Camera focusing device with two-dimensional feedback control capability

Abstract

The invention provides a camera focusing device with two-dimensional feedback control capability, and relates to the technical field of camera focusing, and the camera focusing device comprises an image collection unit, a two-dimensional target characterization unit, a focusing driving unit, an image response characteristic analysis unit, and a cooperative iteration control unit. A two-dimensional target characterization unit is used for generating target characterization containing a geometric contour and texture complexity, and an image response characteristic analysis unit is used for evaluating the presentation state of a characteristic attribute set in real time to guide a collaborative iteration control unit to dynamically adjust the focus depth and optimize characterization parameters. The two-way feedback mechanism enables the system to adaptively cope with complex scenes such as light change, multi-target interference or target motion, the focusing speed and precision are remarkably improved, dynamic optimization of target characterization and focus adjustment are deeply fused, and a brand new technical path is provided for high-performance automatic focusing in the dynamic complex scenes.

Description

Technical Field

[0001] The present invention relates to the technical field of camera focusing, and specifically to a camera focusing device with two-dimensional feedback control ability. Background Art

[0002] According to a camera module autofocus control method and device disclosed in Chinese Patent Publication No. "CN114598813A", it includes: obtaining at least three preset calibration distances; performing focus calibration on the preset calibration distances, including using two of the calibration distances to perform focus calibration tests to obtain the actual motor stroke encodings corresponding to the two calibration distances respectively; based on the calibration distances and the actual motor stroke encodings, determining the calculated motor stroke encodings corresponding to the remaining calibration distances. In the process of determining the calculated motor stroke encodings corresponding to the remaining calibration distances based on the calibration distances and the actual motor stroke encodings, it is determined by calculation means, which can save more time compared to the method of using a machine tool for focus calibration all the time, thereby improving the production efficiency of the camera module.

[0003] The above patent document and the prior art have the following technical problems when in use: Problem 1: In the above document and the prior art, the existing camera focusing technology usually relies on a single clarity evaluation index or a fixed focus search strategy. In complex dynamic scenarios such as drastic light changes, fast target movement, or strong background interference, these methods often face problems of slow focusing speed and insufficient accuracy. The limitation of traditional methods lies in the lack of dynamic understanding of target features and adaptive adjustment ability, resulting in the focusing process being easily interfered by the environment and difficult to quickly converge to the best focus. Problem 2: In the above document and the prior art, the focusing system is usually designed to focus on a single target and is difficult to effectively process scenarios containing multiple potential focusing objects. The limitation of single-target focusing stems from the lack of multi-target parallel processing ability and intelligent decision-making mechanism, restricting the applicability of the system in complex scenarios. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides a camera focusing device with two-dimensional feedback control ability, which solves the problems of the prior art.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: A camera focusing device with two-dimensional feedback control ability, the camera focusing device includes an image acquisition unit, a two-dimensional target characterization unit, a focusing drive unit, an image response characteristic analysis unit, and a collaborative iterative control unit, and further includes the following: The image acquisition unit is used to capture the image data stream of the scene; A two-dimensional target representation unit, connected to the image acquisition unit, for generating and iteratively updating the target representation of one or more potential focus targets on the two-dimensional image plane according to the received image data, and the target representation includes the position information of the target and a set of characteristic attributes describing its recognizability; A focus driving unit, for adjusting the depth of focus of the optical component according to the control instruction; An image response characteristic analysis unit, connected to the image acquisition unit, for analyzing the presentation state of the image characteristics and the set of characteristic attributes corresponding to the target representation in the current image at the current depth of focus, and the image characteristics include at least one sharpness metric; A collaborative iterative control unit, connected to the two-dimensional target representation unit, the focus driving unit, and the image response characteristic analysis unit, for performing the following steps: Sp1: Generating a first control instruction to the focus driving unit to adjust the depth of focus based on the target representation of the current target output by the two-dimensional target representation unit; Sp2: Receiving the image characteristic analysis result at the depth of focus adjusted in Sp1 feedback by the image response characteristic analysis unit; Sp3: Generating an update instruction to the two-dimensional target representation unit based on the presentation state of the set of characteristic attributes in the image characteristic analysis result to adjust and optimize its two-dimensional target representation of one or more potential focus targets; Sp4: Generating a second control instruction to the focus driving unit based on the updated and optimized two-dimensional target representation in Sp3 and combining the sharpness metric for subsequent depth of focus adjustment; Sp5: Repeatedly executing Sp2 to Sp4, until the preset focus completion condition is met, through the iterative enhancement between the optimization effect of the depth of focus adjustment on the two-dimensional target representation and the precise guiding effect of the optimized two-dimensional target representation on the subsequent depth of focus adjustment.

[0006] Preferably, the target representation generated by the two-dimensional target representation unit includes the geometric contour parameters of the target and the description of the texture complexity inside the contour. When the two-dimensional target representation unit receives the update instruction, it adjusts the weight of the geometric contour parameters and the sensitivity threshold of the texture complexity description to achieve the adjustment and optimization of the two-dimensional representation.

[0007] Preferably, the presentation state of the set of characteristic attributes output by the image response characteristic analysis unit includes the gradient intensity, the spatial frequency energy distribution, and the statistical dispersion of the characteristic attributes at the current depth of focus.

[0008] Preferably, in Sp3, if the analysis result indicates that the current depth of focus enhances the recognizability of a certain specific characteristic attribute in the target representation, the update instruction will strengthen the significance of the specific characteristic attribute in the two-dimensional representation.

[0009] Preferably, in Sp4, the collaborative iterative control unit uses a prediction model based on the change trend of sharpness metric and the stability of two-dimensional target representation to determine the amplitude and direction of subsequent focus depth adjustment.

[0010] Preferably, the preset focus completion conditions of the collaborative iterative control unit include at least one of the following: the sharpness metric reaches a stable maximum value, the continuous iterative update amount of the two-dimensional target representation is less than a predetermined threshold, and the number of iterations reaches the upper limit.

[0011] Preferably, when there are multiple potential focus targets, the two-dimensional target representation unit maintains independent two-dimensional target representations for each potential focus target, and the collaborative iterative control unit executes the iterative enhancement process for each target representation in parallel and selects the final focus target according to a preset priority strategy.

[0012] Preferably, the collaborative iterative control unit includes an initialization module for providing an initial two-dimensional target representation for the two-dimensional target representation unit by quickly analyzing the global image and receiving external specifications before the first iteration.

