Multi-frame synchronous fusion imaging system and method based on motion compensation driving

Abstract

The application discloses a multi-frame synchronous fusion imaging system and method based on motion compensation driving, and belongs to the technical field of motion imaging; the application comprises an imaging supervision platform constructed by networking database; the imaging supervision platform is in communication connection with a motion multi-point data acquisition module for acquiring multi-source motion data in a monitoring area in real time, and a dynamic database is constructed; a motion data normalization analysis module is used for standardizing the data in the dynamic database, extracting motion trajectory, speed vector and interframe offset characteristics; the application combines principal component analysis (PCA) and support vector machine (SVM) to realize data dimension reduction and good-bad classification, solves the problems of single data and one-sided analysis of the traditional imaging system, improves the comprehensiveness and accuracy of feature extraction, multi-source data fusion and dynamic analysis; realizes fine regulation and control from micro displacement compensation to image fusion, effectively deals with motion interference of different degrees, and realizes real-time risk assessment and grading compensation.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of motion imaging technology, in particular to a multi-frame synchronous fusion imaging system and method based on motion compensation driving. BACKGROUND

[0002] Multi-frame image fusion algorithm is widely used in the industry due to its good denoising effect and good detail texture retention. The traditional method is to continuously shoot multiple images by a camera, select one frame as a reference frame, and then fuse the other images after registration to obtain a denoised image. However, there are many challenges in practical applications. For example, in the field of robot vision, the six-degree-of-freedom random motion of the robot within the exposure time will cause the loss of high-frequency information after rotation, scaling and translation of the originally clear image, increasing the difficulty of image analysis and recognition.

[0003] In the image acquisition and display process, due to the limitations of CCD / CMOS image sensors in cameras and ordinary displays, the dynamic range of the image is low, and the details of the overexposed or underexposed areas are poor. People expect to expand the dynamic range of the image through technical means, adjust the overexposure and underexposure parts, enhance the detail performance, and make the image closer to the effect captured by the human eye. At the same time, in some special imaging scenarios such as non-visual imaging, the limited signal photons and low signal-to-noise ratio limit the imaging speed, range and quality, and realizing real-time high-quality imaging of complex scenes becomes a problem to be solved.

[0004] Motion compensation technology is a key technology in digital video processing and is widely used in video compression and display enhancement. It uses the similarity between consecutive frames in a video to estimate the motion vector of the object by analyzing the difference between the previous and subsequent frames, and then generates intermediate frames or compensates for the current frame motion. For example, in video encoding, it can reduce data volume, improve transmission efficiency and storage space utilization; in video playback devices, it can improve picture smoothness, reduce motion blur and trailing phenomenon; in medical imaging, virtual reality and augmented reality fields, motion compensation technology also plays an important role, such as eliminating image jitter caused by patient or doctor hand movement in medical imaging, improving diagnostic accuracy and surgical safety, and reducing picture delay caused by head or body movement in VR / AR, improving user immersion and interactivity.

[0005] In combination with the above content, it needs to be explained that the Chinese patent with application number CN2022101488801 discloses a motion imaging compensation device and a motion compensation method. The linear velocity component of the main rotating structure at the axis of the secondary rotating structure on the main optical axis of the imaging light path is used for compensation of the relative motion of the imaging target and the imaging unit. The secondary rotating structure drives the optical components to rotate to realize attitude angle compensation. In the set motion compensation interval, the imaging field is imaged in a relatively static state during the exposure of the imaging system focal plane.

[0006] In fact, only relying on a single type of sensor or data source, lacking the integration of multi-dimensional information such as motion trajectory, environmental disturbance, etc., leads to one-sided feature extraction, making it difficult to accurately identify the source of motion disturbance, overcompensating for slight motion disturbance, wasting computing resources, and responding insufficiently to severe motion disturbance, resulting in image blur or frame loss phenomenon.

[0007] In view of the above technical defects, the present solution is proposed. SUMMARY

[0008] The purpose of the present application is to provide a multi-frame synchronous fusion imaging system and method based on motion compensation driving to solve the problems.

[0009] To achieve the above purpose, the present application provides the following technical scheme: a multi-frame synchronous fusion imaging system based on motion compensation driving, comprising an imaging supervision platform constructed by a networked database, the imaging supervision platform being communicatively connected with a motion multi-point data acquisition module for acquiring multi-source motion data in a monitoring area in real time, and a dynamic database being constructed;

[0010] The motion data normalization analysis module is used for standardizing the data in the dynamic database, extracting the motion trajectory, speed vector and inter-frame offset feature, screening the effective data based on the preset threshold and marking the abnormal branch information label.

