device information retrieval command corresponding to the camera device whose metadata is to be retrieved, and obtain the device metadata corresponding to the camera device whose metadata is to be retrieved. For ordinary video cameras, information such as supported formats, resolutions, frame rates, and control parameter ranges can be obtained through commands such as v4l2-ctl; for Orbbec 3D cameras, the device serial number and device status information can be obtained through the corresponding device query command; for Intel RealSense depth cameras, the firmware version information and device connection information can be obtained through the corresponding firmware query command. Device metadata may include one or more of the following: device serial number, firmware version number, USB connection type, supported resolution, supported frame rate, current device status, and device control parameter range.

[0036] Next, the embedded device can perform structured processing on the device metadata to obtain device management information. This structured processing can include field extraction, standardized field naming, exception filtering, and data format conversion. Device management information can be stored in the form of JSON files, tabular files, or database records. For example, camera type information, device node information, device serial number, firmware version number, and USB connection type can be written into the same structured record to form device management information.

[0037] Finally, the embedded device can update the camera device list based on device management information, ensuring the updated list includes camera type information and device metadata. This updated list can be used for device information queries, device diagnostics, shooting task selection, and after-sales maintenance. When a camera device is replaced, its firmware version changes, or its connection method changes, the camera device list can be updated by re-acquiring the device metadata.

[0038] Through the above implementation methods, this method can not only identify camera types but also automatically acquire and structure and store the device metadata of camera devices, enabling device maintenance personnel or upper-layer applications to quickly obtain information such as device serial number, firmware version number, and connection status. Therefore, this method can improve the automation level of camera device management and enhance the efficiency of device replacement, fault diagnosis, and after-sales maintenance.

[0039] In some embodiments, in response to a target shooting task, determining at least one target camera device from a plurality of camera devices based on the target shooting task and camera type information may include the following steps.

[0040] First, the embedded device can parse the target capture task to obtain the corresponding task type information and task requirement information. The task type information includes image capture tasks, video recording tasks, and / or video compositing tasks. The task requirement information can include one or more of the following: target field of view requirements, target resolution requirements, target frame rate requirements, target imaging type requirements, target output format requirements, and target save path requirements. For example, an image capture task may correspond to a single image output requirement, a video recording task may correspond to a continuous video stream acquisition requirement, and a video compositing task may correspond to a composite output requirement from multiple video sources or multiple image sources.

[0041] Then, the embedded device can determine the target camera type corresponding to the target shooting task based on the task type information and task requirement information. For example, when the task requirement information indicates that a large-scale scene observation is needed, the target camera type can be determined to be a wide-angle camera; when the task requirement information indicates that spatial distance or depth information needs to be acquired, the target camera type can be determined to be a depth camera or a 3D camera; when the task requirement information indicates that heat distribution or temperature anomaly information needs to be acquired, the target camera type can be determined to be a thermal imaging camera; when the task type information is a video synthesis task, multiple target camera types can be determined according to the synthesis requirements.

[0042] Subsequently, the embedded device can match the target camera type with the camera type information corresponding to each camera device to obtain at least one candidate camera device. If multiple camera devices meet the target camera type, they can be further filtered based on device resolution, frame rate, connection status, historical usage status, or task priority.

[0043] Finally, the embedded device can determine at least one target camera device from at least one candidate camera device based on the operational status information of the candidate camera devices. Operational status information may include whether the device is online, whether it is in use, whether it has been successfully initialized, whether there are any abnormalities, and whether the device's current acquisition capabilities meet the target shooting task. For image capture or video recording tasks, one target camera device can be determined; for video compositing tasks, multiple target camera devices can be determined.

[0044] Through the above implementation methods, this method can automatically select the target camera device according to the type and requirements of the target shooting task, enabling tasks such as taking pictures, recording videos, and video compositing to be executed under a unified control framework. Therefore, this method can reduce the awareness of differences between upper-layer applications and lower-layer camera devices, and improve the automation and adaptability of multi-task image acquisition processes.

[0045] In some embodiments, acquiring a preview image through a target camera device and determining environmental feature information to characterize the current shooting environment's lighting conditions based on the preview image may include the following steps.

