A camera focusing device with two-dimensional feedback control ability

Abstract

The present invention provides a camera focusing device with two-dimensional feedback control ability, which relates to the technical field of camera focusing. 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. The two-dimensional target characterization unit is used to generate a target characterization including geometric contours and texture complexity, and the image response characteristic analysis unit is used to evaluate 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 characterization parameters. This two-way feedback mechanism enables the system to adaptively cope with complex scenarios such as light changes, multi-target interference, or target movement, significantly improving the focusing speed and accuracy, deeply integrating the dynamic optimization of target characterization with the focus adjustment, and providing a new technical path for high-performance autofocus in dynamic complex scenarios.

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 Technique

[0002] According to a camera module autofocus control method and device disclosed in Chinese Patent Publication No. "CN114598813A", it includes: obtaining a preset calibration distance, and the number of the preset calibration distances is at least three; performing focus calibration on the preset calibration distances, including performing focus calibration tests using two of the calibration distances to obtain the actual motor stroke encodings respectively corresponding to the two calibration distances; 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 with the method of using a machine tool for focus calibration in all cases, 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:

[0004] 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 such as slow focusing speed and insufficient accuracy. The limitation of the traditional method lies in the lack of dynamic understanding and adaptive adjustment ability of target features, resulting in the focusing process being easily interfered by the environment and difficult to quickly converge to the best focus.

[0005] 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

[0006] 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.

[0007] To achieve the above object, 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 contents:

[0008] An image acquisition unit for capturing an image data stream of a scene;

[0009] A two-dimensional target characterization unit, connected to the image acquisition unit, for generating and iteratively updating the target characterization of one or more potential focus targets on a two-dimensional image plane according to the received image data, and the target characterization includes the position information of the target and a set of feature attributes describing its recognizability;

[0010] A focus drive unit for adjusting the depth of focus of an optical component according to a control instruction;

[0011] 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 feature attributes corresponding to the target characterization in the current image at the current depth of focus, and the image characteristics include at least one sharpness metric;

[0012] A collaborative iterative control unit, connected to the two-dimensional target characterization unit, the focus drive unit, and the image response characteristic analysis unit, for performing the following steps:

[0013] Sp1: Generating a first control instruction to the focus drive unit to adjust the depth of focus based on the target characterization of the current target output by the two-dimensional target characterization unit;

[0014] Sp2: Receiving the image characteristic analysis result at the depth of focus adjusted in Sp1 feedback by the image response characteristic analysis unit;

[0015] Sp3: Generating an update instruction to the two-dimensional target characterization unit based on the presentation state of the set of feature attributes in the image characteristic analysis result to adjust and optimize its two-dimensional target characterization of one or more potential focus targets;

[0016] Sp4: Generating a second control instruction to the focus drive unit based on the updated and optimized two-dimensional target characterization in Sp3 and combining the sharpness metric for subsequent depth of focus adjustment;

[0017] Sp5: Repeatedly executing Sp2 to Sp4, until a 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 characterization and the precise guiding effect of the optimized two-dimensional target characterization on the subsequent depth of focus adjustment.

[0018] Preferably, the target characterization generated by the two-dimensional target characterization unit includes the geometric contour parameters of the target and the description of the texture complexity inside the contour. When the two-dimensional target characterization unit receives an 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 characterization.

[0019] Preferably, the presentation states of the feature attribute set output by the image response characteristic analysis unit include the gradient intensity of the feature attribute at the current focus depth, the spatial frequency energy distribution, and the statistical dispersion.

[0020] Preferably, in Sp3, if the analysis result indicates that the current focus depth enhances the identifiability 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.

[0021] Preferably, in Sp4, the collaborative iterative control unit uses a prediction model based on the combination of the change trend of the sharpness metric and the stability of the two-dimensional target representation to determine the amplitude and direction of the subsequent focus depth adjustment.

[0022] 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.

[0023] 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.

[0024] Preferably, the collaborative iterative 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 specifications before the first iteration.