[0013] Preferably, the camera focusing device includes the following steps: Sp1: Capturing image data through an image acquisition unit; Sp2: The two-dimensional target representation unit generates and iteratively updates the two-dimensional target representation of the potential focus target; Sp3: The focus driving unit adjusts the focus depth according to the first control instruction of the collaborative iterative control unit, and the first control instruction is based on the current two-dimensional target representation; Sp4: The image response characteristic analysis unit analyzes the image characteristics after adjusting the focus depth, including the sharpness metric and the presentation state of the feature attribute set in the two-dimensional representation; Sp5: The collaborative iterative control unit updates the two-dimensional target representation based on the presentation state of the feature attribute set in the image characteristic analysis result; Sp6: The collaborative iterative control unit generates a second control instruction based on the updated two-dimensional target representation and the sharpness metric and gives it to the focus driving unit for subsequent focus depth adjustment; Sp7: Loop and execute Sp4 to Sp6 until the preset focus completion conditions are met.

[0014] The present invention provides a camera focusing device with two-dimensional feedback control ability, having the following beneficial effects: 1. The present invention introduces a two-dimensional feedback mechanism, tightly coupling the focus depth adjustment with the optimization of the target representation, forming a unique iterative enhancement closed-loop system. The two-dimensional target representation unit is used to generate a target representation including geometric contours and texture complexity, and the image response characteristic analysis unit evaluates the presentation state of the feature attribute set in real time, guiding the collaborative iterative control unit to dynamically adjust the focus depth and optimize the representation parameters. This two-way feedback mechanism enables the system to adaptively handle complex scenarios such as light changes, multi-target interference, or target movement, significantly improving the focusing speed and accuracy. By deeply integrating the dynamic optimization of the target representation with the focus adjustment depth, it breaks through the limitations of traditional focusing methods and provides a new technical path for high-performance autofocus in dynamic complex scenarios.

[0015] 2. The present invention adopts a parallel optimization and preset priority strategy for supporting multiple potential focusing targets, greatly enhancing the intelligent adaptability and application flexibility of the system. Traditional focusing systems usually only focus on a single target and are difficult to handle scenarios containing multiple potential focusing objects, such as multi-person scenarios in surveillance systems or multi-tissue imaging in medical images. The two-dimensional target representation unit independently maintains a representation for each target, and the collaborative iterative control unit performs iterative optimization in parallel and dynamically selects the final focusing target according to comprehensive priority strategies such as user specification, target position, size, or motion state. This multi-target processing ability not only improves the robustness of the system but also significantly enhances its applicability in complex environments. By combining multi-target parallel optimization with intelligent decision-making, it breaks through the limitation of single-target focusing and provides an efficient solution for intelligent vision applications in multi-scenarios and multi-targets. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is the structural composition diagram of the camera focusing device of the present invention; Figure 2 is the control step diagram of the collaborative iterative control unit of the present invention; Figure 3 is the operation flow diagram of the camera focusing device of the present invention; Figure 4 is the operation relationship diagram of the camera focusing device of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0017] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention. Specific Embodiment 1: As Figures 1 to 4As shown in the figure, a camera focusing device with two-dimensional feedback control ability, the camera focusing device includes an image acquisition unit, a two-dimensional target characterization unit, a focusing drive unit, an image response characteristic analysis unit, and a collaborative iterative control unit, and further includes the following: The image acquisition unit is the visual input source of the entire focusing device. Its core components are usually high-resolution CMOS and CCD image sensors, which can continuously capture the dynamic image data stream of the scene at a preset frame rate of 30 frames per second or 60 frames per second. The image data stream includes unprocessed raw Bayer format data and YUV or RGB color space data preliminarily processed by the built-in image signal processor of the sensor. The image acquisition unit transmits the captured digital image frames to the two-dimensional target characterization unit and the image response characteristic analysis unit in real time through the MIPI CSI-2 high-speed serial interface, providing the necessary visual information for subsequent target recognition, feature analysis, and clarity evaluation. In the working mode, the image acquisition unit can respond to the precise trigger signal of the collaborative iterative control unit and capture a single frame or a limited number of multi-frame images at a specific moment to cooperate with specific analysis and focusing strategies.

[0019] The two-dimensional target representation unit, as a key module in the focusing device for understanding and describing potential focusing targets, is closely connected to the image acquisition unit. According to the real-time image data received from the image acquisition unit, in the two-dimensional image plane, that is, the X-Y plane, it intelligently generates and continuously optimizes in subsequent iterative processes the digital representation of one or more potential focusing targets. The target representation includes the position information of the target in the two-dimensional image and a set of characteristic attributes describing the current recognizable state of the target. The position information in the target representation is composed of a set of parameters that precisely define the geometric contour of the target, and the boundary of the target is dynamically depicted through the active contour model and the control point sequence of the parametric spline curve. At the same time, the set of characteristic attributes deeply describes the visual characteristics of the target. Among them, the description of the texture complexity inside the contour is extracted by analyzing statistical features such as contrast, energy, and entropy from the gray-level co-occurrence matrix, and is quantified by calculating structural and frequency-domain features such as the local binary pattern histogram and the response energy distribution of the Gabor filter bank. When the collaborative iterative control unit issues an update instruction, the two-dimensional target representation unit actively responds. By finely adjusting the weights of the parameters constituting the geometric contour, enhancing the influence of the control points of the contour segments that become clearer due to focusing improvement, adjusting the sensitivity threshold of the texture complexity description, and adjusting the sampling parameters of the LBP operator or the frequency selection of the Gabor filter when the texture details are richer, the dynamic adjustment and continuous optimization of the two-dimensional target representation are achieved, making it more accurately reflect the true state of the target under the current visual conditions. The target representation data processed by the two-dimensional target representation unit, including the unique identifier of the target, the precise contour definition, the quantified texture feature vector, as well as a comprehensive recognizability score and representation stability metric, will be output as key information to the collaborative iterative control unit and the image response characteristic analysis unit.

[0020] The focusing drive unit is the actuator that realizes the focus adjustment at the physical level. It directly responds to the precise control instructions issued by the collaborative iterative control unit and changes the focus depth of the camera by driving the internal optical components, mainly the movable lens group. It includes voice coil motors VCM, stepper motors, piezoelectric motors, and other precision drive mechanisms. These mechanisms can convert the digital control signals issued by the collaborative iterative control unit, the encoded position value corresponding to the target focus depth, the required number of movement steps, or a pulse sequence of a specific frequency, into the precise physical displacement of the lens group. At the same time, a position sensor is integrated, specifically a Hall effect sensor, to real-time monitor the current position of the lens group and feedback this position information to the collaborative iterative control unit, thus forming a closed-loop control of the physical position of the lens, further improving the accuracy and response speed of the focus adjustment.