[0046] First, the embedded device can continuously acquire multiple preview images using the target camera, obtaining a preview image sequence. This preview image sequence can be acquired before the actual acquisition of the target image data. Continuously acquiring multiple preview images reduces the probability of a single frame being affected by accidental occlusion, instantaneous brightness changes, or acquisition noise, making the determination of ambient lighting conditions more stable. Images in the preview image sequence can be acquired at lower resolution or with shorter acquisition times to reduce the impact of ambient lighting condition detection on system resources and shooting latency.

[0047] Then, the embedded device can perform brightness statistics on each frame of the preview image sequence to obtain the image brightness information corresponding to each frame. Brightness statistics can include converting the preview image to grayscale and calculating the average pixel brightness, or extracting brightness channels and calculating the average brightness, median brightness, or brightness distribution information. The image brightness information can be used to characterize the overall brightness level of a single frame of the preview image.

[0048] Next, the embedded device can determine the ambient brightness information corresponding to the current shooting environment based on the image brightness information corresponding to each frame of the preview image. For example, the ambient brightness information can be obtained by averaging the brightness information of multiple frames; or abnormal brightness values ​​can be removed before statistical analysis to obtain more stable ambient brightness information. In one example, the ambient brightness information can be normalized to a brightness level within a preset range, such as a light level between 0 and 100.

[0049] Finally, the embedded device can determine environmental feature information based on ambient brightness information. This environmental feature information can directly include ambient brightness information, or it can include illumination level information, low-light judgment information, or shooting environment classification information generated based on the ambient brightness information. This environmental feature information can be used in subsequent steps to determine the target shooting strategy.

[0050] Through the above implementation method, the target camera device acquires preview images and performs brightness statistics, enabling the embedded device to perceive the current shooting environment's lighting conditions before the actual shooting.

[0051] In some embodiments, determining the target shooting strategy corresponding to the target shooting task based on environmental feature information may include the following steps.

[0052] First, the embedded device can compare environmental feature information with preset lighting conditions to obtain a lighting judgment result. Preset lighting conditions may include a preset low-light threshold, a preset normal lighting range, or other conditions used to distinguish lighting states. For example, when the environmental feature information includes a normalized light level, this normalized light level can be compared with a preset low-light threshold; when the normalized light level is lower than the preset low-light threshold, a lighting judgment result indicating that the current shooting environment meets the preset low-light conditions can be obtained.

[0053] If the lighting assessment indicates that the current shooting environment does not meet the preset low-light conditions, the embedded device can determine that the target shooting strategy for the target shooting task is the normal light shooting strategy. The normal light shooting strategy is suitable for scenes with sufficient light or relatively stable lighting conditions. Under the normal light shooting strategy, the factory-optimized parameters, automatic exposure parameters, and automatic white balance parameters of the target camera device can be used preferentially to obtain images or videos with natural colors, moderate brightness, and relatively complete details.

[0054] If the lighting assessment indicates that the current shooting environment meets the preset low-light conditions, the embedded device can determine that the target shooting strategy for the target shooting task is a low-light enhancement shooting strategy. The low-light enhancement shooting strategy is suitable for scenes with low light, insufficient local lighting, or low overall image brightness. Under the low-light enhancement shooting strategy, the basic imaging brightness can be improved by adjusting camera control parameters, and noise and shadow details can be improved through image processing.

[0055] After determining the normal light shooting strategy or the low light enhancement shooting strategy, the embedded device can determine the camera control parameters and image processing method corresponding to the target shooting task based on the normal light shooting strategy or the low light enhancement shooting strategy. Thus, environmental feature information can be converted into executable parameter configurations and image processing procedures.

[0056] Through the above implementation method, it is possible to adaptively select between normal light shooting strategy and low light enhancement shooting strategy based on environmental feature information, so that embedded devices no longer handle all environments with a fixed shooting process.

[0057] In some embodiments, determining the camera control parameters and image processing method corresponding to the target shooting task based on a normal light shooting strategy or a low light enhancement shooting strategy may include the following steps.

[0058] When the target shooting strategy is a normal light shooting strategy, the embedded device can determine the camera control parameters, including automatic exposure parameters and automatic white balance parameters, and determine the image processing method, including conventional image output processing. Automatic exposure parameters enable the target camera to automatically adjust the exposure state according to the current scene brightness, while automatic white balance parameters enable the target camera to automatically adjust the color balance according to the current light source conditions. Conventional image output processing can include image format conversion, encoding compression, quality parameter settings, and file saving. In well-lit conditions, using automatic exposure and automatic white balance parameters can reduce unnecessary additional enhancement processing, maintaining the naturalness of image color and detail.