[0025] Preferably, the camera focusing device includes the following steps:

[0026] Sp1: Capture image data through the image acquisition unit;

[0027] Sp2: The two-dimensional target representation unit generates and iteratively updates the two-dimensional target representation of the potential focus target;

[0028] Sp3: The focus drive unit adjusts the focus depth according to the first control instruction of the collaborative iterative control unit, and this first control instruction is based on the current two-dimensional target representation;

[0029] 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;

[0030] 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;

[0031] Sp6: The collaborative iterative control unit generates a second control instruction based on the updated two-dimensional target representation and sharpness metric, and gives it to the focus drive unit for subsequent focus depth adjustment;

[0032] Sp7: Loop through Sp4 to Sp6 until the preset focus completion condition is met.

[0033] The present invention provides a camera focusing device with two-dimensional feedback control ability. It has the following beneficial effects:

[0034] 1. The present invention introduces a two-dimensional feedback mechanism, tightly couples focus depth adjustment with target representation optimization, forms a unique iterative enhancement closed-loop system, uses the two-dimensional target representation unit to generate a target representation including geometric contours and texture complexity, and evaluates the presentation state of the feature attribute set in real time through the image response characteristic analysis unit, 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, deeply integrating the dynamic optimization of the target representation with the focus adjustment, breaking through the limitations of traditional focusing methods, and providing a new technical path for high-performance autofocus in dynamic complex scenarios.

[0035] 2. The present invention adopts the parallel optimization of multiple potential focusing targets and the preset priority strategy, greatly improving 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 multiple people scenarios in surveillance systems or multi-tissue imaging in medical images. The two-dimensional target representation unit independently maintains representations for each target, the collaborative iterative control unit executes 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, combines multi-target parallel optimization with intelligent decision-making, breaks through the limitation of single-target focusing, and provides an efficient solution for intelligent vision applications in multi-scene and multi-target scenarios. Description of the Drawings

[0036] Figure 1 It is the composition structure diagram of the camera focusing device of the present invention;

[0037] Figure 2 It is the control step diagram of the collaborative iterative control unit of the present invention;

[0038] Figure 3 It is the operation flow diagram of the camera focusing device of the present invention;

[0039] Figure 4 It is the operation relationship diagram of the camera focusing device of the present invention. Detailed implementation mode

[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to 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 the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention. Specific embodiment 1:

[0042] As Figures 1 to 4 shown, 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, a collaborative iterative control unit, and further includes the following:

[0043] 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 image signal processor built into 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.

[0044] 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, i.e., the X-Y plane, it intelligently generates and continuously optimizes in subsequent iterative processes a 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 an active contour model and the control point sequence of a 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 improved focusing, 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 a representation stability metric, will be output as key information to the collaborative iterative control unit and the image response characteristic analysis unit.

[0045] 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 a voice coil motor VCM, a stepper motor, and a piezoelectric motor, and other precision drive mechanisms. These mechanisms can convert the digital control signal 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, which is used to monitor the current position of the lens group in real time 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.

[0046] The image response characteristic analysis unit is closely connected to the image acquisition unit, and it 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 focal depth currently 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 gradient, 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 set of feature attributes defined in the target representation at the current image and the current focal depth. For each feature attribute described by the two-dimensional target representation unit, including the edge sharpness of a specific segment of the contour and the richness of texture details in a specific area, 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 set of feature attributes at the current focal depth 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 set of feature attributes, 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 decision-making.

[0047] 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 instructs the image acquisition unit to capture an initial scene image, and by performing a quick saliency analysis on the global image or directly receiving user-specified input from the outside, it identifies an initial potential focusing target area. 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:

[0048] Sp1: Generate a first control instruction to the focus drive unit to adjust the focal depth 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 iteration 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 predetermined trajectory. In subsequent iterations, that is when, the collaborative iteration control unit will use its internal prediction model to determine the focus adjustment strategy; the prediction model preferably combines the historical change trend of the sharpness metric with the stability information of the two-dimensional target representation itself; if the historical data shows that the current focus adjustment direction can continuously improve sharpness, and the target representation relative to the representation of 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 the sharpness decreases, oscillates 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 the stabilization and optimization of 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.

