Video image intelligent analysis method and system based on big data technology

Abstract

The invention discloses a video image intelligent analysis method and system based on a big data technology. The method specifically comprises the following steps: S1, constructing a framework; s2, performing time sequence arrangement; s3, target screening; s4, correlation analysis; the invention relates to the technical field of video image analysis. According to the video image intelligent analysis method and system based on the big data technology, after a main target and an associated target are determined in advance, a corresponding image recognition model is built, and after frame extraction is carried out on video data under a time sequence, a target analysis frame with main target information and associated target information is obtained; through a video recombination mode, convenient and efficient analysis of the association characteristics of the main target and the association target is realized, the analysis processing efficiency of the video data is effectively improved, user-defined recombination of the association video can be carried out, the convenience is good, and the display effect is more flexible and comprehensive.

Description

Technical Field

[0001] The present invention relates to the technical field of video image analysis, and in particular to a video image intelligent analysis method and system based on big data technology. Background Art

[0002] Big data technology has been widely used to analyze customer behavior in industries such as advertising and e-commerce, but its application in the field of video surveillance has been very slow. The main reason is that the core data in the field of video surveillance is video stream, which is unstructured data and cannot be directly analyzed and processed using mature big data technology. Although the number of industries using video resources has increased, the utilization rate is not high, and the efficiency of manual viewing and analysis of videos is low. For example, a method and system for intelligent analysis of video images based on big data technology as described in application number 202311826503.X obtains the information of each frame of the image itself and the correlation information between each frame of the image, and establishes a video recognition model based on the information of each frame of the image itself and the correlation information between each frame of the image, so as to realize intelligent analysis of video resources.

[0003] Based on the retrieval of the above information, it can be seen that the existing analysis of video resources has a large amount of data processing and is easily affected by background interference. In addition, when performing target association analysis based on the video stream, there is a high data storage intensity and it is impossible to quickly obtain target association information. For this reason, a video image intelligent analysis method and system based on big data technology is proposed, which can realize the rapid capture of targets and their related information from the video stream. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides a video image intelligent analysis method and system based on big data technology, which solves the problems that the existing analysis of video resources has a large amount of data processing and is easily affected by background interference, and when performing target association analysis based on the video stream, there is a high data storage intensity and the target association information cannot be quickly obtained.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a video image intelligent analysis method based on big data technology, specifically comprising the following steps: S1. Framework building: determine the target to be analyzed as the main target, determine the targets related to the target as the associated targets, and establish a distributed processing framework of the main target and several associated targets as the big data processing framework; S2, time sequence arrangement: read the video data frames, mark the frames in time sequence, and obtain the original frame data corresponding to the time mark; S3, target screening: the original frame data is segmented and imported into the big data processing framework, and several frame images with main targets and associated targets are screened as target analysis frames; S4. Association analysis: reorganize the target analysis frames according to the time sequence, and analyze the target position association characteristics based on the reorganized video.

[0006] The present invention is further configured as follows: the establishment method of the big data processing framework in S1 includes: According to the relevant structural features of the main target and several related targets, image recognition models of the main target and several related targets are constructed, and a main target recognition model and several related target recognition models based on the image recognition model are built.

[0007] The present invention is further configured as follows: the method for acquiring the target analysis frame in S3 includes: The main target is taken as the main region of interest, and the associated target is taken as the secondary region of interest. After the main target recognition model performs the main region of interest on the frame, the associated target recognition model performs the secondary region of interest on the frame. Obtain edge coordinates of the primary region of interest and the secondary region of interest, and add a primary target mark to the frame image with the primary region of interest, and add a secondary target mark to the frame image with the secondary region of interest; Frame images with primary target markers and frame images with secondary target markers are screened, and background images are removed to serve as target analysis frames.

[0008] The present invention is further configured as follows: the recombination method in S4 includes: Reconstruct the target analysis frames with the main target in time sequence to obtain the main target trajectory; The target analysis frames with associated targets are reassembled according to the time series to obtain the associated target trajectory.

[0009] The present invention is further configured as follows: the target position correlation characteristic analysis method in S4 includes: After selecting and setting the associated target and obtaining the target associated trajectory, a judgment is made based on the intersection, overlap, inclusion and separation states of the edge coordinates of the secondary region of interest corresponding to the associated target and the edge coordinates of the primary region of interest corresponding to the primary target. The judgment result is the target position association characteristic.