[0059] When the target shooting strategy is a low-light enhancement shooting strategy, the embedded device can determine the camera control parameters, including manual exposure parameters, gain parameters, and basic image adjustment parameters, and determine the image processing methods, including multi-frame fusion noise reduction and local contrast enhancement. Manual exposure parameters are used to switch the target camera to manual exposure mode and set the exposure time; gain parameters are used to increase the amplification of the image signal; basic image adjustment parameters can include one or more of brightness, contrast, and saturation parameters. These parameters can improve the basic brightness and usable information content of the low-light image during the acquisition phase. Subsequently, multi-frame fusion noise reduction is used to suppress random noise, and local contrast enhancement improves shadow details.

[0060] After obtaining at least some of the parameters from automatic exposure, automatic white balance, manual exposure, gain, and basic image adjustment parameters, the embedded device can generate parameter configuration instructions for controlling the target camera device. These instructions can be sent to the target camera device via its corresponding driver interface, control interface, or system commands. For example, parameters such as exposure mode, exposure time, gain, brightness, contrast, and saturation can be set through the camera control interface. Once the parameter configuration instructions are executed, the target camera device can acquire target image data according to the configured parameters.

[0061] Through the above implementation methods, the normal light shooting strategy and the low light enhanced shooting strategy are configured with different camera control parameters and image processing methods, enabling the target camera device to adopt different acquisition control methods under different lighting conditions. Therefore, this method can ensure natural imaging effects under normal lighting conditions, and improve image clarity and detail under low light conditions through acquisition parameter adjustment and image enhancement processing.

[0062] In some embodiments, processing the target image data according to an image processing method to generate the shooting result corresponding to the target shooting task may include the following steps.

[0063] When image processing includes multi-frame fusion noise reduction, the embedded device can determine multiple consecutively acquired images to be fused from the target image data. These multiple images can be consecutively acquired under a low-light enhancement shooting strategy, or consecutive frames from a video frame sequence. Since increasing exposure-related parameters and gain parameters under low-light conditions may increase random noise, temporal fusion of multiple images can be performed to reduce the impact of random noise on the shooting results.

[0064] Then, the embedded device can perform temporal fusion processing on multiple frames of images to be fused to obtain a denoised image. Temporal fusion processing may include averaging the pixel values ​​at corresponding pixel positions in the multiple frames of images to be fused, or it may include weighted fusion processing after determining weights based on image brightness, image sharpness, or frame quality. Through temporal fusion processing, random noise is weakened during the multi-frame statistical process, while relatively stable scene content is preserved, thus obtaining a denoised image with lower noise.

[0065] Next, the embedded device can generate intermediate image results corresponding to the target shooting task based on the denoised image. The intermediate image results can be images directly formed from the denoised image, or they can be input images for subsequent local contrast enhancement processing. If the target shooting task only requires multi-frame fusion denoising processing, the denoised image can be used as the intermediate image result; if the target shooting task also requires local contrast enhancement processing, the denoised image can be input into the subsequent enhancement processing flow.

[0066] Finally, the embedded device can encode the intermediate image results according to the output format corresponding to the target shooting task to generate the shooting result corresponding to the target shooting task. The output format can be JPEG image format, PNG image format, video encoding format, or other output formats suitable for embedded devices. For image shooting tasks, the intermediate image results can be JPEG compressed and saved according to preset image quality parameters; for video recording tasks, the processed image frames can be encoded into video files; for video compositing tasks, image data from multiple target camera devices can be processed and combined into the target video result.

[0067] By employing the above-described implementation method and performing temporal fusion processing on multiple consecutively acquired images, random noise caused by increased gain or prolonged exposure can be suppressed under low-light conditions. Therefore, this method can improve the clarity and usability of low-light shooting results and address the problem of insufficient image quality stability in complex environments.

[0068] In some embodiments, generating intermediate image results corresponding to the target shooting task based on the denoised image may include the following steps.

[0069] First, the embedded device can perform luminance channel extraction processing on the denoised image to obtain the luminance channel to be enhanced. Luminance channel extraction processing can include converting the denoised image to a color space containing luminance information and extracting the luminance channel therein; it can also include calculating the luminance channel to be enhanced based on the color components of the denoised image. The luminance channel to be enhanced is used to characterize the brightness distribution of different regions in the denoised image.