[0021] The Image Response Characteristic Analysis Unit is closely connected to the Image Acquisition Unit and analyzes in real time the specific image characteristics corresponding to the target representation area currently output by the Two-Dimensional Target Representation Unit in the image captured by the Image Acquisition Unit at the current depth of focus set by the Focus Drive Unit. On the one hand, it calculates the image sharpness metric of the target representation area, using a variety of mature sharpness evaluation algorithms, including the Tenengrad function based on image gradients, the variance of the Laplacian operator, and the high-frequency energy ratio based on image frequency domain analysis, and outputs one or more quantified sharpness scores. On the other hand, it deeply analyzes the specific presentation state of the feature attribute set defined in the target representation at the current image and the current depth of focus. For each feature attribute described by the Two-Dimensional Target Representation Unit, including the edge sharpness of specific segments of the contour and the richness of texture details in specific regions, the Image Response Characteristic Analysis Unit will re-measure and evaluate these features in the current image frame. The analysis of the presentation state of the feature attribute set at the current depth of focus includes precisely calculating the current gradient intensity values of these feature attributes, analyzing the specific distribution of their spatial frequency energy, evaluating their statistical dispersion, the variance of pixel gray values, and the histogram entropy of specific texture descriptors. These analysis results, including the quantified sharpness metric and a detailed report on the presentation state of the feature attribute set, will be used as a structured data packet, containing the target ID, various sharpness scores, and a vector of the current values of each feature attribute, and will be fed back to the Collaborative Iterative Control Unit in real time as an important basis for its subsequent decisions.

[0022] The Collaborative Iterative Control Unit is the intelligent core and command center of the entire focusing device. It is connected to the Two-Dimensional Target Representation Unit, the Focus Drive Unit, and the Image Response Characteristic Analysis Unit. By executing a series of carefully designed iterative steps, it realizes the closed-loop feedback and optimization of the entire focusing process. At the beginning of the focusing task, the initialization module inside the Collaborative Iterative Control Unit is first activated. This module will instruct the Image Acquisition Unit to capture an initial scene image and identify a preliminary potential focusing target area by performing a quick saliency analysis on the global image or directly receiving user-specified input from the outside. This initial area information is then transmitted to the Two-Dimensional Target Representation Unit, which generates the first version of the target representation data structure based on this. Subsequently, the system enters the core iterative focusing loop, marked as the current iteration number k, and the specific steps include the following: Sp1: Generate a first control instruction to the Focus Drive Unit to adjust the depth of focus based on the target representation of the current target output by the Two-Dimensional Target Representation Unit. The Collaborative Iterative Control Unit receives the current target representation data from the Two-Dimensional Target Representation Unit, which is represented as at the k-th iteration. If it is the first iteration, that is, , the Collaborative Iterative Control Unit will instruct the Focus Drive Unit to execute a preset exploratory focus search mode based on the initial target representation A small - range search for sharpness or a focus scan along a predefined trajectory. In subsequent iterations, that is when, the collaborative iteration control unit will use its internal prediction model to determine the focus adjustment strategy; this prediction model preferably combines the historical change trend of sharpness metrics with the stability information of the two - dimensional target representation itself; if historical data shows that the current focus adjustment direction can continuously improve sharpness, and the target representation relative to the representation in the previous iteration remains highly stable, indicating that the system's understanding of the target tends to be consistent, then the prediction model will instruct to continue adjusting the focus depth along the current direction, while appropriately increasing the adjustment step size to accelerate convergence; conversely, if sharpness decreases, fluctuates violently, or the target representation changes significantly, indicating that the current focusing strategy is poor and the understanding of the target is not yet mature, then the prediction model will instruct to reduce the focus adjustment step size, switch to a more conservative exploratory adjustment, and conduct a more detailed search around the current focus position to assist in stabilizing and optimizing the target representation; this step finally generates an accurate first control instruction, which is a specific target focus depth encoded value or a relative displacement amount, and sends it to the focus drive unit.

[0023] Sp2: Receive the image characteristic analysis results at the focus depth adjusted in Sp1, which are fed back by the image response characteristic analysis unit. After the focus drive unit completes the physical displacement of the lens group according to the first control instruction issued in Sp1, enabling the camera to reach the new focus depth the image acquisition unit will capture the scene image at this new focus depth. The image response characteristic analysis unit immediately analyzes this new image, especially for the image area corresponding to the target representation determined by the two - dimensional target representation unit in the previous iteration Calculate its sharpness metric and the specific presentation status of each feature attribute set in the target representation in this new image. After the analysis is completed, the image response characteristic analysis unit will feedback the image characteristic analysis result structure containing these detailed information to the collaborative iteration control unit in real - time.

[0024] Sp3: Generate an update instruction to the two - dimensional target representation unit based on the presentation status of the feature attribute set in the image characteristic analysis results to adjust and optimize its two - dimensional target representation of one or more potential focusing targets. After receiving the image characteristic analysis results fed back in Sp2, the collaborative iteration control unit will carefully compare the data on the presentation status of the feature attribute set in it with the expected values and previous states of the corresponding attributes recorded in the current target representation If the analysis results clearly indicate that the current new focus depth significantly enhances the target representation Regarding the identifiability of a specific feature attribute in [text], if the gradient intensity of a certain edge segment of the target contour is significantly enhanced and the texture detail contrast of a certain area becomes more distinct, the update instruction generated by the collaborative iterative control unit will direct the two-dimensional target representation unit to strengthen the significance and weight of this specific feature attribute in its internal two-dimensional representation model. This update instruction is a structured message that precisely guides the two-dimensional target representation unit on how to adjust its internal model parameters. By adjusting the weights of the control points defining the geometric contour to make it more inclined towards the newly enhanced edge and modifying the parameters of the texture descriptor to better capture the fine texture revealed by the focus improvement. After the two-dimensional target representation unit executes this update instruction, the target representation within it evolves into a better .

[0025] Sp4: Based on the updated and optimized two-dimensional target representation in Sp3 and combined with the sharpness metric, generate a second control instruction to the focus drive unit for subsequent focus depth adjustment. After the two-dimensional target representation unit completes its representation optimization and outputs the updated target representation , the collaborative iterative control unit combines the sharpness metric values obtained from Sp2 corresponding to the focus depth and the previous representation to decide the next focus adjustment strategy. The second control instruction generated in this step is essentially to prepare the focus adjustment instruction for the next iteration loop, that is, the Sp1 step of the th iteration. Its decision logic is similar to the prediction model in Sp1, but the key difference is that it makes a judgment based on the latest target representation with a deeper and more accurate understanding of the target; if is significantly optimized and changed compared to , the focus adjustment strategy will be adjusted accordingly, resetting the search range or changing the search step size in order to reach a better focus faster under the guidance of the new representation.

