An underwater scene AI adaptive multi-camera cooperative imaging system, method and mobile terminal

Abstract

The application discloses an underwater scene AI adaptive multi-camera cooperative imaging system and method and a mobile terminal, and relates to the technical field of mobile terminal imaging; the system comprises: a pressure sensing and sealing linkage module, which is used for dynamically adjusting the sealing strength of a lens according to water depth; a light refraction compensation module, which is used for correcting image blur caused by underwater refraction through an optical and digital combined mode; a multi-camera cooperative control module, which is used for intelligently switching at least macro, main camera and other lenses based on water depth to adapt to different shooting scenes; and an underwater color restoration module, which is used for inhibiting blue-green color deviation and restoring real colors through spectral compensation; the method comprises the following steps: water entry detection and sealing adjustment, light refraction compensation, intelligent lens selection and color restoration; through cooperation of multiple modules, the application solves the problems of low underwater sealing reliability, imaging blur, unintelligent lens switching and color distortion in the prior art, and realizes full-automatic and high-quality underwater shooting experience.

Description

Technical Field

[0001] This invention relates to the field of mobile terminal imaging technology, specifically to an adaptive imaging system, method, and mobile terminal suitable for underwater shooting scenarios, and particularly to an intelligent imaging scheme that integrates pressure sensing, light refraction compensation, multi-camera collaboration, and color optimization. Background Technology

[0002] With the rapid development and widespread adoption of waterproof technology for mobile devices (such as smartphones and action cameras), IP68 and above waterproof ratings have become common features in mid-to-high-end devices. This has driven the rapid expansion of underwater shooting scenarios from professional photography equipment to consumer-grade mobile devices. Users are increasingly demanding underwater imaging using mobile devices in their daily water activities (such as swimming and snorkeling), outdoor adventures (such as shallow-sea diving), and even light underwater operations (such as underwater inspections and aquaculture records). This demand is not only limited to the safety of the equipment for underwater use, but also places higher demands on the clarity, color accuracy, and adaptability of the imaging effect to different shooting scenarios. Therefore, consumer-grade underwater photography is becoming an important direction for the expansion of mobile device photography technology.

[0003] Currently, industry solutions for underwater imaging using mobile terminals mainly focus on the following three aspects: The first is passive protection based on hardware, which involves using fixed protective structures such as sealing rings, waterproof coatings, and nano-hydrophobic materials to give the equipment a certain underwater tolerance. This approach relies on the design and sealing level of the equipment before it leaves the factory and cannot be dynamically adjusted once determined.

[0004] Secondly, there is the stacking of hardware functions, namely, configuring multiple camera modules, such as macro lenses, main cameras, ultra-wide-angle lenses, telephoto lenses, etc., and using the inherent optical characteristics of different lenses (such as focal length and field of view) to cover different shooting distances from very close to far away; on land, such multi-camera systems can usually achieve automatic switching through simple distance detection or scene recognition.

[0005] Thirdly, there is the initial optimization of software algorithms, which involves improving the brightness and color perception of underwater images by increasing the overall exposure and making simple white balance adjustments in the camera software. Some models have launched "underwater mode", but these modes are mostly minor adjustments to the parameters of existing land imaging algorithms and have not been fundamentally redesigned for the physical characteristics of the underwater environment.

[0006] However, the aforementioned existing technologies have a series of inherent defects and shortcomings when dealing with complex and ever-changing underwater shooting environments: 1. Lack of dynamic adaptability in sealing protection: Existing fixed protective structures rely on the initial waterproof rating, and their sealing strength is constant. In water depths of 0-5m, water pressure is dynamic. Fixed sealing force may be "overkill" in shallow water, causing unnecessary energy consumption and structural fatigue. At depths close to the waterproof limit, there may be a risk of water infiltration due to insufficient sealing fit. The reliability of protection cannot be optimized with changes in the environment.

[0007] 2. Failure to address underwater light refraction interference: When light enters the air (lens protective glass) from the water, it undergoes significant refraction, causing a shift in the light propagation path. Existing technologies rely solely on the lens's optical design and autofocus algorithm, which is typically based on an air medium model. Underwater, this mismatch can lead to problems such as blurred edges, loss of detail, inaccurate focusing, and even image distortion, severely impacting image clarity.