[0011] The normal data frame loss risk module is used for combining the effective data and the abnormal branch information label, constructing a multi-dimensional risk assessment model, and generating a frame loss risk level signal.

[0012] The linkage frame loss decision-making and control module generates a motion compensation instruction and a multi-frame fusion strategy according to the risk level signal, drives the execution mechanism to complete the imaging compensation, and feeds back the control result to the imaging supervision platform, so that the imaging supervision platform provides a visual monitoring interface for medical staff, displays the risk warning and compensation state in real time, and supports dynamic adjustment of compensation parameters.

[0013] Further, the preliminary management and classification process of the motion multi-point data acquisition module on the collected data is as follows:

[0014] The multi-source motion data includes device motion trajectory data, inter-frame offset vector data and environmental disturbance data, which are collected in real time by high-precision sensors and image acquisition devices, and are classified and stored according to the time sequence of collection time marked with time stamp UTC; the collected motion data is normalized and pretreated, and a sliding window filtering algorithm is used to eliminate high-frequency noise.

[0015] Further, the analysis process of the motion data normalization analysis module on the screened motion data is as follows:

[0016] Extract the motion trajectory curvature, speed vector length, and inter-frame offset from the filtered motion data, mark them as key features and sequentially mark them as , and , construct a feature vector matrix:

[0017]

[0018] Reduce the dimension of the feature vector matrix by the principal component PCA algorithm, obtain a preset variance threshold from the imaging supervision platform, process the product of the preset variance threshold and the total amount of principal components, obtain effective features, and perform two classification on the effective features based on a support vector machine SVM model, and distinguish to obtain high-quality data and degraded data, and assign corresponding branch information labels to them.

[0019] Further, the process of generating branch information according to the high and low data by the motion data normalization analysis module is as follows:

[0020] Call the preset classification data reference table and the frame loss threshold from the inside of the imaging supervision platform , compare the high-quality data and degraded data with the classification data reference table: if the classification result is high-quality data, generate a low-risk imaging branch label and retain the complete frame data.

[0021] Further, if the classification result is degraded data, further judge whether the inter-frame offset is greater than the frame loss threshold . If > , a high-offset frame loss risk branch label is generated; if < , a low-offset quality degradation branch label is generated, all branch information labels and corresponding motion data are stored in the dynamic database, and sent to the normal data frame loss risk module.

[0022] Further, the frame loss risk analysis process of the normal data frame loss risk module combined with the screened data and branch information is as follows:

[0023] Call the effective data and branch information labels in the dynamic database, construct a risk analysis feature set, the risk analysis feature set includes motion trajectory curvature fluctuation, speed vector change rate and branch label type, and sequentially mark them , and , train the feature set using a random forest model, output a frame loss risk probability , and compare and analyze the pre-stored risk threshold range and the output frame loss risk probability from the inside of the imaging supervision platform.

[0024] Further, if If the risk threshold range is minimum, the risk is determined to be low, and a normal imaging signal is generated; if If the risk threshold range is within the range, the risk is determined to be medium, and a motion compensation warning signal is generated, triggering inter-frame motion vector estimation; if If the risk threshold range is greater than or equal to the maximum risk threshold range, the risk is determined to be high, and a multi-frame fusion emergency signal is generated, starting a multi-frame image registration and fusion process, calculating the inter-frame mapping relationship based on the improved optical flow method, and constructing a fusion weight matrix .

[0025] Further, the normal data frame loss risk module generates a regulation and compensation signal as follows:

[0026] The risk level signal and the corresponding motion compensation warning signal and multi-frame fusion emergency signal are packaged into a regulation data packet and sent to the linkage frame loss decision regulation module; the regulation data packet includes a "parameter maintenance instruction" corresponding to low risk, a "motion vector compensation instruction" corresponding to medium risk, and a "multi-frame fusion instruction" corresponding to high risk; after receiving the regulation data packet, the linkage frame loss decision regulation module analyzes the instruction type and drives the execution mechanism: for the motion vector compensation instruction, the imaging device is controlled to perform micro-displacement compensation in the opposite direction; for the multi-frame fusion instruction, the image fusion algorithm is called to perform weighted fusion on the marked frame loss risk frame, generating a high-definition image after compensation.