[0026] Sp5: Loop through Sp2 to Sp4. Through the iterative enhancement between the optimization effect of the two-dimensional target representation by adjusting the focus depth and the precise guidance of the subsequent focus depth adjustment by the optimized two-dimensional target representation, until the preset focus completion condition is met, the iteration counter k inside the collaborative iterative control unit is incremented. The system repeats the closed-loop operations of Sp2, Sp3, and Sp4. At the end of each loop, the collaborative iterative control unit strictly checks whether the preset focus completion condition has been met. These completion conditions preferably include at least one of the following combinations: First, the sharpness metric reaches stability and is close to the maximum value in several consecutive iterations, and its growth rate is less than a preset small threshold; Second, the continuous iterative update amount of the two-dimensional target representation, that is, the parameter change amplitude of the representation model, is less than another preset stability threshold in several consecutive iterations, indicating that the system's understanding of the two-dimensional shape and features of the target has tended to be stable; Third, the total number of iterations reaches the preset upper limit to prevent infinite loops. Once any one or a combination of conditions is met, the iterative process terminates, the collaborative iterative control unit announces that the focus is completed, and the current focus depth and the finally optimized two-dimensional target representation are the results of this focus task. If the conditions are not met, continue to the next iteration loop.

[0027] When dealing with the situation where there are multiple potential focus targets in the processing scene, this camera focusing device shows higher intelligence. At this time, the two-dimensional target representation unit independently generates and maintains the two-dimensional target representation for each identified potential focus target. The collaborative iterative control unit can then perform the above iterative enhancement process of Sp1 to Sp5 for each target representation in parallel or in a time-slice rotation manner. At the same time, inside the collaborative iterative control unit, according to a set of preset priority strategies, the target specified by the user through touch interaction has the highest priority, and is comprehensively sorted according to factors such as the central position, apparent size, motion state, convergence speed of its two-dimensional target representation, and identifiability score in the picture to dynamically select the main focus target to be optimized in the current iteration cycle. After all targets complete their respective iterative optimizations, a final and optimal focus target is selected.

[0028] The overall working mode of this camera focusing device and its focusing method specifically include the following steps: First, continuously capture the dynamic image data of the scene through the image acquisition unit; Second, the two-dimensional target representation unit actively generates and continuously optimizes the two-dimensional target representation of the potential focus target based on the captured image data in subsequent iterations; Then, the focus drive unit precisely adjusts the focus depth of the camera in strict accordance with the first control instruction issued by the collaborative iterative control unit based on the current two-dimensional target representation; Then, the image response characteristic analysis unit carefully analyzes the image characteristics at the new depth of focus, especially calculates the sharpness metric, and evaluates the feature attributes in the two-dimensional target representation and their specific presentation states in the current image; Subsequently, the collaborative iteration control unit intelligently updates the two-dimensional target representation based on the analysis results of the presentation states of the feature attribute sets feedback by the image response characteristic analysis unit, making it closer to the real target situation; Then again, the collaborative iteration control unit comprehensively considers the updated two-dimensional target representation and the current sharpness metric, generates a second control instruction, and guides the focus drive unit to perform the next more accurate depth of focus adjustment; Finally, the system iteratively executes the process from image characteristic analysis to depth of focus readjustment in a loop until the preset focus completion condition is met, thereby achieving efficient and accurate autofocus.

[0029] In summary, the precise division of labor and efficient collaboration of each unit, especially the ingenious regulation of the bidirectional enhanced feedback between focus adjustment and two-dimensional target representation optimization by the collaborative iteration control unit, achieve a strong adaptability to dynamic complex scenes and excellent autofocus performance. Specific Embodiment 2: As Figures 1 to 4 shown, based on the content in the above specific embodiment, the following content is further disclosed: The image acquisition unit is the visual input source of the entire focusing device, and the image data stream captured by it mainly includes the following forms: Original Bayer format data: This is the unprocessed pixel data directly output from the image sensor, usually arranged in formats such as RGGB, BGGR, etc., and contains the original light intensity information; YUV color space data: After being preliminarily processed by the in-sensor image signal processor (ISP), it generates luminance (Y) and chrominance (U, V) component data for subsequent analysis; RGB color space data: The red, green, and blue channel data after being processed by the ISP, which is a common image representation form; These data constitute the dynamic image data stream output by the image acquisition unit and provide a basis for subsequent target recognition and sharpness evaluation.

[0031] The process of capturing image data is as follows: The image acquisition unit uses a high-resolution CMOS or CCD image sensor as the core component; the sensor converts the optical signal in the scene into an electrical signal through photoelectric conversion; continuously captures image frames at a preset frame rate (such as 30fps or 60fps) to form a dynamic image data stream; the captured digital image frames are transmitted in real time to the subsequent two-dimensional target characterization unit and image response characteristic analysis unit through a high-speed serial interface (such as MIPI CSI-2); the image acquisition unit can respond to the trigger signal of the collaborative iterative control unit and capture a single frame or a limited number of multiple frames of images at a specific moment to support a specific focusing strategy.

[0032] The working modes of the image acquisition unit mainly include the following two: Continuous capture mode: Continuously acquires an image data stream at a fixed frame rate (such as 30 frames per second or 60 frames per second), which is suitable for real-time monitoring or dynamic scene analysis; Trigger capture mode: After receiving the trigger signal from the collaborative iterative control unit, acquires a single frame or a specified number of image frames. This mode is suitable for scenarios that require precise control of the acquisition timing, specific analysis or optimization steps during the focusing process; The specific moment refers to that during the focusing process, the collaborative iterative control unit decides to acquire an image at a certain precise time point according to the current state and optimization requirements, including immediately acquiring an image after the focus is adjusted and the focus drive unit completes the focus depth adjustment for analyzing the characteristics of the adjusted image; after the target characterization is updated, the two-dimensional target characterization unit acquires an image after updating the target characterization to verify the effectiveness of the new characterization.

