CN-121582842-B - Image processing method and system for dynamic monitor

CN121582842BCN 121582842 BCN121582842 BCN 121582842BCN-121582842-B

Abstract

The invention discloses an image processing method and system for a dynamic monitor, which relate to the technical field of image processing and comprise the steps of deriving image frames in a monitoring window through a target dynamic monitor to obtain a monitoring image frame sequence; the method comprises the steps of traversing a monitoring image frame sequence to score image quality, marking, obtaining a marking monitoring image frame sequence, carrying out semantic conversion analysis and multi-scale feature recognition to obtain a structural description text sequence and a marking multi-scale feature group sequence, carrying out intra-group feature interaction on the marking multi-scale feature group sequence to determine a marking interaction multi-scale feature sequence, carrying out double memory iteration on the marking interaction multi-scale feature sequence and the structural description text sequence to determine target iterative memory, carrying out recognition, and determining an image processing result of a monitoring window. The method solves the technical problems of low monitoring and identifying precision and poor reliability of the dynamic image in the prior art, and achieves the technical effect of improving the accuracy and stability of abnormal identification.

Inventors

LIU QIMING
ZHAO YIN
WANG HAOYU
LUO YANFENG
ZHANG XIAO
Lv Tianci

Assignees

首都医科大学附属北京中医医院

Dates

Publication Date: 20260508
Application Date: 20251111

Claims (9)

1. An image processing method for a dynamic monitor, the method comprising: the method comprises the steps of exporting image frames in a monitoring window through an SDK driving interface of a target dynamic monitor to obtain a monitoring image frame sequence; performing image quality scoring on the monitoring image frame sequence according to a preset scoring index, and marking according to a scoring result to obtain a marking monitoring image frame sequence; traversing the identification monitoring image frame sequence to perform semantic conversion analysis and multi-scale feature recognition to obtain a structural description text sequence and an identification multi-scale feature group sequence; performing intra-group feature interaction according to the identification multi-scale feature group sequence to determine an identification interaction multi-scale feature sequence; performing double memory iteration on the identification interaction multi-scale feature sequence and the structural description text sequence to determine target iterative memory; identifying the target iterative memory, and determining an image processing result of a monitoring window; Performing double memory iteration on the identification interaction multi-scale feature sequence and the structural description text sequence to determine target iterative memory, wherein the method comprises the following steps: Extracting a first identification interaction multiscale feature, a first structural description text, a second identification interaction multiscale feature and a second structural description text from the identification interaction multiscale feature sequence and the structural description text sequence; Performing double memory iteration on the first identification interaction multi-scale feature, the first structural description text, the second identification interaction multi-scale feature and the second structural description text to obtain a first iterative memory; performing double memory iteration on a third identification interaction multi-scale feature and a third structural description text in the identification interaction multi-scale feature sequence and the structural description text sequence by using a second identification interaction multi-scale feature and a second structural description text, supplementing the first iteration memory according to an iteration result to obtain a second iteration memory, and analogizing to obtain a stage iteration memory; And respectively adding the last identification interaction multi-scale feature and the last structural description text sequence in the identification interaction multi-scale feature sequence and the structural description text sequence into a stage iteration memory to obtain the target iteration memory.
2. An image processing method for a dynamic monitor as claimed in claim 1, wherein the predetermined scoring criteria include peak signal-to-noise ratio, sharpness, brightness, exposure and contrast.
3. The image processing method for a dynamic monitor according to claim 1, wherein traversing the sequence of identified monitored image frames for semantic conversion analysis and multi-scale feature recognition, obtaining a sequence of structured descriptive text and a sequence of identified multi-scale feature sets, comprises: invoking a semantic converter to perform semantic conversion analysis on the identification monitoring image frame sequence respectively to obtain a structural description text sequence; Acquiring a multi-scale set and constructing a multi-scale feature analyzer set; and carrying out multi-scale feature recognition on the identification monitoring image frame sequence by utilizing the multi-scale feature analyzer set to obtain the identification multi-scale feature group sequence.
4. A method of image processing for a dynamic monitor as claimed in claim 3, wherein obtaining a set of multi-scale features to construct a set of multi-scale feature analyzers comprises: Acquiring a historical image abnormal log set of the dynamic monitor; Performing similar division on the historical image abnormal log set according to the abnormal type to obtain a plurality of divided historical image abnormal log sets; Extracting abnormal interval duration in the log from the plurality of partition history image abnormal log sets to obtain a plurality of partition abnormal interval duration sets; and screening based on the multiple divided abnormal interval duration sets to obtain the multi-scale set.
5. The method for image processing for a dynamic monitor according to claim 4, wherein filtering based on the plurality of divided abnormal interval duration sets to obtain the multi-scale set comprises: respectively extracting the maximum partition abnormal interval duration and the minimum partition abnormal interval duration in the partition abnormal interval duration sets to obtain a plurality of maximum partition abnormal interval durations and a plurality of minimum partition abnormal interval durations; And performing union calculation on the maximum division abnormal interval duration and the minimum division abnormal interval duration to obtain the multi-scale set.
6. The image processing method for a dynamic monitor of claim 1, wherein performing intra-group feature interactions based on the identified multi-scale feature group sequence, determining an identified interactive multi-scale feature sequence, comprises: respectively calculating intra-group feature mean values of the identification multi-scale feature group sequences to obtain leading identification multi-scale feature sequences; Respectively carrying out intra-group feature interaction on corresponding identification multi-scale feature groups in the identification multi-scale feature group sequence based on the leading identification multi-scale feature sequence to obtain an identification intra-group interaction multi-scale feature group sequence; and carrying out intra-group average analysis on the intra-identification-group interaction multi-scale feature group sequence to obtain an identification interaction multi-scale feature sequence.
7. The image processing method for a dynamic monitor according to claim 6, wherein performing intra-group feature interaction on corresponding ones of the identified multi-scale feature group sequences based on the leading identified multi-scale feature sequences, respectively, to obtain an identified intra-group interaction multi-scale feature group sequence, comprises: Calculating the similarity of the leading identification multi-scale features and the same type of sub-features of the corresponding identification multi-scale feature group in the identification multi-scale feature group sequence, and constructing a leading interaction matrix group sequence according to a calculation result; and carrying out convolution interaction on the corresponding identification multi-scale feature group in the identification multi-scale feature group sequence based on the leading interaction matrix group sequence to obtain the interaction multi-scale feature group sequence in the identification group.
8. The image processing method for a dynamic monitor according to claim 1, wherein performing a double memory iteration on the first identified interactive multi-scale feature, the first structured descriptive text, the second identified interactive multi-scale feature, and the second structured descriptive text to obtain a first iterative memory comprises: Trend analysis is carried out on the first identification interaction multi-scale feature and the first identification interaction multi-scale feature, and first feature iteration trend information is obtained; carrying out semantic trend analysis on the first structural description text and the second structural description text to obtain first semantic iteration trend information; And adding the first characteristic iteration trend information and the first semantic iteration trend information into a first iteration memory.
9. An image processing system for a dynamic monitor, wherein the system is configured to perform an image processing method for a dynamic monitor as claimed in any one of claims 1 to 8, the system comprising: The image frame deriving module is used for deriving the image frames in the monitoring window through the SDK driving interface of the target dynamic monitor to obtain a monitoring image frame sequence; the scoring module is used for traversing the monitoring image frame sequence, scoring the image quality according to a preset scoring index, and marking according to a scoring result to obtain a marking monitoring image frame sequence; the analysis and identification module is used for traversing the identification monitoring image frame sequence to perform semantic conversion analysis and multi-scale feature identification to obtain a structural description text sequence and an identification multi-scale feature group sequence; the intra-group feature interaction module is used for performing intra-group feature interaction according to the identification multi-scale feature group sequence to determine an identification interaction multi-scale feature sequence; the double-memory iteration module is used for carrying out double-memory iteration on the identification interaction multi-scale feature sequence and the structural description text sequence to determine target iterative memory; and the identification module is used for identifying the target iteration memory and determining an image processing result of the monitoring window.