[0049] Sp2: Receive the image characteristic analysis result at the focus depth adjusted in Sp1, feedback 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, making the camera reach the new focus depth after that, 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 target representation determined by the two-dimensional target representation unit in the previous iteration corresponding image region, calculate its sharpness metric and the specific presentation state 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.

[0050] Sp3: Generate an update instruction to the two-dimensional target representation unit based on the presentation state of the feature attribute set in the image characteristic analysis result to adjust and optimize its two-dimensional target representation of one or more potential focusing targets. After receiving the image characteristic analysis result feedback by Sp2, the collaborative iteration control unit will carefully compare the data on the presentation state 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 result clearly indicates the current new focus depth Significantly enhances the target representation The identifiability of a certain specific feature attribute in, the gradient intensity of a certain edge segment of the target contour is greatly enhanced, and the texture detail contrast of a certain area is more distinct. Then the update instruction generated by the collaborative iteration control unit will instruct 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, making it more inclined to the newly enhanced edge by adjusting the weights of the control points defining the geometric contour, and modifying the parameters of the texture descriptor to better capture the fine texture revealed due to the focus improvement. After the two-dimensional target representation unit executes this update instruction, the target representation inside it evolves into a better .

[0051] Sp4: Based on the updated and optimized two-dimensional target representation in Sp3, and combined with the clarity metric, generate the 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 iteration control unit combines the clarity 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.

[0052] Sp5: Loop through Sp2 to Sp4, and through the iterative enhancement between the optimization effect of the two-dimensional target representation by focal depth adjustment and the precise guidance effect of the optimized two-dimensional target representation on subsequent focal depth adjustment until the preset focus completion condition is met. Increment the iteration counter k inside the collaborative iterative control unit. 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 approaches 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 range 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 combination of conditions is met, the iterative process terminates, the collaborative iterative control unit announces that the focus is completed, and the current focal depth and the finally optimized two-dimensional target representation are the results of this focusing task. If the conditions are not met, continue to the next iteration loop.

[0053] When dealing with the situation where there are multiple potential focusing 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 focusing target. The collaborative iterative control unit can execute the iterative enhancement process of Sp1 to Sp5 described above 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 focusing target to be optimized in the current iteration cycle. After all targets complete their respective iterative optimizations, select a final and optimal focusing target.

[0054] The overall working mode of this camera focusing device and its focusing method specifically include the following steps:

[0055] First, continuously capture the dynamic image data of the scene through the image acquisition unit;

[0056] Second, the two-dimensional target representation unit actively generates and continuously optimizes the two-dimensional target representation of potential focusing targets based on the captured image data in subsequent iterations;

[0057] Next, the focus drive unit precisely adjusts the focal depth of the camera in strict accordance with the first control instruction issued by the collaborative iteration control unit based on the current two-dimensional target representation;

[0058] Then, the image response characteristic analysis unit carefully analyzes the image characteristics at the new focal depth, 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;

[0059] Subsequently, the collaborative iteration control unit intelligently updates the two-dimensional target representation according to the analysis results of the feature attribute set presentation state feedback by the image response characteristic analysis unit to make it closer to the real target situation;

[0060] Then again, the collaborative iteration control unit comprehensively considers the updated two-dimensional target representation and the current sharpness metric, generates the second control instruction, and guides the focus drive unit to perform a more precise focal depth adjustment in the next round;

[0061] Finally, the system repeatedly executes the iterative process from image characteristic analysis to focal depth readjustment until the preset focus completion condition is met, thereby achieving efficient and precise autofocus.

[0062] In summary, the precise division of labor and efficient collaboration of each unit, especially the ingenious regulation of the two-way 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:

[0064] As Figures 1 to 4 shown, based on the content in the above specific embodiment, the following content is further disclosed:

[0065] The image acquisition unit is the visual input source of the entire focusing device, and the image data stream it captures mainly includes the following forms:

[0066] 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., containing the original light intensity information;

[0067] YUV color space data: After being preliminarily processed by the image signal processor (ISP) built into the sensor, it generates luminance (Y) and chrominance (U, V) component data for subsequent analysis;

[0068] RGB color space data: The red, green, and blue three-channel data after being processed by the ISP, which is a common image representation form;

[0069] These data constitute the dynamic image data stream output by the image acquisition unit, providing a basis for subsequent target recognition and clarity evaluation.