[0033] The two-dimensional target characterization unit is responsible for generating and optimizing the digital characterization of potential focusing targets on the two-dimensional image plane, that is, the X-Y plane, based on the real-time image data provided by the image acquisition unit. The process is divided into the following two steps: Initial generation: Identifies the preliminary focusing target area through the saliency analysis of the global image (including detecting the salient area) or user-specified input; uses image processing techniques (including edge detection, region segmentation) to extract the geometric contour parameters and texture features of the target to form an initial characterization; Iterative update: When the collaborative iterative control unit issues an update instruction, this unit dynamically adjusts the target characterization according to the change of the image data; adjusts the weight of the geometric contour parameters: enhances the influence of the clearer contour segments due to improved focusing (including adjusting through control points); adjusts the sensitivity threshold of the texture complexity description: optimizes the feature extraction parameters when the texture details are richer (including adjusting the sampling parameters of the local binary pattern LBP or the frequency selection of the Gabor filter).

[0034] Image representation is a digital description of an object on a two-dimensional plane, specifically including object position information: the geometric contour of the object is defined by a set of parameters, including the control point sequence of the active contour model and the boundary description of the parametric spline curve; feature attribute set: describes the recognizable state of the object, including the visual characteristics of texture and structure; The feature attribute set specifically includes: description of the texture complexity inside the contour: gray-level co-occurrence matrix statistical features: including contrast (reflecting the intensity of gray-level changes), energy (reflecting uniformity), entropy (reflecting complexity); structure and frequency-domain features: local binary pattern (LBP) histogram: quantifies the local structure of the texture; Gabor filter bank response energy distribution: extracts the multi-scale and multi-directional frequency-domain features of the object; These parameters together constitute the feature attribute set of the object representation and are continuously optimized during iteration.

[0035] The optical components of the focus drive unit mainly include a movable lens group, and the depth of focus of the camera is adjusted by changing the position of the lens group. The focus drive unit is the actuator for physical focus adjustment, and its adjustment process: receives precise digital control signals from the cooperative iteration control unit, including the encoded position value corresponding to the target depth of focus, the number of movement steps or pulse sequence, uses precision drive mechanisms such as voice coil motors (VCMs), stepper motors or piezoelectric motors to drive the lens group, drives the lens group to move to the specified position according to the control signal, and a Hall effect sensor that real-time monitors the current position of the lens group and feeds back the position information to the cooperative iteration control unit to form a closed-loop control. The cooperative iteration control unit fine-tunes the instruction according to the feedback information to ensure that the lens group accurately reaches the target position.

[0036] The image response characteristic analysis unit analyzes the images captured by the image acquisition unit in real time, and evaluates the image characteristics at the current depth of focus for the target area defined by the two-dimensional object representation unit; the analysis process includes: obtaining the current frame data from the image acquisition unit, extracting the image characteristics of the corresponding area according to the object representation of the two-dimensional object representation unit, and using multiple algorithms to quantify the clarity and the state of the feature attributes; The analysis metrics are the image clarity metric and the presentation state of the feature attribute set. Specifically, the image clarity metric: outputs a quantified clarity score for evaluating the focus quality; the presentation state of the feature attribute set: re-measures the feature attributes in the object representation, reflecting their performance in the current image; The image characteristics of the image representation include the image clarity metric and the presentation state of the feature attribute set, and the specific parameters are shown in Table 1 below: Table 1 Image characteristic parameters of the image representation analyzed by the image response characteristic analysis unit:

[0037] The presentation states of the feature attribute set include gradient intensity: measuring the sharpness of edges; spatial frequency energy distribution: analyzing the frequency domain distribution of details; statistical dispersion: evaluating the variance of pixel gray values or the entropy of the texture histogram, reflecting texture complexity; The sharpness metrics include multiple metric parameters, specifically: the Tenengrad function: calculating the sum of the image gradient magnitudes based on the Sobel operator, reflecting the overall sharpness; the variance of Laplacian: calculating the variance of the pixel values after the image is processed by the Laplacian operator, measuring the focus sharpness; the proportion of high-frequency energy: extracting the proportion of high-frequency component energy through Fourier transform or wavelet transform, indicating the richness of details.

[0038] The preset focus completion conditions in the collaborative iterative control unit include at least one of the following: The sharpness metric reaches a stable maximum value: the growth of the sharpness score in several consecutive iterations is less than a preset small threshold; The continuous iterative update amount of the two-dimensional target representation is less than a predetermined threshold: the change amplitude of the representation parameters is less than the stable threshold in multiple iterations; The number of iterations reaches the upper limit: setting the maximum number of iterations to avoid infinite loops.

[0039] If the analysis result in the collaborative iterative control unit indicates that the current focus depth enhances the discernibility of a certain specific feature attribute in the target representation, the update instruction will strengthen the significance of this specific feature attribute in the two-dimensional representation. Discernibility refers to the sharpness and distinguishability of the feature attribute in the current image. The evaluation criteria include gradient intensity: a high edge sharpness indicates an obvious feature; spatial frequency energy distribution: concentrated in the high frequency indicates rich details; statistical dispersion: a large variance of gray values or a high entropy indicates high texture complexity. Significance refers to the importance and influence of the feature attribute in the target representation. The evaluation criteria include weight and sensitivity threshold. A high feature weight indicates its high importance, and a low threshold indicates sensitivity to feature changes and easy optimization. If the analysis result shows that the discernibility of a certain feature is enhanced (such as an increase in gradient intensity), the update instruction will: increase the weight of this feature in the target representation. Reduce its sensitivity threshold to make it more easily detected and strengthened in subsequent iterations.

[0040] The prediction model of the collaborative iterative control unit is a decision-making algorithm based on historical data and the current state, used to guide the direction and step size of focus depth adjustment. By recording the sharpness metrics and the stability of the target representation in past iterations, analyzing trends, including the change rate and direction of the sharpness metrics, and the iterative update amount of the target representation parameters, to decide on adjustment strategies, including continuous improvement of sharpness and stable representation: adjust along the current direction and increase the step size; sharpness decreases or oscillates, or the representation is unstable: reduce the step size and make a conservative adjustment;

[0041] The depth-of-focus adjustment amount (amplitude and direction) for the next iteration; is the momentum coefficient, controlling the influence of historical adjustment amounts; is the depth-of-focus adjustment amount for the current iteration; is the learning rate, controlling the influence of the current gradient; is the gradient of the sharpness metric with respect to the depth of focus, representing sensitivity; is the sign of the sharpness change, +1 indicating an increase and -1 indicating a decrease; is the target representation stability index (such as the reciprocal of the change in representation parameters); Through this model, the collaborative iteration control unit intelligently adjusts the depth of focus to achieve efficient focusing. Specific Embodiment Three: As Figures 1 to 4 shown, based on the content in the above specific embodiments, the following content is further disclosed: Based on the above specific embodiments, the hardware component structure corresponding to the camera focusing device during actual use is further disclosed: The hardware components of the image acquisition unit are centered around a high-resolution CMOS or CCD image sensor, capturing scene optical signals through photoelectric conversion to generate image data in Bayer, YUV, or RGB formats, supporting high frame rates of 30 / 60fps and resolutions above 12MP; the lens module (F / 1.8, wide-angle / telephoto) focuses light onto the sensor surface, and some integrate OIS anti-shake functions to improve stability; the built-in image signal processor performs denoising, color interpolation, and white balance processing to optimize data quality. The MIPI CSI-2 interface controller enables high-speed transmission of image data to the two-dimensional target representation unit and the image response characteristic analysis unit. The clock and power management module provides stable power supply and clock signals to ensure efficient operation of the sensor. The hardware of this unit supports continuous or trigger capture modes to adapt to the requirements of dynamic scenes.