[0070] Then, the embedded device can perform local contrast enhancement processing on the luminance channel to be enhanced, resulting in an enhanced luminance channel. Local contrast enhancement processing can employ contrast-limited adaptive histogram equalization. Specifically, the luminance channel to be enhanced can be divided into multiple local regions, and histogram statistics and equalization processing can be performed on each local region separately. Noise amplification caused by over-enhancement is suppressed by limiting contrast. This processing can improve visible details in dark areas and avoid excessive stretching of local contrast in the image.

[0071] Next, the embedded device can generate an enhanced image based on the enhanced luminance channel. Specifically, the enhanced luminance channel can be used to replace or update the luminance information in the original denoised image, and combined with other image information from the original denoised image to generate the enhanced image. The enhanced image has more obvious dark area gradation and better local detail representation compared to the denoised image.

[0072] Finally, the embedded device can identify the enhanced image as the intermediate image result corresponding to the target shooting task. This intermediate image result can then proceed to subsequent processing steps such as encoding compression, quality adjustment, file saving, or video compositing. For example, in the intelligent shooting process, the enhanced image can be compressed using JPEG and saved as an image file according to the target save path.

[0073] Through the above implementation method, after multi-frame fusion and noise reduction processing, local contrast enhancement processing is further performed on the luminance channel to improve the details in dark areas of the low-light image. At the same time, by limiting the contrast enhancement, the risk of noise being excessively amplified is reduced. Therefore, this method can balance noise reduction and detail enhancement in low-light environments, improving the visual quality of the shooting results and the usability of subsequent image analysis.

[0074] In some embodiments, the multi-camera intelligent adaptive shooting system for embedded devices can be applied to embedded devices that connect to multiple camera devices. This system can be implemented through software modules, hardware modules, or a combination of software and hardware, and can execute... Figure 2 The main program flow shown and Figure 3The intelligent photo-taking process is shown below. The system includes a device identification and management module, a task analysis module, an environmental perception module, a strategy determination module, an image acquisition module, and an image processing module.

[0075] The device identification and management module scans the camera devices connected to the embedded device, obtains the device characteristic information corresponding to each camera device, and determines the camera type information corresponding to each camera device based on the device characteristic information. The device characteristic information includes device node information, device capability information, and / or device name information. The device identification and management module can scan device nodes in the form of / dev / video*, obtain device capability information and device name information through the device capability query interface, and match the device capability information and device name information with preset device configuration relationships to determine the camera type information. The device identification and management module can also obtain device metadata such as device serial number, firmware version number, and USB connection type, and store the device metadata in a structured manner as device management information.

[0076] The task parsing module, in response to a target shooting task, determines at least one target camera device from multiple camera devices based on the target shooting task and camera type information. The task parsing module can parse image shooting tasks, video recording tasks, video compositing tasks, device information query tasks, or device diagnostic tasks, and select the target camera device based on task type information, task requirement information, and camera type information. When the target shooting task is a video compositing task, the task parsing module can determine multiple target camera devices.

[0077] The environment perception module is used to acquire preview images through the target camera device and determine environmental feature information to characterize the current shooting environment's lighting conditions based on the preview images. The environment perception module can continuously acquire multiple frames of preview images, perform brightness statistics on each frame, and determine environmental brightness information and environmental feature information based on the brightness information of multiple frames.

[0078] The strategy determination module is used to determine the target shooting strategy corresponding to the target shooting task based on environmental feature information. The target shooting strategy includes camera control parameters and image processing methods. The strategy determination module can compare environmental feature information with preset lighting conditions. When the current shooting environment does not meet the preset low-light conditions, a normal light shooting strategy is determined; when the current shooting environment meets the preset low-light conditions, a low-light enhancement shooting strategy is determined. The normal light shooting strategy may include automatic exposure parameters, automatic white balance parameters, and conventional image output processing methods; the low-light enhancement shooting strategy may include manual exposure parameters, gain parameters, basic image adjustment parameters, multi-frame fusion noise reduction processing methods, and local contrast enhancement processing methods.

[0079] The image acquisition module is used to control the target camera device to acquire target image data according to the camera control parameters. The image acquisition module can set parameters such as exposure mode, exposure time, gain, brightness, contrast, saturation, resolution, and frame rate for the target camera device according to the parameter configuration instructions generated by the strategy determination module, and acquire single-frame images, multi-frame images, or video frame sequences.