[0008] 3. Multi-camera collaboration logic is not adapted to underwater scenarios: Existing multi-camera switching logic (such as laser or TOF ranging, AI scene recognition) is mainly designed for terrestrial environments; the medium (water) and light conditions in underwater environments are completely different from those on land, which may cause distance measurement and scene recognition algorithms that are effective on land to be inaccurate underwater; therefore, lens switching may become abrupt or inaccurate, and cannot accurately match specific needs such as close-up underwater creatures, mid-range coral reef shooting, and long-range underwater terrain shooting.

[0009] 4. Poor color reproduction: Water absorbs light of different wavelengths differently, with longer wavelengths like red and yellow being absorbed much more strongly than shorter wavelengths like blue and green. This results in a severe blue-green tint in underwater imaging. Existing color optimization algorithms are mostly based on white balance adjustment, which can only adjust the hue globally and cannot compensate for the red and yellow spectral information absorbed by the water at the physical level. Therefore, it is difficult to accurately reproduce the colors of underwater scenes.

[0010] In summary, existing technical solutions are isolated, static, and do not fully consider the physical characteristics of the underwater environment, leading to multiple challenges for mobile terminals when shooting underwater, such as sealing risks, blurry images, poor scene adaptation, and color distortion. Therefore, there is an urgent need for an integrated solution that can intelligently sense the underwater environment, dynamically adjust hardware status, and systematically optimize optics and algorithms for underwater physical characteristics. Summary of the Invention

[0011] To overcome the shortcomings of existing technologies, this invention provides an underwater scene AI adaptive multi-camera collaborative imaging system, method, and mobile terminal. The primary objective of this invention is to improve sealing reliability at different water depths by dynamically linking pressure sensors and sealing structures to adjust sealing strength in real time according to water depth. Secondly, it aims to improve image clarity by compensating for imaging deviations caused by underwater light refraction through a combination of optics and algorithms. Thirdly, it intelligently and accurately switches between different cameras based on environmental perception data to adapt to the needs of different underwater shooting distances. Finally, it effectively suppresses blue-green color cast and restores the true colors of underwater objects through a spectral-level color restoration algorithm.

[0012] To achieve the above-mentioned objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides an underwater scene AI adaptive multi-camera collaborative imaging system integrated into a mobile terminal, wherein the mobile terminal includes at least a main camera lens and a macro lens, and the system includes: 1) Pressure sensing and sealing linkage module, including a miniature pressure sensor and a sealing drive circuit; the miniature pressure sensor is set on the edge of the lens module of the mobile terminal to detect the external environmental pressure value in real time; the main control chip converts the pressure value into the current water depth data; the sealing drive circuit is connected to the electromagnetic sealing ring of the lens module and is used to dynamically adjust the current intensity flowing through the electromagnetic sealing ring according to the instructions issued by the main control chip, so that the electromagnetic adsorption force generated by the sealing ring is proportional to the water depth (water pressure), thereby realizing the adaptive adjustment of the sealing fit.

[0013] 2) The light refraction compensation module includes an optical compensation unit and a digital correction unit; the optical compensation unit is a customized optical prism set in front of the lens optical path, used to pre-compensate the incident light and partially offset the refraction shift caused by the water-air interface; the digital correction unit, based on a preset underwater refraction model, performs further digital geometric transformation and aberration correction on the image data after optical compensation, so as to completely eliminate image blur and distortion caused by refraction.

[0014] 3) A multi-camera collaborative control module is connected to the pressure sensing and sealing linkage module and multiple cameras of the mobile terminal. This module is configured to: receive current water depth data and, in conjunction with the analysis results of the shooting scene by the image recognition unit, generate a lens switching command; in a preferred embodiment, when the system determines that the shooting distance is less than or equal to 1 meter, it automatically activates the macro lens for focusing and imaging; and when the shooting distance is greater than 1 meter, it automatically switches to the main camera lens; the response time of the entire switching process is less than 300 milliseconds.

[0015] 4) The underwater color restoration module is integrated into the image signal processor (ISP). This module has a built-in color restoration algorithm for underwater spectral characteristics. Its core is to dynamically analyze the spectral distribution of the image through multi-frame image synthesis and signal processing technology, and to suppress the excessively enhanced blue and green channel signal intensity caused by water scattering. At the same time, through algorithm enhancement and synthesis, it compensates for the red and yellow channel spectral information that is severely absorbed by the water, thereby correcting the color deviation from the source and making the final image color closer to the true color of the object under natural light.