[0043] The two-dimensional target representation unit is centered around an embedded digital signal processor or AI accelerator, running saliency analysis, contour extraction, and texture analysis algorithms to generate and optimize target representations; high-performance memory stores image data and representation parameters, supporting fast read and write; FPGA or ASIC can be used to accelerate the calculation of gray-level co-occurrence matrices or Gabor filters to improve real-time performance. The data interface module receives the image data from the image acquisition unit and interacts with the collaborative iteration control unit. The hardware of this unit supports multi-target parallel processing, suitable for optimizing target representations in complex scenes.

[0044] The focus driving unit takes a movable lens group (3 - 5 pieces, with micron - level precision) and a driving mechanism as the core, including a voice - coil motor, a stepper motor or a piezoelectric motor, and adjusts the depth of focus according to the control signal; the Hall - effect sensor monitors the position of the lens group in real - time and feeds back to the cooperative iterative control unit to form a closed - loop control; the driving circuit converts the PWM or digital signal into a driving signal, and the microcontroller processes the instructions and coordinates the operation. The hardware of this unit ensures fast and accurate focus adjustment, suitable for a variety of applications from mobile devices to high - end cameras.

[0045] The image response characteristic analysis unit takes a digital signal processor or a GPU as the core, and executes clarity measurement and feature analysis algorithms. The memory caches the image data and intermediate results, and the data interface module receives the data from the image acquisition unit and transmits the analysis results to the cooperative iterative control unit. A dedicated accelerator can be used to accelerate frequency - domain analysis or statistical calculations; the hardware of this unit supports real - time analysis of the image characteristics of multiple target areas, ensuring efficient feedback to optimize the focusing process.

[0046] The cooperative iterative control unit takes a central processing unit or an embedded SoC as the core, runs a prediction model, coordinates each unit, and generates control instructions and update instructions. The memory stores historical data and algorithm parameters, and the real - time operating system manages task scheduling to ensure low - latency iteration. The data interface module supports interaction with each unit, receives target representations and analysis results, and sends instructions. The initialization module can integrate a low - power coprocessor to run saliency analysis or process user input. The hardware of this unit supports high real - time performance and multi - target optimization, driving the intelligent operation of the entire focusing system. Specific Embodiment Four: As Figures 1 to 4 shown, based on the content in the above - mentioned specific embodiments, the following content is further disclosed: To verify the feasibility of the entire technical solution, the following application case content is further disclosed: Case Illustration One: Precise Focus Tracking in Dynamic Pet Photography; A photography enthusiast tried to use a camera with a camera - locking device to photograph a Siamese cat playing in a theater; the Siamese cat's fur color has obvious focus colors, but its body posture changes rapidly, sometimes curling up and sometimes jumping, which poses significant challenges to traditional automatic close - range systems in continuous tracking and precise focusing. In appropriate light, the background includes household items such as sofas and coffee tables, which can have a certain connection with target recognition. The operation process is as follows: Initialization and Target Preliminary Omen: The front camera of the user's camera; when the system starts, the image acquisition unit captures a video stream at a precision of 60 frames per second. The initialization module controlled by the collaborative iteration unit receives the first few frames of images. Through the saliency analysis algorithm, it detects that there is a certain cat in the moving progress preview and the background in the picture as a potential target; this initial area information is transmitted to the two-dimensional target characterization unit; the two-dimensional target characterization unit then generates an initial target characterization of the camera , whose geometric parameters roughly outline the cat, but due to the rapid movement of the cat and possible out-of-focus of the body, it is not clear. The initially calculated pseudo internal texture complexity description is also relatively low; at this time, the image response feature analysis unit is based on Analyze the initial shape and give a Tenengrad value of 850 units, and the intuitiveness score of the target characterization of the body is 0.6, indicating that the target features are not yet clear; Iterative Goodness and Characterization Optimization (Sp1-Sp5 Loop): The first iteration (k = 1): The collaborative iteration control power is based on the unit , and the instruction is close to the focus drive unit for preliminary focus depth Adjustment, moving forward 5 microns; the image response feature analysis unit analyzes the new image at depth, and the feedback tends to slightly increase to 920 units. The wake-up feature of the key color of the cat's face shows a slight increase in amplitude in the state representation state analysis; the collaborative iteration control unit generates an update instruction according to this. After receiving it, the two-dimensional target characterization unit weights the individual parameters of the corresponding facial area and strengthens the texture description parameters related to the key color, generating the target characterization ; ; Corresponding to The update amount (stability) is 15%; The third iteration (k = 3): After two iterations, the focus depth is adjusted to , tending to rise to 1500 units. At depth, the symmetric image of the cat's body distribution, especially the regression curve, shows a significant increase in its gradient intensity in the evaluation of the image response characteristic analysis unit; the two-dimensional target characterization unit sharply optimizes The geometric shape parameters of, producing a body curve that more precisely fits the actual cat. At the same time, the texture attributes related to the cat's body are also strengthened; the updated target characterization The serializability score of is increased to 0.75, and the update amount compared with is 10%; The seventh iteration (k = 7): At this time, the cat suddenly appears. The prediction model of the collaborative iteration control unit combines the historical trend change trend (first rising continuously) and The stability (the update amount has been lower than 5% for two consecutive iterations), quickly determines the depth of focus has been optimally approximated, but the two-dimensional target representation unit quickly updates the position information of the cat on the XY plane and the simulation parameters caused by the pose change by analyzing the latest image frame, and generates ; The efficient iterative control mainly refers to the updated position information, indicating a small focus adjustment to the drive unit to , while maintaining tight tracking of the adjacent area; after this round of iteration, the obtained reaches 3200 units, the consistency score of reaches 0.88; Completion and results: After a total of 12 iterations, running for 180 milliseconds, the system meets the initial completion conditions; specifically, the subsequent steps are stable at around 3550 units (the gain of three consecutive iterations is less than 50 units), and the two-dimensional target representation has an update amount of less than 2% for two consecutive iterations; the final depth of focus is locked at , in the obtained photo display, even during the dynamic elephant process, the Siamese cat's face, especially the eyes, has obtained very high precision, the hair details are distinct, and the background shows a natural shallow depth-of-field blur, which fully demonstrates the relevant performance of this device in dynamic target tracking through the synergistic effect of depth-of-focus adjustment and two-dimensional target representation optimization.