[0070] 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 cooperative 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.

[0071] The working modes of the image acquisition unit mainly include the following two:

[0072] Continuous capture mode: Continuously acquire the 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;

[0073] Trigger capture mode: After receiving the trigger signal from the cooperative iterative control unit, acquire 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;

[0074] The specific moment refers to during the focusing process, the cooperative 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 adjustment and after the focus drive unit completes the focus depth adjustment for analyzing the characteristics of the adjusted image; after the target characterization is updated, after the two-dimensional target characterization unit updates the target characterization, acquire an image to verify the effectiveness of the new characterization.

[0075] 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:

[0076] Initial generation: Identify the preliminary focusing target area through the saliency analysis of the global image (including detecting the salient area) or user-specified input; use image processing techniques (including edge detection, region segmentation) to extract the geometric contour parameters and texture features of the target to form the initial characterization;

[0077] Iterative update: When the collaborative iterative control unit issues an update instruction, this unit dynamically adjusts the target representation according to the changes in the image data; Adjust the weights of the geometric contour parameters: Enhance the influence of the contour segments that are clearer due to improved focus (including adjustment through control points); Adjust the sensitivity threshold described by the texture complexity: When the texture details are richer, optimize the feature extraction parameters (including adjusting the sampling parameters of the local binary pattern (LBP) or the frequency selection of the Gabor filter).

[0078] Image representation is the digital description of the target on the two-dimensional plane, specifically including the target position information: The geometric contour of the target 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 target, including the visual characteristics of texture and structure.

[0079] The feature attribute set specifically includes: Texture complexity description inside the contour: Gray-level co-occurrence matrix statistical features: including contrast (reflecting the intensity of gray-scale 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 target.

[0080] These parameters together constitute the feature attribute set of the target representation and are continuously optimized during iteration.

[0081] The optical components of the focus drive unit mainly include a movable lens group, which adjusts the depth of focus of the camera 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 collaborative iterative control unit, including the encoded position value corresponding to the target depth of focus, the number of moving steps, or the 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 the Hall effect sensor monitors the current position of the lens group in real time and feeds back the position information to the collaborative iterative control unit to form a closed-loop control. The collaborative iterative control unit fine-tunes the instruction according to the feedback information to ensure that the lens group accurately reaches the target position.

[0082] 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 target 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 target representation of the two-dimensional target representation unit, and using multiple algorithms to quantify the clarity and the state of the feature attributes.

[0083] The analyzed metrics are the image sharpness metric and the presentation state of the feature attribute set. Specifically, the image sharpness metric: outputs a quantified sharpness score for evaluating the focus quality; the presentation state of the feature attribute set: re-measures the feature attributes in the target representation, reflecting their performance in the current image;

[0084] The image characteristics of the image representation include the sharpness metric and the presentation state of the feature attribute set. The specific parameters are shown in Table 1 below:

[0085] Table 1 Image characteristic parameters of the image representation analyzed by the image response characteristic analysis unit:

[0086]

[0087] The presentation state of the feature attribute set includes 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 the texture complexity;

[0088] The sharpness metric includes multiple metric parameters. Specifically: the Tenengrad function: calculates the sum of the image gradient magnitudes based on the Sobel operator, reflecting the overall sharpness; the variance of Laplacian: calculates 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: extracts the proportion of high-frequency component energy through Fourier transform or wavelet transform, indicating the richness of details.

[0089] The preset focus completion conditions of the cooperative iteration control unit include at least one of the following:

[0090] 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;

[0091] The continuous iteration update amount of the two-dimensional target representation is less than a predetermined threshold: the change range of the representation parameters is less than the stable threshold in multiple iterations;

[0092] The number of iterations reaches the upper limit: sets the maximum number of iterations to avoid infinite loops.