[0027] The working method of the multi-frame synchronous fusion imaging system based on motion compensation driving includes the following steps:

[0028] Data acquisition and preprocessing stage: real-time acquisition of multi-source motion data, filtering and denoising and marking of abnormal branches;

[0029] Feature extraction and good-bad classification stage: extract features to construct a matrix, and use SVM to distinguish good and bad data after dimension reduction and mark the branches;

[0030] Frame loss risk assessment and signal generation stage: based on random forest risk assessment, generate low, medium and high risk signal driving strategies;

[0031] Motion compensation and multi-frame fusion execution stage: perform compensation according to the risk level, micro-displacement or multi-frame fusion to correct motion errors.

[0032] The beneficial effects of the present application are:

[0033] 1、The present application integrates trajectory, offset, environment and other multi-dimensional data through a motion multi-point data acquisition module, combines principal component analysis PCA and support vector machine SVM to realize data dimension reduction and good-bad classification, solves the problem of single data and one-sided analysis of traditional imaging systems, improves the comprehensiveness and accuracy of feature extraction, and realizes multi-source data fusion and dynamic analysis.

[0034] 2、The application is to construct a multi-dimensional risk assessment system based on a random forest model, according to the frame loss risk probability The dynamic generation of the three-level response strategies of regular imaging, motion compensation and multi-frame fusion realizes the fine regulation and control from micro-displacement compensation to image fusion, effectively deals with motion interference of different degrees, and realizes real-time risk assessment and grading compensation. BRIEF DESCRIPTION OF DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0036] Figure 1 The whole method flowchart of the application is shown in the figure;

[0037] Figure 2 The whole block diagram of the system of the application is shown in the figure;

[0038] Figure 3 The dynamic database construction schematic diagram of the application is shown in the figure. DETAILED DESCRIPTION

[0039] The technical solutions in the embodiments of the application will be described clearly and completely in the following with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only some embodiments of the application, not all. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the application.

[0040] Embodiment one: please refer to Figure 1 - Figure 3 As shown in the figure, the embodiment is a multi-frame synchronous fusion imaging system based on motion compensation driving, which includes an imaging supervision platform constructed by networking database, the imaging supervision platform is in communication connection with a motion multi-point data acquisition module for acquiring multi-source motion data in a monitoring area in real time, and a dynamic database is constructed;

[0041] The linkage frame loss decision control module generates motion compensation instructions and multi-frame fusion strategies according to the risk level signal, drives the execution mechanism to complete imaging compensation, and feeds back the control result to the imaging supervision platform. The imaging supervision platform provides a visual monitoring interface for medical staff, displays the risk early warning and compensation state in real time, supports dynamic adjustment of compensation parameters, and the preliminary management and classification process of the collected data by the motion multi-point data acquisition module is as follows:

[0042] The multi-source motion data includes device motion trajectory data, inter-frame offset vector data and environmental disturbance data, which are collected in real time by high-precision sensors and image acquisition devices, and are stored in classified documents according to the collection time sequence marked with time stamp UTC; the collected motion data is normalized and pre-processed, and a sliding window filtering algorithm is used to eliminate high-frequency noise, and the calculation formula is:

[0043]

[0044] Wherein, The filter output value at time t is represented as The window size is represented as The original data point is represented as The original motion parameter corresponding to the UTC time stamp is obtained from the imaging supervision platform, and the pre-stored motion parameter threshold and Row comparison analysis:

[0045] When The motion parameter threshold, it is marked as an abnormal data branch, and it is sent to the motion data normalization analysis module.

[0046] The motion data normalization analysis module is used for standardizing the data in the dynamic database, extracting the motion trajectory, velocity vector and inter-frame offset features, screening the effective data based on the preset threshold and marking the abnormal branch information label; the analysis process of the motion data normalization analysis module on the screened motion data is as follows:

[0047] The motion trajectory curvature, velocity vector module length and inter-frame offset are extracted from the filtered motion data, which are marked as key features and are marked as , and in turn, and a feature vector matrix is constructed:

[0048]

[0049] The feature vector matrix is reduced by the principal component PCA algorithm, the preset variance threshold is obtained from the imaging supervision platform, and the effective features are obtained by processing the product of the preset variance threshold and the total amount of principal components, the effective features are classified based on the support vector machine SVM model, and the high-quality data and the degraded data are obtained, and the corresponding branch information label is given, and the calculation formula is:

[0050]