[0048] Case illustration two: Detail capture of texture still-life macro photography; Scene overview: A plant photographer plans to take a macro close-up of a blooming orchid petal; the surface of the orchid petal has extremely fine and complex vein textures and delicate color restoration, but in the initial state, these are difficult to distinguish, and the petal edges may also be inconsistent with the background or other petal parts of similar tones; the shooting environment is an indoor still-life photography studio with a smooth surface, but the challenge is how to identify details from the blurred initial state and sharpen the fine structure of the petals. The process of using this solution is as follows: Region initialization and target preliminary preset: The photographer approaches the orchid with the lens and frames the main rose petals for shooting; the image acquisition unit starts to transmit image data; the initialization module of the collaborative iterative control unit analyzes the global preset and color distribution of the image, and combines the center area of the picture that the user may specify by touch to identify the petal main body as the potential expected target; the two-dimensional target representation unit generates the initialization representation accordingly , whose geometric parameters roughly delimit the visible area of the petals. However, due to severe initial defocus, the description of the internal texture complexity is very low, and it is almost impossible to effectively characterize the petal veins; the initial image of the image response characteristic analysis unit, based on the spatial frequency index, walks 0.25 (normalized value), and the target characterization The reconfigurability score of is 0.4; Iterative goodness and characterization optimization (Sp1-Sp5 cycle): First iteration (k = 1): The good iteration control unit, based on the fuzzy , instructs the focus drive unit to perform a tentative focus depth adjustment to move the focal plane forward by 50 microns; the image response characteristic analysis unit detects a slight improvement to 0.30 in the new image; more importantly, although the overall is still blurry, at the depth, the relay network of several main veins on the petals is slightly enhanced in other areas, and this information is captured by the feature attribute set extraction state analysis; the fine iteration control unit generates an update instruction based on this. After receiving it, the two-dimensional target characterization unit begins to in the texture complexity of assign attention weights to the Gabor filter response parameters related to the preliminary description of the vein direction and position, and the target generation characterization ; The update amount of and is 20%, mainly reflected in the preliminary construction of the texture feature attributes; Fifth iteration (k = 5): The focus depth has been adjusted several times to reach ; at this time, it is about to approach 0.65; at the depth, the previously concerned main vein grid is relatively neat, and finer secondary veins begin to emerge around it; the feature attributes and presentation state analysis of the image response characteristic analysis unit clearly point out these new controlled signal structures; the update instruction of the collaborative iteration control unit guides the two-dimensional target characterization unit to not only further strengthen the geometric synthesis parameters of the main veins in (such as making it more precise through the control point coefficients), but also greatly expands the description of the texture complexity, the characterization parameters of these newly emerging fine veins, and increases the sensitivity threshold of the corresponding parameters; the updated target characterization is very close to the optimal; in Sp3, the state display of the feature attribute set fed back by the image response characteristic analysis unit shows that almost all vein features defined in it show extremely high gradient intensity and braking at the current focus depth; the update instruction mainly makes minor optimizations to generate which has an update amount diagram of 1.5% with indicating that the two-dimensional target representation is highly stable; Completion and results: After a total of 11 iterations, lasting about 250 milliseconds, the system determines that the completion condition is met; the final restraint is stabilized at 0.95 (normalized value), and the iterative update amount of the two-dimensional target representation is less than 1%; the final focus depth is locked at The photographer views the shooting result, and the photo perfectly confirms the contour world of the orchid: every delicate vein is clearly visible, and the main vein to the almost invisible capillary structure is accurately reproduced, with natural and soft color transitions, fully reflecting the enhancement and force of the petals; in this case, through focus adjustment and iterative enhancement of two-dimensional feature representation, it can effectively handle the contour problems of complex textures and initially low-clarity targets, achieving extreme capture of tiny details.

[0049] Case illustration three: Portrait face priority in low-light environment; Scene overview: In a park at dusk, the user attempts to take a natural light portrait photo of a friend; at this time, the overall light is dim, and elements such as measurements and benches in the background are visible but lack details. The friend's face, as the main target, has low visibility of its facial features in low light. Traditional shooting systems are prone to problems such as hunting or focusing on a higher background in such scenarios. The steps of running this solution are as follows: Initialization and preliminary target representation: The user composes the picture, and the friend's face is located in the center of the picture; the image acquisition unit works with the above ISO parameter of 1600 to increase the picture brightness in low light, but it may also introduce noise points; the initialization module of the collaborative iterative control unit combines a face detection algorithm, as a special case of external specification or rapid global analysis, and initially determines and locks the friend's area as a high-priority potential target; the two-dimensional target representation unit generates an initial facial target representation whose geometric re-participation number roughly determines the eyebrow area, and the feature attribute set initially records the relative positions and approximate gray-scale features of key areas such as eyes, nose, and mouth; due to insufficient light and initial defocus, the image response feature analysis unit gives an overall score, based on the Laplacian operator brightness value, with a brightness of 35 (the score indicates blurriness), and the quantifiable score of is 0.55; Second iteration (k=2): The collaborative iteration control unit instructs the drive unit to adjust the focal depth to At this depth, the image response characteristic analysis unit feedback signal, more importantly, the characteristic attribute set presents a state analysis indication, The local awakening of the area represented in the image has slightly enhanced details, although it is still darker overall; the update instructions of the collaborative iteration control unit optimize the two-dimensional target representation unit according to this alarm The custom parameters and feature attribute weights of the middle area increase the precision of the eye control points and enhance the importance of grayscale features related to the iris and pupil in the representation. ; and The update volume is 12%; Sixth iteration (k=6): focal depth The overall score is increased to 120; at this time, The facial features represented are emphasized, especially the edges of the eyes and lips, whose intensity is significantly enhanced under the current focus; the two-dimensional target representation unit, under the instruction of the collaborative iteration control unit, greatly optimizes The geometric parameters of these key facial features are reset, which leads to a more realistic shape of the facial features, and the texture complexity descriptions related to these features (such as the direction of eyebrows and the texture complexity of teeth) are also effectively constructed. middle; The olfactory score of The update amount is 8%; and the priority strategy of the collaborative iteration control unit is well prepared to ensure that the optimization of the representation takes precedence over other potential goals that may appear; Ninth iteration (k=9): The inflection point reaches 280, target prediction The key details of the subject can be drawn quite accurately; the prediction model of the collaborative iterative control unit is based on the continuous and steady rise and The focus drive unit performs a slightly larger focus before the image response characteristics analysis unit feedback. Under the , the reflective highlights in the eye area are extremely sharp, and stress parameters of the facial skin begin to appear; arrive The update amount is 3%, which is mainly focused on the characterization parameters of these focal points; Completion and results: The system performed 10 iterations in total, with a total running time of about 320 milliseconds (the processing time may be slightly longer in low light); finally, it stabilized at around 310, and the two-dimensional target representation The iterative update amount is less than 1.5%; the focus depth is locked at ; In the finally taken photo, although the overall ambient light is dim, the face of the friend, especially the eyes, is accurately captured, while the images and benches in the background are blurred naturally because the focus is on the portrait subject, highlighting the subject. This case shows that, under low-light conditions, this camera imaging device can still effectively and precisely identify and align with the human face through its unique two-dimensional feedback control mechanism combined with a priority strategy, overcoming the deficiencies of traditional docking methods in similar scenarios.