[0080] The image processing module processes the target image data according to image processing methods to generate the shooting results corresponding to the target shooting task. When the environmental feature information indicates that the current shooting environment meets preset low-light conditions, the target shooting strategy includes improving the exposure correlation parameters of the target camera device and performing multi-frame fusion noise reduction and local contrast enhancement processing on the target image data. The image processing module can perform temporal fusion processing on continuously acquired multiple frames of images to be fused to obtain a denoised image; perform luminance channel extraction processing and local contrast enhancement processing on the denoised image to obtain an enhanced image; and encode the enhanced image according to the output format corresponding to the target shooting task to generate image files, video files, or composite video files.

[0081] Through the above implementation, the system achieves automatic identification and management of multiple camera devices through a device identification and management module, matches target shooting tasks with target camera devices through a task parsing module, adaptively determines shooting strategies through an environment perception module and a strategy determination module, and acquires target image data, performs noise reduction and enhancement, and outputs shooting results through an image acquisition module and an image processing module. Therefore, the system can form a unified multi-camera intelligent adaptive shooting framework in embedded devices, reducing the complexity of managing heterogeneous cameras, improving imaging quality in complex lighting environments, and enhancing the integration efficiency of multi-task processing such as taking photos, recording videos, and video synthesis.

[0082] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A multi-camera intelligent adaptive shooting method of an embedded device, characterized in that, An embedded device for connecting multiple camera devices, the method includes: Scan the camera devices connected to the embedded device, obtain the device feature information corresponding to each camera device, and determine the camera type information corresponding to each camera device based on the device feature information; wherein, the device feature information includes device node information, device capability information and / or device name information; In response to a target shooting task, at least one target camera device is determined from a plurality of camera devices based on the target shooting task and the camera type information; The target camera device acquires preview images, and environmental feature information that characterizes the current shooting environment's lighting conditions is determined based on the preview images; Based on the environmental feature information, a target shooting strategy corresponding to the target shooting task is determined, and the target shooting strategy includes camera control parameters and image processing methods; The target camera device is controlled to acquire target image data according to the camera control parameters. The target image data is processed according to the image processing method described above to generate the shooting result corresponding to the target shooting task; Wherein, when the environmental feature information indicates that the current shooting environment meets the preset low light conditions, the target shooting strategy includes improving the exposure correlation parameters of the target camera device, and performing multi-frame fusion noise reduction processing and local contrast enhancement processing on the target image data.

2. The multi-camera intelligent adaptive shooting method of the embedded device according to claim 1, characterized in that, Scanning the camera devices connected to the embedded device, obtaining device feature information corresponding to each camera device, and determining the camera type information corresponding to each camera device based on the device feature information, including: Scan the camera device nodes of the embedded device to obtain a set of candidate device nodes; Perform a device capability query on the candidate device nodes in the candidate device node set to obtain the device capability information and device name information corresponding to each candidate device node; The device capability information and device name information are matched with the preset device configuration relationship to obtain the camera type information corresponding to each candidate device node; A list of camera devices is generated based on each candidate device node and the camera type information corresponding to each candidate device node.

3. The multi-camera intelligent adaptive shooting method of the embedded device according to claim 2, characterized in that, After generating the list of camera devices, the method further includes: Based on the camera device list, determine the camera devices whose metadata needs to be obtained; Invoke the device information acquisition instruction corresponding to the camera device whose metadata is to be acquired, and obtain the device metadata corresponding to the camera device whose metadata is to be acquired; The device metadata is structured to obtain device management information; The camera device list is updated based on the device management information so that the updated camera device list includes camera type information and device metadata.

4. The multi-camera intelligent adaptive shooting method of the embedded device according to claim 1, characterized in that, In response to a target shooting task, determining at least one target camera device from a plurality of camera devices based on the target shooting task and the camera type information includes: The target shooting task is analyzed to obtain the task type information and task requirement information corresponding to the target shooting task; wherein, the task type information includes image shooting task, video recording task and / or video compositing task; Based on the task type information and the task requirement information, determine the target camera type corresponding to the target shooting task; The target camera type is matched with the camera type information corresponding to each camera device to obtain at least one candidate camera device; Based on the operating status information of the candidate camera devices, at least one target camera device is determined from the at least one candidate camera device.