[0093] If the analysis result in the collaborative iteration control unit indicates that the current focus depth enhances the distinguishability 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. Distinguishability refers to the clarity and ease of differentiation 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 or high entropy of the gray value 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 great importance, and a low threshold indicates sensitivity to feature changes and ease of optimization. If the analysis result shows an increase in the distinguishability of a certain feature (such as an increase in gradient intensity), the update instruction will: increase the weight of this feature in the target representation. Lower its sensitivity threshold to make it more easily detected and strengthened in subsequent iterations.

[0094] The prediction model of the collaborative iteration 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 clarity metric and target representation stability in past iterations, analyzing trends, including the change rate and direction of the clarity metric, and the iterative update amount of target representation parameters, and deciding on adjustment strategies, including continuous improvement in clarity and stable representation: adjust along the current direction and increase the step size; decrease in clarity or oscillation, or unstable representation: reduce the step size and make a conservative adjustment;

[0095]

[0096] is the focus depth adjustment amount (amplitude and direction) for the next iteration;

[0097] is the momentum coefficient, controlling the influence of the historical adjustment amount;

[0098] is the focus depth adjustment amount for the current iteration;

[0099] is the learning rate, controlling the influence of the current gradient;

[0100] is the gradient of the clarity metric with respect to the focus depth, indicating sensitivity;

[0101] is the sign of the clarity change, +1 indicates an increase, -1 indicates a decrease;

[0102] is the target representation stability index (such as the reciprocal of the change in representation parameters);

[0103] Through this model, the collaborative iterative control unit intelligently adjusts the depth of focus to achieve efficient focusing. Specific Embodiment Three:

[0105] As Figures 1 to 4 shown, based on the content in the above specific embodiments, the following content is further disclosed:

[0106] Based on the above specific embodiments, the hardware component structure corresponding to the camera focusing device during actual use is further disclosed:

[0107] The hardware components of the image acquisition unit are centered around a high-resolution CMOS or CCD image sensor, which captures scene light signals through photoelectric conversion to generate image data in Bayer, YUV, or RGB format, supports high frame rates of 30 / 60fps, and has a resolution of over 12MP; the lens module (F / 1.8, wide-angle / telephoto) focuses light onto the sensor surface, and some integrate OIS anti-shake function 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 characterization unit and the image response characteristic analysis unit. The clock and power management module provides stable power supply and clock signals to ensure the efficient operation of the sensor. The hardware of this unit supports continuous or trigger capture modes to adapt to the requirements of dynamic scenes.

[0108] The two-dimensional target characterization unit is centered around an embedded digital signal processor or an AI accelerator, runs saliency analysis, contour extraction, and texture analysis algorithms to generate and optimize target characterizations; high-performance memory stores image data and characterization parameters to support fast read and write; FPGA or ASIC can be used to accelerate the calculation of gray-level co-occurrence matrix or Gabor filter to improve real-time performance. The data interface module receives the image data from the image acquisition unit and interacts with the collaborative iterative control unit. The hardware of this unit supports multi-target parallel processing and is suitable for optimizing target characterizations in complex scenes.

[0109] The focusing drive unit is centered around a movable lens group (3 - 5 pieces, micron-level precision) and a drive mechanism, including a voice coil motor, a stepper motor, or a piezoelectric motor, which 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 collaborative iterative control unit to form a closed-loop control; the drive circuit converts PWM or digital signals into drive signals, and the microcontroller processes instructions and coordinates operations. The hardware of this unit ensures fast and accurate focus adjustment and is suitable for a variety of applications from mobile devices to high-end cameras.

[0110] The image response characteristic analysis unit takes a digital signal processor or GPU as the core and executes sharpness measurement and feature analysis algorithms. The memory caches image data and intermediate results. The data interface module receives data from the image acquisition unit and transmits the analysis results to the collaborative 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 regions to ensure efficient feedback for optimizing the focusing process.