[0050] It should be noted that, in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising a reference structure" does not exclude the existence of additional identical elements in the process, method, article or device comprising the element.

[0051] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A camera focusing device with two-dimensional feedback control capability, characterized in that: The camera focusing device includes an image acquisition unit, a two-dimensional target representation unit, a focus driving unit, an image response characteristic analysis unit, and a collaborative iteration control unit, and further includes the following contents: An image acquisition unit, used for capturing an image data stream of a scene; A two-dimensional target representation unit is connected to the image acquisition unit and is used to generate and iteratively update target representations of one or more potential focus targets on the two-dimensional image plane according to the received image data, and the target representation includes the position information of the target and a feature attribute set describing its identifiability; A focus drive unit, used for adjusting the focus depth of the optical component according to the control instruction; An image response characteristic analysis unit, connected to the image acquisition unit, for analyzing the presentation state of the image characteristics and feature attribute set corresponding to the target representation in the current image at the current focus depth, wherein the image characteristics include at least one clarity metric; The collaborative iteration control unit is connected to the two-dimensional target representation unit, the focus driving unit and the image response characteristic analysis unit, and is used to perform the following steps: Sp1: Based on the target representation of the current target output by the two-dimensional target representation unit, generate a first control instruction to the focus driving unit to adjust the focus depth; Sp2: receiving the image characteristic analysis result fed back by the image response characteristic analysis unit at the focal depth adjusted by Sp1; Sp3: Based on the presentation status of the feature attribute set in the image characteristic analysis result, generate an update instruction to the two-dimensional target representation unit to adjust and optimize its two-dimensional target representation of one or more potential focus targets; Sp4: Based on the two-dimensional target representation updated and optimized in Sp3 and combined with the clarity measurement, a second control instruction is generated to the focus drive unit for subsequent focus depth adjustment; Sp5: Loop through Sp2 to Sp4, iteratively enhancing the optimization effect of focus depth adjustment on two-dimensional target representation and the precise guidance effect of the optimized two-dimensional target representation on subsequent focus depth adjustment until the preset focus completion condition is met.

2. A camera focusing device with two-dimensional feedback control capability according to claim 1, characterized in that: The target representation generated by the two-dimensional target representation unit includes the geometric contour parameters of the target and the texture complexity description inside the contour. When the two-dimensional target representation unit receives an update instruction, it adjusts and optimizes the two-dimensional representation by adjusting the weights of the geometric contour parameters and the sensitivity threshold of the texture complexity description.

3. A camera focusing device with two-dimensional feedback control capability according to claim 1, characterized in that: The presentation status of the characteristic attribute set output by the image response characteristic analysis unit includes the gradient intensity, spatial frequency energy distribution and statistical discreteness of the characteristic attribute at the current focal depth.

4. A camera focusing device with two-dimensional feedback control capability according to claim 1, characterized in that: In Sp3, if the analysis result indicates that the current focus depth enhances the recognizability of a specific feature attribute in the target representation, the collaborative iteration control unit will update the instruction to enhance the significance of the specific feature attribute in the two-dimensional representation.

5. The camera focusing device with two-dimensional feedback control capability according to claim 1, characterized in that: The collaborative iterative control unit uses a prediction model based on a combination of a clarity metric change trend and a two-dimensional target representation stability in Sp4 to determine the magnitude and direction of subsequent focus depth adjustment.

6. A camera focusing device with two-dimensional feedback control capability according to claim 1, characterized in that: The focusing completion condition preset by the collaborative iteration control unit includes at least one of the following: the clarity metric reaches a stable maximum value, the continuous iterative update amount of the two-dimensional target representation is less than a predetermined threshold, and the number of iterations reaches an upper limit.

7. A camera focusing device with two-dimensional feedback control capability according to claim 1, characterized in that: When there are multiple potential focus targets, the two-dimensional target representation unit maintains an independent two-dimensional target representation for each potential focus target, and the collaborative iteration control unit executes the iterative enhancement process for each target representation in parallel, and selects the final focus target according to a preset priority strategy.

8. The camera focusing device with two-dimensional feedback control capability according to claim 1, characterized in that: The collaborative iteration control unit includes an initialization module, which is used to provide an initial two-dimensional target representation for the two-dimensional target representation unit by quickly analyzing the global image and receiving external designation before the first iteration.

9. The camera focusing device with two-dimensional feedback control capability according to claim 1, characterized in that: The camera focusing device comprises the following steps: Sp1: Capture image data through the image acquisition unit; Sp2: The 2D target representation unit generates and iteratively updates the 2D target representation of the potential focus target; Sp3: The focus driving unit adjusts the focus depth according to a first control instruction of the collaborative iteration control unit, where the first control instruction is based on the current two-dimensional target representation; Sp4: The image response characteristic analysis unit analyzes the image characteristics after adjusting the focal depth, including the clarity measurement and the presentation status of the feature attribute set in the two-dimensional representation; Sp5: The collaborative iteration control unit updates the two-dimensional target representation based on the presentation status of the feature attribute set in the image characteristic analysis result; Sp6: The collaborative iteration control unit generates a second control instruction to the focus drive unit for subsequent focus depth adjustment based on the updated two-dimensional target representation and clarity measurement; Sp7: Execute Sp4 to Sp6 in a loop until the preset focus completion condition is met.

Images