[0016] Preferably, in the pressure sensing and sealing linkage module, for every 1 meter increase in water depth, the current output by the sealing drive circuit increases linearly by 10%-20%, most preferably by 15%.

[0017] Preferably, in the light refraction compensation module, the refractive index and shape of the prism in the optical compensation unit are precisely calculated to match the typical refraction angle within a specific water depth range (e.g., 0-5 meters). The digital correction unit can fine-tune the correction parameters based on real-time water depth data to achieve more accurate compensation.

[0018] Preferably, the decision logic of the multi-camera collaborative control module further includes: when the water depth is greater than 3 meters and the image recognition unit identifies it as a "large scene" (such as an underwater cave or a vast coral reef), the ultra-wide-angle lens can be used for shooting first or simultaneously.

[0019] Secondly, this invention provides an AI-adaptive multi-camera collaborative imaging method for underwater scenes applied to the above-mentioned system, comprising the following steps: S1: The system monitors the ambient pressure in real time using a pressure sensor. When the pressure value continuously exceeds the first threshold (e.g., corresponding to a water depth of 0.2 meters) for a predetermined time, it determines that the mobile terminal has entered the underwater environment and automatically triggers the "underwater shooting mode". S2: In the underwater shooting mode, the main control chip calculates the real-time water depth based on the pressure sensor data, and generates control commands according to the preset "water depth-sealing current" mapping model to dynamically adjust the working current of the electromagnetic sealing ring and enhance the sealing force between the lens and the body. S3: Synchronously activate the light refraction compensation module. The incident light first undergoes physical refraction pre-compensation through the optical prism. Then, the raw image data acquired by the image sensor is sent to the digital correction unit. Combined with the current water depth data, the underwater refraction correction model is applied for digital processing to obtain a preliminary clear image. S4: The multi-camera collaborative control module selects the optimal imaging lens from macro lens, main camera lens, and ultra-wide-angle lens based on the current water depth and the shooting distance and scene type estimated by the AI ​​scene recognition unit through the preview image, and controls it to focus and acquire images. S5: The image signal processor runs an underwater color restoration algorithm on the image acquired in step S4. This algorithm analyzes the spectral intensity of each channel of the image, adaptively suppresses the blue and green channels, and performs spectral compensation and enhancement on the red and yellow channels, finally synthesizing and outputting a high-definition image with restored color.

[0020] Thirdly, the present invention provides a mobile terminal, including a housing, a motherboard, a processor and a memory disposed on the motherboard, and an underwater scene AI adaptive multi-camera collaborative imaging system as described in the first aspect above; the memory stores a computer program, and when the computer program is executed by the processor, it implements the underwater scene AI adaptive multi-camera collaborative imaging method as described in the second aspect above.

[0021] Compared with the prior art, the present invention has the following beneficial technical effects: 1. Achieves dynamic and highly reliable sealing protection: Through the closed-loop linkage between the pressure sensor and the sealing mechanism, the sealing strength can be adaptively adjusted with changes in water depth (water pressure), thereby solving the contradiction of fixed seals being "either too strong or too weak" at different water depths. Within the working water depth range of 0-5 meters, the risk of leakage caused by water pressure changes can be reduced by about 80%, significantly improving the safety boundary for underwater use of mobile terminals.

[0022] 2. Significantly improves the clarity of underwater imaging: The dual approach of "optical pre-compensation + digital fine correction" works together to combat underwater refraction effects from both the physical optical path and digital image levels. This effectively corrects image blurring, distortion and inaccuracy caused by refraction. Actual measurements show that the detail retention and overall clarity of underwater imaging can be improved by more than 60%.

[0023] 3. Intelligent multi-camera scene adaptation is achieved: The key environmental parameter of water depth is introduced into the multi-camera switching decision logic, and combined with AI scene recognition, the lens switching is more in line with the actual needs of underwater shooting. For example, when shooting corals at close range in shallow water, the macro lens is automatically called, and when shooting mid-to-long-range scenes in deeper water, the main camera or ultra-wide-angle lens is automatically called, which improves the shooting flexibility and scene matching by more than 50%. Users do not need to switch manually and the experience is smoother.