[0111] The collaborative iterative control unit takes a central processor or 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 co-processor 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 4:

[0113] As Figures 1 to 4 shown, based on the content in the above specific embodiments, the following content is further disclosed:

[0114] To verify the feasibility of the entire technical solution, the following application case content is further disclosed:

[0115] Case Illustration 1: Precise tracking focus in dynamic pet photography;

[0116] A photography enthusiast tried to use a camera with a camera locking device to photograph a Siamese cat playing on a theater stage; the Siamese cat's fur color has obvious focal colors, but its body posture changes rapidly, sometimes curling up and sometimes jumping, which poses a significant challenge 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:

[0117] Initialization and preliminary omen of the target: The user's camera front 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 iterative unit receives the first few frames of images. Through the saliency analysis algorithm, it detects that there is a certain cat in the picture with a moving progress preview different from the background as a potential target; this initial area information is transmitted to the two-dimensional target representation unit; the two-dimensional target representation unit then generates an initial target representation of the camera , whose geometric parameters roughly outline the cat's, but due to the cat's rapid movement and possible out-of-focus of its body, it is not blurred. 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, giving a Tenengrad value of 850 units, the target representation The ontology of has an intuitive score of 0.6, indicating that the target features are not yet clear;

[0118] Good iteration and representation optimization (Sp1 - Sp5 loop):

[0119] First iteration (k = 1): The collaborative iteration control power supply is based on the unit , instructing the proximity focus drive unit to perform a preliminary depth of focus adjustment, moving forward 5 microns; The image response feature analysis unit analyzes the new image at the depth, and the feedback tends to slightly increase to 920 units. The wake - up feature of the cat's face key color shows a small increase in amplitude in the state representation status analysis; The collaborative iteration control unit generates an update instruction based on this. After receiving it, the two - dimensional target representation unit weights the individual parameters of the corresponding facial area and strengthens the texture description parameters related to the key color, generating the target representation ; Corresponding to The update amount (stability) is 15%;

[0120] Third iteration (k = 3): After two iterations, the depth of focus is adjusted to , tending to rise to 1500 units. At the 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 representation unit sharply optimizes the geometric shape parameters, producing a more accurate fit to the actual body curve of the cat. At the same time, the texture attributes related to the cat's body are also strengthened; The updated target representation has its serializability score increased to 0.75, and the update amount for is 10%;

[0121] 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 of (the update amount has been less than 5% for two consecutive iterations), and quickly judges that the depth of focus has been best approximated. However, the two - dimensional target representation unit quickly updates the position information of the cat in the XY plane and the simulation parameters due to posture changes by analyzing the latest image frame, generating ; The efficient iteration control mainly refers to the updated position information, instructing the drive unit to perform a small focus adjustment to , while maintaining tight tracking of adjacent areas; after this round of iteration, the obtained value reaches 3,200 units, The consistency score of

[0122] Completion and results: After a total of 12 iterations, running for 180 milliseconds, the system meets the initial completion conditions; specifically, subsequently: The step stabilizes around 3,550 units (the gain in three consecutive iterations is less than 50 units), and the two-dimensional target representation The update amount of two consecutive iterations of

[0123] is less than 2%; finally, the focus depth is locked at

[0124] In the obtained photo display, even during the dynamic process of the elephant, the Siamese cat's face, especially the eyes, has achieved very high precision, the hair details are distinct, and the background presents 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 focus depth adjustment and two-dimensional target representation optimization.

[0124] Case illustration 2: Detail capture in texture still life macro photography;

[0125] 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 of similar tones or other petal parts; 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 petal. The process of processing with this solution is as follows:

[0126] Region initialization and target preliminary preset: The photographer approaches the orchid with the lens and frames the main rose petal for shooting; the image acquisition unit starts transmitting image data; the initialization module of the collaborative iteration 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 body as the potential expected target; the two-dimensional target representation unit generates an initialization representation , whose geometric parameters roughly enclose the visible area of the petal, but due to severe initial defocus, the description of its internal texture complexity is very low, and it can hardly effectively represent the petal veins; the image response characteristic analysis unit initializes the image, based on the spatial frequency index, with a value of 0.25 (normalized value), and the target representation The reconfigurability score of

[0127] Good iteration and representation optimization (Sp1-Sp5 cycle):