Description

Image processing method and system for dynamic monitor Technical Field The invention relates to the technical field of image processing, in particular to an image processing method and system for a dynamic monitor. Background Along with the wide application of intelligent monitoring equipment, dynamic monitors play an increasingly important role in the fields of industrial detection, safety protection and the like, but because the acquired dynamic images are often influenced by factors such as definition, brightness, exposure, contrast and the like, the image quality is unstable, and a single frame image is difficult to comprehensively express deep semantics in the dynamic change process and correlate with multi-scale features, so that the accuracy and reliability of abnormal identification are insufficient, and the requirement of high-accuracy dynamic monitoring is difficult to meet. Disclosure of Invention The application provides an image processing method and system for a dynamic monitor, which are used for solving the technical problems of low dynamic image monitoring and identifying precision and poor reliability in the prior art. In view of the above, the present application provides an image processing method and system for a dynamic monitor. In a first aspect of the present application, there is provided an image processing method for a dynamic monitor, the method comprising: The method comprises the steps of obtaining a monitoring image frame sequence by deriving an image frame in a monitoring window through an SDK driving interface of a target dynamic monitor, conducting image quality grading on the monitoring image frame sequence according to a preset grading index, conducting identification according to a grading result to obtain an identification monitoring image frame sequence, conducting semantic conversion analysis and multi-scale feature identification on the identification monitoring image frame sequence to obtain a structural description text sequence and an identification multi-scale feature group sequence, conducting intra-group feature interaction according to the identification multi-scale feature group sequence to determine an identification interaction multi-scale feature sequence, conducting double memory iteration on the identification interaction multi-scale feature sequence and the structural description text sequence to determine target iterative memory, conducting identification on the target iterative memory, and determining an image processing result of the monitoring window. In a second aspect of the present application, there is provided an image processing system for a dynamic monitor, the system comprising: The system comprises an image frame deriving module, a scoring module, an analysis and identification module, an intra-group feature interaction module, a dual memory iteration module and an identification module, wherein the image frame deriving module is used for deriving an image frame in a monitoring window through an SDK driving interface of a target dynamic monitor to obtain a monitoring image frame sequence, the scoring module is used for traversing the monitoring image frame sequence to score the image quality according to a preset scoring index and identify according to a scoring result to obtain an identification monitoring image frame sequence, the analysis and identification module is used for traversing the identification monitoring image frame sequence to conduct semantic conversion analysis and multi-scale feature identification to obtain a structural description text sequence and an identification multi-scale feature group sequence, the intra-group feature interaction module is used for conducting intra-group feature interaction according to the identification multi-scale feature group sequence to determine an identification interaction multi-scale feature sequence, the dual memory iteration module is used for conducting dual memory iteration on the identification interaction multi-scale feature sequence and the structural description text sequence to determine a target iterative memory, and the identification module is used for identifying the target iterative memory to determine an image processing result of the monitoring window. One or more technical schemes provided by the application have at least the following technical effects or advantages: The method comprises the steps of deriving an image frame in a monitoring window through an SDK driving interface of a target dynamic monitor to obtain a monitoring image frame sequence, traversing the monitoring image frame sequence to score the image quality according to a preset scoring index, marking according to a scoring result to obtain a marking monitoring image frame sequence, traversing the marking monitoring image frame sequence to perform semantic conversion analysis and multi-scale feature recognition to obtain a structural description text sequence and a marking multi-scale feature group sequ