[0024] 4. Achieves accurate underwater color reproduction: Unlike simple white balance adjustment, the color reproduction algorithm of this invention is designed based on the selective absorption spectral characteristics of water. By enhancing the signal of missing bands and suppressing the redundant bands, color compensation is performed at the physical level. As a result, the color reproduction error of the final image can be controlled within 5%, which can more realistically reproduce the colorful underwater world and solve the core pain point of blue-green color cast in underwater imaging.

[0025] 5. Provides a highly integrated and automated user experience: The entire system automatically completes the process from water entry detection, seal adjustment, refraction compensation, lens selection to color optimization. Users only need to press the shutter to obtain high-quality underwater photos, realizing "instant optimization" of underwater scenes, which greatly reduces the user's operating threshold and professional requirements. Attached Figure Description

[0026] Figure 1 This is an overall structural block diagram of an underwater scene AI adaptive multi-camera collaborative imaging system provided in an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the working principle of the pressure sensor and the lens sealing protection current linkage in one embodiment of the present invention; Figure 3 This is a flowchart of an underwater scene AI adaptive multi-camera collaborative imaging method provided in an embodiment of the present invention (i.e., a flowchart of multi-camera collaborative switching and light refraction compensation process). Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and not for limiting the invention. It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings, not the entire structure.

[0028] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the term "connection" should be interpreted broadly, for example, it can refer to an electrical connection, or the internal connection or signal interaction between two components; those skilled in the art can understand the specific meaning of the above term in this invention according to the specific circumstances.

[0029] ( Example 1 Please see Figure 1 This embodiment provides an underwater scene AI adaptive multi-camera collaborative imaging system, which is integrated into a smartphone with an IP68 waterproof rating. The system mainly includes: a pressure sensing and sealing linkage module, a light refraction compensation module, a multi-camera collaborative control module, an underwater color restoration module, and a main control chip as the control core.

[0030] The pressure sensing and sealing linkage module includes a MEMS (Micro-Electro-Mechanical Systems) miniature pressure sensor and a sealing drive circuit. The MEMS pressure sensor is precisely embedded inside the metal decorative ring of the rear camera module of the mobile phone, and its pressure-sensing diaphragm is in contact with the external environment to detect the absolute environmental pressure in real time. The sealing drive circuit is electrically connected to an annular electromagnetic sealing ring around the camera module. The electromagnetic sealing ring is made of magnetic material and has a coil embedded inside. When not energized, it provides basic sealing force, and when energized, it generates additional electromagnetic force to enhance the sealing and tightening effect.

[0031] The main control chip (which can be the main SoC of a mobile phone or a dedicated image coprocessor) pre-stores a pressure-water depth conversion formula (P=ρgh+P0, where P is the pressure measured by the sensor, ρ is the water density, g is the gravitational acceleration, h is the water depth, and P0 is the atmospheric pressure) and a "water depth-current" lookup table. When the system is running, the data from the pressure sensor is read in real time and converted into a water depth value h. The main control chip, based on the current water depth h, looks up the "water depth-current" lookup table to obtain the corresponding target current value I, and sends it to the sealing drive circuit. The command is issued; the sealing drive circuit is typically a programmable constant current source, which, upon receiving the command, precisely adjusts the output current to I; the current I flows through the coil of the electromagnetic sealing ring, generating an electromagnetic attraction force proportional to the current magnitude, which presses the sealing gasket between the camera lens barrel and the phone casing more tightly; in this embodiment, the "water depth-current" lookup table can be set as follows: when the water depth is 0m, the current is the baseline value I0 (used to maintain the basic seal); for every 1m increase in water depth, the current I increases linearly by 15%; for example, at a water depth of 5m, the current is I0. (1+15%× 5)=1.75I0. This dynamic adjustment ensures that the sealing force is always optimally matched with the water pressure from the water surface to the maximum allowable depth, which avoids unnecessary power consumption in shallow water areas and ensures the sealing safety in deep water areas.