5. The multi-camera intelligent adaptive shooting method for embedded devices according to claim 1, characterized in that, The target camera device acquires a preview image, and based on the preview image, determines environmental feature information to characterize the current shooting environment's lighting conditions, including: Multiple preview images are continuously captured by the target camera device to obtain a preview image sequence; Brightness statistics are performed on each frame of the preview image sequence to obtain the image brightness information corresponding to each frame of the preview image; Based on the image brightness information corresponding to each frame of the preview image, determine the ambient brightness information corresponding to the current shooting environment; The environmental feature information is determined based on the ambient brightness information.

6. The multi-camera intelligent adaptive shooting method of the embedded device according to claim 5, characterized in that, Determining the target shooting strategy corresponding to the target shooting task based on the environmental feature information includes: The environmental feature information is compared with the preset illumination judgment conditions to obtain the illumination judgment result; If the illumination judgment result indicates that the current shooting environment does not meet the preset low illumination conditions, the target shooting strategy corresponding to the target shooting task is determined to be the normal light shooting strategy. If the illumination judgment result indicates that the current shooting environment meets the preset low illumination conditions, the target shooting strategy corresponding to the target shooting task is determined to be the low light enhancement shooting strategy. Based on the normal light shooting strategy or the low light enhancement shooting strategy, determine the camera control parameters and image processing method corresponding to the target shooting task.

7. The multi-camera intelligent adaptive shooting method of embedded devices according to claim 6, characterized in that, Based on the normal light shooting strategy or the low light enhancement shooting strategy, determine the camera control parameters and image processing method corresponding to the target shooting task, including: When the target shooting strategy is a normal light shooting strategy, the camera control parameters are determined to include automatic exposure parameters and automatic white balance parameters, and the image processing method is determined to include conventional image output processing method. When the target shooting strategy is a low-light enhancement shooting strategy, the camera control parameters are determined to include manual exposure parameters, gain parameters, and basic image adjustment parameters, and the image processing methods are determined to include multi-frame fusion noise reduction processing and local contrast enhancement processing. Based on at least some of the parameters among the automatic exposure parameters, the automatic white balance parameters, the manual exposure parameters, the gain parameters, and the image basic adjustment parameters, a parameter configuration instruction for controlling the target camera device is generated.

8. The multi-camera intelligent adaptive shooting method of the embedded device according to claim 1, characterized in that, The target image data is processed according to the image processing method to generate the shooting result corresponding to the target shooting task, including: When the image processing method includes a multi-frame fusion noise reduction processing method, the continuously acquired multi-frame images to be fused are determined from the target image data; The multiple frames of images to be fused are subjected to temporal fusion processing to obtain a denoised image; Generate intermediate image results corresponding to the target shooting task based on the denoised image; The intermediate image results are encoded according to the output format corresponding to the target shooting task to generate the shooting result corresponding to the target shooting task.

9. The multi-camera intelligent adaptive shooting method of the embedded device according to claim 8, characterized in that, Generating intermediate image results corresponding to the target shooting task based on the denoised image includes: The luminance channel of the denoised image is extracted to obtain the luminance channel to be enhanced. The brightness channel to be enhanced is subjected to local contrast enhancement processing to obtain the enhanced brightness channel; An enhanced image is generated based on the enhanced luminance channel; The enhanced image is determined as the intermediate image result corresponding to the target shooting task.

10. A multi-camera intelligent adaptive shooting system of an embedded device, characterized in that, An embedded device for connecting multiple camera devices, the system comprising: The device identification and management module is used to scan the camera devices connected to the embedded device, obtain the device feature information corresponding to each camera device, and determine the camera type information corresponding to each camera device based on the device feature information; wherein, the device feature information includes device node information, device capability information and / or device name information; The task parsing module is used to, in response to a target shooting task, determine at least one target camera device from a plurality of camera devices based on the target shooting task and the camera type information; An environmental perception module is used to acquire preview images through the target camera device and determine environmental feature information that characterizes the current shooting environment's lighting status based on the preview images. The strategy determination module is used to determine the target shooting strategy corresponding to the target shooting task based on the environmental feature information. The target shooting strategy includes camera control parameters and image processing methods. The image acquisition module is used to control the target camera device to acquire target image data according to the camera control parameters; The image processing module is used to process the target image data according to the image processing method to generate the shooting result corresponding to the target shooting task; Wherein, when the environmental feature information indicates that the current shooting environment meets the preset low light conditions, the target shooting strategy includes improving the exposure correlation parameters of the target camera device, and performing multi-frame fusion noise reduction processing and local contrast enhancement processing on the target image data.

Images