[0051] Wherein, The feature vector matrix is represented as The feature vector of The feature vector in the training sample, such as the K, V, Ensure that the formula input is consistent with the textual description; is expressed as a Lagrange multiplier; is expressed as a sample label; is expressed as a kernel function, which is used to calculate the similarity of the current data with historical data ; is expressed as a bias term coefficient; sign(...) is the "sign function", which extracts the sign of the calculation result, only retains +1 or -1, similar to "non-superior or inferior" judgment;

[0052] is expressed as the decision function of the support vector machine (SVM) model. The SVM model is trained by historical data to obtain an "optimal boundary line" that separates high-quality data and degraded data. Accordingly, when new motion data features are input, f (M) calculates the distance and direction of M to the boundary line, judges which category of the imaging regulatory platform preset numerical planning M belongs to, and according to the preset numerical planning type, marks the classification result as "high-quality data" or "degraded data", and generates the corresponding branch information label.

[0053] The process of generating branch information by the motion data normalization analysis module according to high-quality and degraded data is as follows:

[0054] Retrieve the preset classification data reference table and frame loss threshold from the imaging regulatory platform Compare high-quality data and degraded data with the classification data reference table:

[0055] If the classification result is high-quality data, generate a low-risk imaging branch label and retain complete frame data;

[0056] If the classification result is degraded data, further judge whether the frame offset is greater than the frame loss threshold :

[0057] If > , generate a high-offset frame loss risk branch label;

[0058] If < , generate a low-offset quality degradation branch label. Store all branch information labels and corresponding motion data synchronously to the dynamic database, and send them to the normal data frame loss risk module.

[0059] Embodiment two: the embodiment is a multi-frame synchronous fusion imaging system based on motion compensation driving, including normal data missing frame risk module: for combining effective data and abnormal branch information label, constructing multi-dimensional risk assessment model, generating missing frame risk level signal; the normal data missing frame risk module combines the missing frame risk analysis process of the screened data and branch information as follows:

[0060] Call effective data and branch information label in dynamic database, construct risk analysis feature set, risk analysis feature set includes motion trajectory curvature fluctuation, speed vector change rate and branch label type, and mark them in turn 、 and , train the feature set by using random forest model, output missing frame risk probability Each decision tree of random forest model Receive three specific parameters as input, output the risk prediction result of single tree by learning the association between different parameter combinations and missing frame risk, and finally get the comprehensive risk probability by summing the results of all trees , the calculation formula is:

[0061]

[0062] Among them, Indicates the motion trajectory curvature fluctuation, the difference between the current curvature and the historical average curvature, reflecting the stability of the motion path; Indicates the speed vector change rate, the change of speed with time, reflecting the risk of sudden change of motion speed; Indicates the branch label type, including low risk imaging branch, high offset missing frame risk branch and low offset quality recession branch, reflecting the type of data anomaly; Indicates the total number of decision trees; Indicates the prediction result of the tth decision tree to the risk analysis feature set;

[0063] Compare the pre-stored risk threshold range with the output missing frame risk probability

[0064] If < risk threshold range minimum value, it is judged as low risk, and the normal imaging signal is generated, the linkage frame decision control module receives the normal imaging signal, maintains the current imaging device parameters such as lens focal length, exposure time, sends instructions to the imaging device controller through the imaging supervision platform, keeps the device stationary or moves at a uniform speed according to the preset trajectory, collects device state data such as motor speed and lens position every 5 frames, verifies that the parameters have not drifted, which is used for data feedback;

[0065] If ​In the risk threshold range, it is determined as medium risk, a motion compensation warning signal is generated, and inter-frame motion vector estimation is triggered, and the calculation formula is:

[0066]

[0067] wherein, represents the motion vector of the pixel point ; and respectively represent the motion displacement of the horizontal and vertical coordinates;

[0068] After the linkage frame compensation decision control module receives the motion compensation warning signal, the motion vector is extracted, and the device is controlled to move in the opposite direction . Specifically, if = (+5, -3) indicates that the image is offset to the right and down, the driving part or motor of the device needs to be operated to move to the left and up by 5 pixels of equivalent physical distance, and lens focal length conversion or adjustment needs to be combined. Adjust the parameters according to the actual needs to complete the sub-millimeter level compensation;

[0069] If ≥ the maximum value of the risk threshold range, it is determined as high risk, a multi-frame fusion emergency signal is generated, a multi-frame image registration and fusion process is started, an inter-frame mapping relationship is calculated based on an improved optical flow method, and a fusion weight matrix is constructed, and the calculation formula is:

[0070]

[0071] wherein, represents the fusion weight of the pixel point , and represents the second order norm of the motion vector of the point;

[0072] After the linkage frame compensation decision control module receives the multi-frame fusion emergency signal, it is executed based on the operation of the networking database connected by the imaging supervision platform. The frame number registration can additionally use, for example, the ORB feature point algorithm detection, and is not limited thereto. The adjacent frames are matched, the pixel points are weighted and averaged according to the fusion weight matrix , the motion blur area is suppressed, and finally the imaging supervision platform is used to output the fused high-definition image based on accelerated image processing.