[0032] The light refraction compensation module includes an optical compensation unit and a digital correction unit. The optical compensation unit is a specially designed wedge-shaped optical glass prism that is attached to the inside of the camera's protective glass. Its wedge angle and refractive index are calculated so that light rays incident from the water at a certain angle are refracted by the prism and exit at an angle closer to perpendicular to the image sensor (such as CMOS) plane, thus partially pre-correcting the large-angle refraction at the water-air interface. The digital correction unit is implemented by an FPGA (Field Programmable Gate Array) chip, which has an underwater refraction distortion correction model built based on a ray tracing algorithm. This correction model not only considers the compensation of the fixed prism but also fine-tunes the residual distortion according to the real-time water depth h. The raw RAW data acquired by the image sensor is first processed by the FPGA for digital geometric transformation and aberration correction before entering the ISP (Image Signal Processor) for conventional processing (such as de-mosaicing and noise reduction) to eliminate the remaining edge blur and shape distortion.

[0033] The logic of the multi-camera collaborative control module is executed by the image processing unit (IPU) in the main control chip. The phone in this embodiment is equipped with three rear cameras: a macro lens with a 2.5cm focal length, a main camera lens with an equivalent 24mm focal length, and an ultra-wide-angle lens with an equivalent 16mm focal length. The input signals to the multi-camera collaborative control module include: the current water depth h from the pressure sensing module, and the real-time analysis results of the camera preview image from the AI ​​scene recognition unit (integrated in the ISP or NPU) (including the estimated object distance d and scene labels such as "microorganisms," "coral close-up," "distant landscape," "wide scene," etc.). Its collaborative decision-making logic (i.e., a multi-condition decision tree, which includes at least the following decision branches) is as follows: When the AI ​​identifies the scene as "microorganisms" or "texture details" and estimates the object distance d≤1m, the macro lens will be activated first, regardless of the water depth. When the estimated object distance d > 1m: If the water depth h ≤ 3m, or the scene is labeled "normal distant view", then the main camera lens will be activated; If the water depth h > 3m and the scene label is "wide scene" (e.g., the image has large edge curvature and wide content), the ultra-wide-angle lens will be activated first to obtain a more stunning view. When switching modes (such as switching from macro to main camera), the system ensures a smooth transition of focus and exposure parameters. The entire switching process (including physical lens activation, focusing, and locking) takes less than 300 milliseconds, and users can hardly perceive any lag.

[0034] The underwater color restoration module is deeply integrated into the ISP pipeline as a software algorithm. It runs after completing routine steps such as white balance and color matrix conversion. Its core algorithm is a color mapping network trained by deep learning. The training data of this network includes a large number of paired images taken at the same location, one in air (true color) and the other underwater (color cast). When the algorithm is working, it first analyzes the statistical characteristics of the input image (already refraction compensated) in color spaces such as RGB and Lab, especially evaluating the intensity of the blue and green channels relative to the red channel. Then, it predicts the missing red and yellow components of each pixel through the network model and compensates for them. At the same time, it adaptively suppresses oversaturated blue and green areas. This process is not a simple global tone adjustment, but a spectral reconstruction based on local image content, which can control the color restoration error of the final output image (calculated using the ΔE color difference formula) within 5%, significantly improving the color realism.

[0035] Specifically, the underwater color restoration module includes a spectral feature extraction unit, a color mapping and reconstruction unit, and a post-processing and fusion unit. The spectral feature extraction unit performs temporal and spatial analysis on the input single or multiple frames of raw image data, calculates the intensity distribution histograms of the image in the red, green, and blue channels, and estimates the absorption and attenuation coefficients of the water body for red and yellow wavelengths. The color mapping and reconstruction unit is connected to the spectral feature extraction unit and internally stores or performs real-time calculations of a color conversion function. This color conversion function is obtained by collecting a large number of "true colors in the air" under the same lighting conditions. The image is obtained by training a deep learning network model on a paired dataset of "image-corresponding underwater color-distorted image". The color mapping and reconstruction unit applies the color conversion function to operate on each pixel or pixel region of the input image: nonlinear enhancement and signal reconstruction are performed on the red and yellow spectral components that are severely attenuated due to water absorption; adaptive suppression is performed on the blue and green spectral components that are relatively strong due to water scattering. The post-processing and fusion unit is used to perform noise reduction, edge enhancement and color smoothing on the image after color mapping and reconstruction, and finally synthesize and output the final image.