[0128] The first iteration (k = 1): The good iteration control unit is based on the blurred , the instruction points to the focus drive unit to perform a tentative depth of focus 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, but at 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 proposed state analysis; the fine iterative control unit generates an update instruction based on this, and after receiving it, the two-dimensional target characterization unit starts to texture complexity, assigns attention weights to the Gabor filter response parameters related to the preliminary description of the vein direction and position, and the target generates a characterization ; and the update amount is 20%, mainly reflected in the preliminary construction of the texture feature attributes;

[0129] The fifth iteration (k = 5): The depth of focus has been adjusted several times to reach ; At this time, it is about to approach 0.65; at 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 control signal structures; the update instruction of the collaborative iterative 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 expand the description of the texture complexity, the characterization parameters of these newly emerging fine veins, and increase the sensitivity threshold of the corresponding parameters; the updated target characterization the delayability score is increased to 0.78, and the update amount with is 12%, and the characterization model begins to be able to effectively describe complex textures;

[0130] The tenth iteration (k = 10): The quantifiable score reaches 0.92; at this time, the target characterization has very precisely drawn the details of the veins of the petals like this and static; the prediction model controlled by the collaborative iterative control unit judges that the current depth of focus is already very close to the best; in Sp3, the feature attribute set presentation state feedback by the image response characteristic analysis unit shows that almost all vein features defined in show extremely high gradient intensity and braking at the current depth of focus; the update instruction mainly makes minor optimizations to to generate , and its update amount with is shown as 1.5%, indicating that the two-dimensional target characterization is highly stable;

[0131] Completion and Results: After a total of 11 iterations, lasting approximately 250 milliseconds, the system determined that the completion condition was met; the final restraint stabilized at 0.95 (normalized value), and the iterative update amount of the two-dimensional target representation was less than 1%; the final focal depth was locked at , and the photographer checked the shooting result. The photo perfectly confirmed the contour world of the orchid: every delicate vein was clearly visible, and the main vein to the almost invisible capillary structure was accurately reproduced. The colors transitioned naturally and softly, fully demonstrating the enhancement and force of the petals. In this case, through the iterative enhancement of focus adjustment and two-dimensional feature representation, it is possible to effectively address the contour problems of complex textures and initially low-clarity targets, achieving extreme capture of tiny details.

[0132] Case Illustration Three: Portrait Face Priority in Low-Light Environment;

[0133] 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 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 this solution are as follows:

[0134] Initialization and Preliminary Target Representation: The user composes the picture, and the friend's face is located slightly off-center in the frame. The image acquisition unit operates with the above ISO parameter of 1600 to enhance 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, it 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 accordingly , 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 value of the Laplacian operator for brightness, the brightness is 35 (the score indicates blurriness), and the quantifiable score is 0.55;

[0135] Iterative Optimization and Representation Enhancement (Sp1-Sp5 Loop):

[0136] Second Iteration (k = 2): The collaborative iterative control unit instructs the drive unit to adjust the focal depth to ; at this depth, the image response characteristic analysis unit feeds back a signal. More importantly, the state analysis indication of the feature attribute set shows that There is a slight enhancement in the local wake-up details of the characterized region, although it remains overall darker; the update instruction of the collaborative iterative control unit optimizes the custom parameters and feature attribute weights of the two-dimensional target characterization unit for the region in ; The update amount is 12%;

[0137] Sixth iteration (k = 6): The focal depth raises the overall integral to 120; at this time, the emphasized facial features represented, especially the edges of the eyes and lips, have their intensity significantly enhanced under the current focus; the two-dimensional target characterization unit, under the instruction of the collaborative iterative control unit, greatly optimizes the reset of the geometric parameters of these key facial features in ; and the description of the texture complexity related to these features (such as the direction of the eyebrows and the texture complexity of the teeth) also begins to be effectively constructed into ; the olfactory score of is increased to 0.80, and the update amount of

[0138] Ninth iteration (k = 9): The inflection point reaches 280, and the target prediction can already quite accurately draw the key details of the subject; the prediction model of the collaborative iterative control unit, based on the continuously stable upward and highest stability trend, conducts a feedback analysis of the image response characteristics of the focus drive unit with a slightly larger amplitude before the focus. Under , the reflected highlight points in the eye region are extremely sharp, and the stress parameters of the facial skin also begin to appear; From to