[0073] The process of generating the control and compensation signal by the normal data frame loss risk module is as follows: the risk level signal and the corresponding motion compensation warning signal and multi-frame fusion emergency signal are packaged into a control data packet and sent to the linkage frame compensation decision control module.

[0074] The regulation data packet comprises a "parameter maintenance instruction" corresponding to a low risk, a "motion vector compensation instruction" corresponding to a medium risk, and a "multi-frame fusion instruction" corresponding to a high risk. It should be noted that the motion vector compensation instruction is accompanied by data so as to calibrate and accurately regulate according to the data; and the multi-frame fusion instruction is accompanied by a matrix and registration parameters, and the calibration and accurate regulation are performed according to actual numerical values and the data;

[0075] After the linkage frame decision regulation module receives the regulation data packet, the instruction type is parsed and the actuator is driven to execute: for the motion vector compensation instruction, the imaging device is controlled to move in the opposite direction for micro-displacement compensation;

[0076] For the multi-frame fusion instruction, an image fusion algorithm is called to perform weighted fusion on the marked frame risk frame, to generate a high-definition image after compensation.

[0077] In combination with Embodiment One and Embodiment Two, the motion multi-point data acquisition module integrates trajectory, offset, environment and other multi-dimensional data, and the dimensionality is reduced through principal component analysis PCA and classified through support vector machine SVM, to solve the problem of single data and one-sided analysis of the traditional imaging system, and to improve the comprehensiveness of feature extraction;

[0078] At the same time, a multi-dimensional risk assessment system is constructed based on a random forest model, a three-level response strategy of "regular imaging, motion compensation, multi-frame fusion" is dynamically generated according to the frame risk probability, fine regulation and control from micro-displacement compensation to image fusion are realized, different degrees of motion interference are effectively coped with, and multi-source data fusion analysis and real-time grading and accurate compensation of motion risk are realized.

[0079] The above content is only an example and description of the structure of the present application, and those skilled in the art can make various modifications or supplements or use similar ways to replace the described specific embodiments, as long as the modifications or supplements do not deviate from the structure of the present application or exceed the scope defined by the present application, and the modifications or supplements shall belong to the protection scope of the present application.

[0080] In the description of the present application, the description of the terms "one embodiment", "example", "specific example" and the like means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are contained in at least one embodiment or example of the present application. In the present application, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner, and the relevant fittings include couplings, lead screws, gears, gaskets and other commonly used mechanical connecting parts in the field, and are not limited thereto, and the connecting mode is replaced and used according to the actual use.

[0081] The preferred embodiments of the application disclosed above are only to facilitate the elucidation of the application. The preferred embodiments do not describe all the details of the application and limit the application to the specific embodiments. Obviously, many modifications and variations can be made in light of the teachings above. The description is chosen and described in order to provide the best illustration of the application and its practical application to those skilled in the art and to enable those skilled in the art to best utilize the application. The application is limited only by the claims and their full scope and equivalents.