[0036] Preferably, before the spectral feature extraction unit performs analysis, a preliminary global white balance correction is performed using the gray area or a preset white reference area in the image data to eliminate the overall color cast underwater; the color conversion function is applied to the image data after the preliminary global white balance correction to perform more refined local and spectral color restoration.

[0037] Combination Figure 2As shown, the working mechanism of pressure sensing and sealing linkage is based on the linear relationship between current and water depth. The driving current value I and the water depth value h can satisfy the relationship: I = f(h), where f(h) is a monotonically increasing function in the water depth range from 0 to h_max, and the slope of f(h) increases when h is large to cope with the nonlinear increase of water pressure. Under different water depths, the energization state of the electromagnetic sealing ring is as follows: In the 1m shallow water area, the current is small, and the electromagnetic sealing ring generates a medium electromagnetic force (F1), which is superimposed with the basic sealing force to provide sufficient sealing; while in the 5m deep water area, the current increases to the maximum, and the electromagnetic sealing ring generates a strong electromagnetic force (F5), ensuring that the sealing gasket is tightly pressed under high water pressure to effectively prevent water leakage.

[0038] Please see Figure 3 This embodiment also provides an imaging method based on the above system, including the following steps: Step S310, Environmental Perception and Mode Triggering: The pressure sensor continuously monitors the environmental pressure. When the detected pressure value exceeds a certain threshold of the land atmospheric pressure (corresponding to the pressure of about 0.2 meters of water column) for more than 2 seconds, the main control chip determines that the device has entered the water and automatically triggers the "underwater shooting mode". A water droplet icon appears on the camera UI interface as a prompt. Step S320, Adaptive sealing adjustment; while triggering the underwater mode, the main control chip calculates the precise water depth h based on the real-time pressure value, and immediately queries the "water depth-current" lookup table to send a command to the sealing drive circuit to adjust the working current of the electromagnetic sealing ring to the appropriate value I(h) corresponding to h, thereby achieving an instantaneous enhancement of the sealing force. Step S330, Light Refraction Compensation; After the underwater mode is triggered, the light refraction compensation module is automatically activated. Light from the scene being photographed passes through the water, the phone's protective glass, and the optical compensation prism in sequence, and is finally captured by the image sensor. The captured raw image data is sent to the digital correction unit (FPGA), which calls the correction coefficient related to the current water depth h to perform pixel-level geometric correction and sharpening processing on the image, and outputs preliminary clear image data. Step S340: Intelligent multi-camera collaborative imaging; The multi-camera collaborative control module works in parallel. It receives the current water depth h and starts the AI ​​scene recognition unit to analyze the camera preview stream, quickly estimate the object distance d and determine the scene type; Based on the built-in decision tree logic (as described above), the module selects the optimal lens from macro, main camera and ultra-wide-angle, and controls the lens motor to complete the focusing; The selected lens performs the final image capture to obtain high-quality raw image data; Step S350: Underwater Color Restoration and Output; The image data captured in step S340 is sent to the ISP pipeline; In the underwater color restoration module, a dedicated color restoration algorithm is applied, for example, a convolutional neural network with a U-Net structure. Its input is an sRGB image that has undergone white balance pre-correction, and its output is a color-corrected image. The loss function of the convolutional neural network combines L1 loss, perceptual loss, and color difference loss in the Lab color space to simultaneously ensure color accuracy and the visual naturalness of the image; This color restoration algorithm can perform spectral analysis and local adjustment on the image, strongly compensate for red and yellow information, and moderately suppress blue and green, ultimately generating a bright and realistic high-quality underwater photo or video frame, which is presented to the user or stored in the memory.

[0039] ( Example 2 This embodiment is a cost-optimized variation of Embodiment 1, mainly targeting entry-level or cost-sensitive waterproof mobile phones. In this embodiment, the MEMS pressure sensor in the pressure sensing and sealing linkage module is replaced with a capacitive water depth sensor. This sensor estimates the water depth by detecting the capacitance change between the exposed metal contacts of the lens module caused by the change in the dielectric constant of the water. This method has a lower cost, and although its absolute accuracy is slightly lower than that of the pressure sensor, it is sufficient to distinguish different depth ranges within the 0-5 meter range, which can meet the needs of graded adjustment of sealing force.