[0139] Completion and result: The system performs 10 iterations in total, and the total running time is about 320 milliseconds (the processing time in low light may be slightly longer); finally, it stabilizes at around 310, and the iterative update amount of the two-dimensional target characterization is less than 1.5%; the focal 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 naturally blurred because the focus is on the portrait subject, highlighting the subject. This case shows that the imaging device of this camera can still effectively and accurately identify and align with the human face through its unique two-dimensional feedback control mechanism and combined with the priority strategy under low-light conditions, overcoming the deficiencies of traditional docking methods in similar scenarios.

[0140] 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.

[0141] 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 ability, characterized in that: 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; The two-dimensional target characterization unit is connected to the image acquisition unit and is used to generate and iteratively update the target characterization of one or more potential focusing targets on the two-dimensional image plane according to the received image data, and the target characterization includes the position information of the target and the feature attribute set describing its recognizability; The focusing drive unit is used to adjust the depth of focus of the optical component according to the control instruction; The image response characteristic analysis unit is connected to the image acquisition unit and is used to analyze the presentation state of the image characteristics and the feature attribute set corresponding to the target characterization in the current image under the current depth of focus, and the image characteristics include at least one sharpness metric; The collaborative iterative control unit is connected to the two-dimensional target characterization unit, the focusing drive unit, and the image response characteristic analysis unit and is used to perform the following steps: Sp1: Generate a first control instruction to the focusing drive unit based on the target characterization of the current target output by the two-dimensional target characterization unit to adjust the depth of focus; Sp2: Receive the image characteristic analysis result under the depth of focus adjusted in Sp1 feedback by the image response characteristic analysis unit; Sp3: Generate an update instruction to the two-dimensional target characterization unit based on the presentation state of the feature attribute set in the image characteristic analysis result to adjust and optimize its two-dimensional target characterization of one or more potential focusing targets; Sp4: Generate a second control instruction to the focusing drive unit based on the updated and optimized two-dimensional target characterization in Sp3 and combined with the sharpness metric for subsequent depth of focus adjustment; Sp5: Loop to execute Sp2 to Sp4, and through the iterative enhancement between the optimization effect of the depth of focus adjustment on the two-dimensional target characterization and the precise guiding effect of the optimized two-dimensional target characterization on the subsequent depth of focus adjustment until the preset focusing completion condition is met.

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

3. The camera focusing device with two-dimensional feedback control ability according to claim 1, characterized in that: The presentation state of the feature attribute set output by the image response characteristic analysis unit includes the gradient intensity, spatial frequency energy distribution, and statistical dispersion of the feature attribute under the current depth of focus.

4. The camera focusing device with two-dimensional feedback control ability according to claim 1, characterized in that: In Sp3, if the analysis result indicates that the current depth of focus enhances the recognizability of a certain specific feature attribute in the target characterization, the update instruction in the collaborative iterative control unit will strengthen the significance of this specific feature attribute in the two-dimensional characterization.

5. The camera focusing device with two-dimensional feedback control ability according to claim 1, characterized in that: In Sp4, the collaborative iterative control unit uses a prediction model combined with the change trend of the sharpness metric and the stability of the two-dimensional target characterization to determine the amplitude and direction of the subsequent depth of focus adjustment.

6. The camera focusing device with two-dimensional feedback control ability according to claim 1, characterized in that: The focus completion conditions preset by 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.

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

8. The camera focusing device with two-dimensional feedback control ability according to claim 1, characterized in that: The collaborative iterative control unit includes an initialization module for providing an initial two-dimensional target representation for the two-dimensional target representation unit before the first iteration by performing a quick analysis of the global image and receiving external specifications.

9. The camera focusing device with two-dimensional feedback control ability according to claim 1, characterized in that: The camera focusing device includes the following steps: Sp1: Capture image data through the 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 drive unit adjusts the focus depth according to the first control instruction of the collaborative iterative control unit, and this 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 drive unit for subsequent focus depth adjustment; Sp7: Loop and execute Sp4 to Sp6 until the preset focus completion conditions are met.

Images