Claims

1. A multi-frame synchronous fusion imaging system based on motion compensation driving, characterized in that, The imaging supervision platform is built with a networked database, and the imaging supervision platform is communicatively connected with a motion multi-point data acquisition module for acquiring multi-source motion data in a monitoring area in real time, and a dynamic database is built; The motion data normalization analysis module is used for standardizing the data in the dynamic database, extracting motion trajectories, speed vectors, and frame offset features, screening effective data based on a preset threshold, and marking abnormal branch information tags; The normal data frame loss risk module is used for combining the effective data and the abnormal branch information tags, building a multi-dimensional risk assessment model, and generating a frame loss risk level signal; The linkage frame loss decision-making and control module generates motion compensation instructions and multi-frame fusion strategies according to the risk level signal, drives an execution mechanism to complete imaging compensation, and feeds back the control result to the imaging supervision platform, and the imaging supervision platform provides a visual monitoring interface for medical staff, displays risk early warning and compensation status in real time, and supports dynamic adjustment of compensation parameters; The analysis process of the motion data normalization analysis module on the screened motion data is as follows: the motion trajectory curvature, the speed vector module length and the inter-frame offset are extracted from the filtered motion data, which are marked as key features and sequentially marked as 、 and , and a feature vector matrix is constructed: The feature vector matrix is reduced in dimension through a principal component PCA algorithm, a preset variance threshold is obtained from the imaging supervision platform, the product of the preset variance threshold and the total amount of principal components is processed, effective features are obtained, the effective features are classified based on a support vector machine SVM model, and high-quality data and degraded data are distinguished and marked with corresponding branch information; The normal data missing frame risk module combines the missing frame risk analysis process of the screened data and branch information as follows: the effective data and branch information tags in the dynamic database are called, a risk analysis feature set is constructed, the risk analysis feature set includes the motion trajectory curvature fluctuation, the speed vector change rate and the branch tag type, and they are sequentially marked , and , a random forest model is used to train the feature set, and a missing frame risk probability is output. Retrieving pre-stored risk threshold range and outputting frame loss risk probability from inside the imaging supervision platform Performing comparison analysis: if <risk threshold range minimum value, determining low risk, generating regular imaging signal; if located in the risk threshold range, determining medium risk, generating motion compensation warning signal, triggering inter-frame motion vector estimation; if ≥risk threshold range maximum value, determining high risk, generating multi-frame fusion emergency signal, starting multi-frame image registration and fusion process, calculating inter-frame mapping relationship based on improved optical flow method, constructing fusion weight matrix .

2. The multi-frame synchronous fusion imaging system based on motion compensation driving according to claim 1, characterized in that, The preliminary management and classification process of the motion multi-point data acquisition module on the collected data is as follows: The multi-source motion data includes device motion trajectory data, inter-frame offset vector data and environmental disturbance data, which are collected in real time through high-precision sensors and image acquisition equipment, and are classified and stored with time stamps UTC according to the collection time sequence; the collected motion data is normalized and preprocessed, and a sliding window filtering algorithm is used to eliminate high-frequency noise.

3. The multi-frame synchronous fusion imaging system based on motion compensation driving according to claim 2, characterized in that, The process of generating branch information according to the high-quality and degraded data by the motion data normalization analysis module is as follows: Retrieving preset classification data reference table and frame loss threshold from inside the imaging regulatory platform Comparing high-quality data and degraded data with the classification data reference table: if the classification result is high-quality data, a low-risk imaging branch label is generated, and complete frame data is retained.

4. The multi-frame synchronous fusion imaging system based on motion compensation driving according to claim 3, characterized in that, If the classification result is the deteriorated data, it is further judged whether the inter-frame offset is greater than a frame loss limit threshold, or not : If > , a high offset frame loss risk branch label is generated; If < a low offset quality degradation branch tag is generated, all branch information tags are stored synchronously with corresponding motion data to a dynamic database, and sent to a normal data missing frame risk module.

5. The multi-frame synchronous fusion imaging system based on motion compensation driving according to claim 1, characterized in that, The process of generating control and compensation signals by the normal data frame loss risk module is as follows: The risk level signal, the corresponding motion compensation warning signal and the multi-frame fusion emergency signal are packaged into a control data packet, and sent to the linkage frame loss decision-making and control module; The control data packet includes a "parameter maintenance instruction" corresponding to a low risk, a "motion vector compensation instruction" corresponding to a medium risk, and a "multi-frame fusion instruction" corresponding to a high risk; After receiving the control data packet, the linkage frame loss decision-making and control module analyzes the instruction type and drives the execution mechanism: for the motion vector compensation instruction, the imaging equipment is controlled to perform micro-displacement compensation in the opposite direction; for the multi-frame fusion instruction, an image fusion algorithm is called to perform weighted fusion on the marked frame loss risk frame, and a high-definition image after compensation is generated.

6. The working method of the multi-frame synchronous fusion imaging system based on motion compensation driving, for the multi-frame synchronous fusion imaging system based on motion compensation driving according to any one of claims 1-5, characterized in that, The method comprises the following steps: The data acquisition and preprocessing stage: real-time acquisition of multi-source motion data, filtering and denoising, and marking of abnormal branches; The feature extraction and quality classification stage: extracting features to build a matrix, distinguishing high-quality and degraded data after dimension reduction, and marking branches; The frame loss risk assessment and signal generation stage: generating low, medium and high risk signals to drive strategies based on random forest risk assessment; The motion compensation and multi-frame fusion execution stage: performing compensation according to the risk level, micro-displacement or multi-frame fusion to correct motion errors.

Images