[0040] The optical compensation unit (glass prism) in the light refraction compensation module has been removed, relying entirely on the digital correction unit, i.e., a purely software-based "digital refraction compensation" scheme is adopted. The algorithm of the digital correction unit is enhanced, and its correction model needs to be more complex to compensate for the missing optical pre-correction. For example, the current water depth value h and the turbidity parameter t from the water quality sensor are obtained; based on h and t, a set of refraction distortion correction parameters are indexed from a pre-stored multidimensional lookup table, including a pixel displacement map and a point spread function correction kernel; the pixel displacement map is applied to remap the original image to correct spatial distortion; the point spread function correction kernel is applied to deconvolve the remapped image to restore image detail clarity. Although this requires higher computing power from the processing chip (such as the DSP of a mid-range mobile phone), it saves the hardware cost of the optical prism.

[0041] Meanwhile, the multi-camera collaborative control module has been simplified, with the phone only equipped with a macro lens and a main camera lens. Its collaborative logic has been simplified to: based on the object distance d estimated by AI, the macro lens is used when d≤1m and the main camera lens is used when d>1m, eliminating the complex judgment of water depth and ultra-wide-angle lens. The algorithm parameters of the color reproduction module can also be appropriately simplified, reducing computational complexity while ensuring significant improvement in color cast. This embodiment can still provide a superior imaging experience compared to the existing "underwater mode" while significantly reducing costs.

[0042] ( Example 3 This embodiment is a functional extension of embodiment 1, mainly targeting high-end flagship mobile phones or professional action cameras, to expand underwater shooting capabilities to a deeper and wider range.

[0043] First, the detection range of the pressure sensing module has been expanded from 0-5 meters to 0-10 meters; correspondingly, the "water depth-current" adjustment model has been modified from a linear model to a piecewise curve model; for example, it maintains linear growth (+15% per meter) in the 0-5 meter depth range, and in the 5-10 meter depth range, due to the accelerated increase in water pressure, the current adjustment curve becomes non-linear (such as exponential) to cope with greater pressure with stronger sealing force; the system will issue a flashing warning to the user on the screen when it detects that the water depth is close to the waterproof design limit of the equipment (such as 9 meters).

[0044] Secondly, a "water quality identification module" (not shown in the figure) is added. This module can work in conjunction with the phone's existing ambient light sensor and flash: the flash emits light pulses of a specific spectrum, and the ambient light sensor detects the attenuation and spectral changes of the reflected light. By analyzing transmittance data, the system can roughly determine whether the water body is clear freshwater, seawater, or turbid water (rich in plankton or silt). Water quality information (such as turbidity level) will be input as parameters to the light refraction compensation module and the underwater color restoration module, and their algorithm parameters will be dynamically fine-tuned. For example, in turbid water, the refraction compensation and color restoration algorithms will adopt more aggressive descattering and color correction strategies.

[0045] Furthermore, in terms of multi-camera collaboration, in addition to the existing lenses, an extra telephoto lens (such as an equivalent 100mm focal length) can be added. The decision logic of the multi-camera collaboration control module is expanded to: when shooting small to medium-sized underwater creatures (such as fish) at a distance, if the object distance d>3m and the AI ​​identifies it as a "small moving target", the telephoto lens can be automatically called to take the picture in order to obtain better composition and details; at the same time, the image stabilization system will also coordinate with the water wave frequency data detected by the pressure sensor to improve the success rate of the shot.

[0046] It should be noted that the technical features in the above embodiments 1, 2 and 3 can be freely combined, as long as they do not violate the basic principles and core concepts of the present invention, they are all within the protection scope of the present invention; for example, the capacitive water depth sensor of embodiment 2 can be combined with the water quality identification module of embodiment 3 to form a new implementation scheme.

[0047] Furthermore, the system and method of the present invention are not only applicable to smartphones, but can also be applied to other mobile or fixed devices with shooting functions, such as tablet computers, action cameras, underwater drones, and waterproof surveillance cameras, after appropriate miniaturization adaptation.

[0048] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural transformations made using the contents of the present invention's specification and drawings under the inventive concept of the present invention, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. An AI-adaptive multi-camera collaborative imaging system for underwater scenes, characterized in that, Integrated into a mobile terminal, the system includes: The pressure sensing and sealing linkage module is used to detect the external environmental pressure in real time and convert it into water depth data, and dynamically adjust the sealing strength of the mobile terminal lens module according to the water depth data. The light refraction compensation module is used to compensate for the deviation of the light path due to refraction at the water-air interface in order to obtain clear image data; The multi-camera collaborative control module is communicatively connected to the pressure sensing and sealing linkage module, and is used to select a target camera that is suitable for the current underwater shooting scene from multiple cameras of the mobile terminal for imaging, based at least on the water depth data. The underwater color restoration module is used to process the image data acquired by the target camera, suppress color deviation caused by selective light absorption by the water, and restore the true colors of the image.

2. The underwater scene AI adaptive multi-camera collaborative imaging system according to claim 1, characterized in that, The pressure sensing and sealing linkage module includes: A pressure sensor is located in the lens module area to detect the ambient pressure value; The main control chip is used to calculate the current water depth based on the environmental pressure value; A sealing drive circuit is used to adjust the current flowing to an electromagnetic sealing ring, so that the electromagnetic attraction force generated by the electromagnetic sealing ring matches the water pressure corresponding to the current water depth.

3. The underwater scene AI adaptive multi-camera collaborative imaging system according to claim 2, characterized in that, The sealing drive circuit is configured such that the current increases linearly by a preset ratio for every unit increase in the current water depth; and the electromagnetic sealing ring is arranged around the lens module and the mobile terminal housing.

4. The underwater scene AI adaptive multi-camera collaborative imaging system according to claim 1, characterized in that, The light refraction compensation module includes: An optical compensation unit, located in the camera's optical path, is used to pre-correct the physical refraction angle of incident light. The digital correction unit is used to perform digital geometric transformation and aberration correction on the image data pre-corrected by the optical compensation unit based on a preset underwater refraction model.

5. The underwater scene AI adaptive multi-camera collaborative imaging system according to claim 4, characterized in that, The digital correction unit is also configured to receive current water depth data and dynamically adjust the correction parameters of the underwater refraction model based on the current water depth data.

6. The underwater scene AI adaptive multi-camera collaborative imaging system according to claim 1, characterized in that, The multi-camera collaborative control module is configured as follows: Receive current water depth data, as well as shooting distance and scene type information obtained based on image recognition; When the shooting distance is less than or equal to the first threshold, the macro lens is activated as the target camera. When the shooting distance is greater than the first threshold, the main camera lens is activated as the target camera.

7. The underwater scene AI adaptive multi-camera collaborative imaging system according to claim 6, characterized in that, The multi-camera collaborative control module is also configured to: when the current water depth is greater than a depth threshold and the scene type information indicates a large scene, activate the ultra-wide-angle lens as the target camera.

8. The underwater scene AI adaptive multi-camera collaborative imaging system according to claim 1, characterized in that, The underwater color restoration module processes image data in the following ways: Analyze the intensity distribution of the image data across different color channels; Enhance the spectral components of color channels whose intensity is below a first threshold in the image data; Suppress the spectral components of color channels whose intensity is higher than a second threshold in the image data.

9. An AI-adaptive multi-camera collaborative imaging method for underwater scenes, characterized in that, The method, applied to the underwater scene AI adaptive multi-camera collaborative imaging system as described in any one of claims 1-8, comprises: In response to the detection that the mobile terminal has entered an underwater environment, an underwater shooting mode is triggered; In the underwater shooting mode, the current water depth is determined based on real-time detected pressure data, and the sealing strength of the lens module is dynamically adjusted based on the current water depth. Activate light refraction compensation to correct refraction deviations in the light signals received by the image sensor; Based at least on the current water depth, a target camera is determined from multiple cameras, and the target camera is controlled to acquire images; An underwater color restoration algorithm is run on the images captured by the target camera to output a color-corrected image.

10. A mobile terminal, characterized in that, include: case; The motherboard is housed within the casing. A processor and memory are mounted on the motherboard, and the memory stores computer programs. And, the underwater scene AI adaptive multi-camera collaborative imaging system as described in any one of claims 1-8; When the computer program is executed by the processor, it implements the underwater scene AI adaptive multi-camera collaborative imaging method as described in claim 9.

Images