[0010] The present invention is further configured as follows: the method for acquiring the target associated trajectory includes: The time series mark corresponding to the main target trajectory is used as the target time series. After obtaining the associated target trajectory of the set associated target, the associated target trajectory is intercepted according to the target time series. The intercepted associated target trajectory is then reassembled with the main target trajectory to obtain the target associated trajectory.

[0011] The present invention also discloses a video image intelligent analysis system based on big data technology, comprising a data input module, a big data processing layer and an output adjustment layer, wherein the data input module is connected to the big data processing layer, and the big data processing layer is connected to the output adjustment layer; The data input module is used to read the video data frames, mark the frames in time series, and obtain the original frame data corresponding to the time mark; The big data processing layer is used to build a main target recognition model and several related target recognition models based on the image recognition model according to the relevant structural features of the main target and several related targets, and perform target recognition on the original frame data to obtain a target analysis frame; The output adjustment layer is used to reorganize the target analysis frames according to the time sequence, and perform target position correlation characteristic analysis based on the reorganized video.

[0012] The present invention is further configured as follows: the big data processing layer includes a main target recognition module and a plurality of associated target recognition modules; The main target recognition module is used as a main target recognition model to recognize and extract the corresponding main region of interest of the frame; The plurality of associated target recognition modules are all used as associated target recognition models, and are respectively used to recognize and extract secondary regions of interest corresponding to different associated targets in the frame.

[0013] The present invention is further configured as follows: the output adjustment layer includes a reorganization analysis module and an information storage module; The reorganization analysis module is used to reorganize the target analysis frames with the main target according to the time sequence to obtain the main target trajectory; It is used to reorganize the target analysis frames with associated targets in time sequence to obtain the associated target trajectory; It is also used to mark the time series corresponding to the main target trajectory as the target time series, and after obtaining the associated target trajectory of the set associated target, intercept the associated target trajectory according to the target time series, and then reassemble the intercepted associated target trajectory with the main target trajectory to obtain the target associated trajectory; The information storage module is used to store the main target trajectory, the associated target trajectory and the target associated trajectory.

[0014] The present invention provides a video image intelligent analysis method and system based on big data technology. It has the following beneficial effects: The present invention determines the main target and the related targets in advance, builds a corresponding image recognition model, extracts frames of video data in a time series, obtains target analysis frames with main target information and related target information, and then realizes convenient and efficient analysis of the correlation characteristics of the main target and the related targets through video reorganization, effectively improves the analysis and processing efficiency of video data, and can perform customized reorganization of related videos, which has good convenience and more flexible and comprehensive display effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the process of the present invention; Figure 2 Schematic diagram of the system architecture of the present invention. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0017] See also Figure 1-2 The embodiment of the present invention provides the following technical solutions: a video image intelligent analysis system based on big data technology, as shown in the attached Figure 2 As shown, it includes a data input module, a big data processing layer and an output adjustment layer. The data input module is used to read the video data frames, mark the frames with time series, and obtain the original frame data corresponding to the time mark.

[0018] As a preferred solution, in order to improve the regular and orderly processing of the original frame data, the data input module is connected to the big data processing layer. The big data processing layer is composed of a main target recognition module and several associated target recognition modules, which are used to build a main target recognition model and several associated target recognition models based on the image recognition model according to the relevant structural features of the main target and the several associated targets. Specifically, the main target recognition module is used as a main target recognition model to identify and extract the corresponding main interest region of the frame, and the main target is used as the main interest region. After the main target recognition model performs the main interest region on the frame, the edge coordinates of the main interest region are obtained, and the main target mark is added to the frame image with the main interest region; Several associated target recognition modules are used as associated target recognition models, which are respectively used to identify and extract secondary regions of interest corresponding to different associated targets in the frame. Taking the associated target as the secondary region of interest, the associated target recognition model performs secondary regions of interest on the frame, obtains the edge coordinates of the secondary region of interest, adds secondary target tags to the frame images with the secondary region of interest, filters the frame images with the main target tags and the frame images with the secondary target tags, and obtains the target analysis frame after removing the background image.

[0019] As a preferred solution, in order to realize the correlation analysis of the target, the big data processing layer is connected to the output adjustment layer. The output adjustment layer is composed of a reorganization analysis module and an information storage module, which is used to reorganize the target analysis frame according to the time sequence. Specifically, the reorganization analysis module is used to reorganize the target analysis frame with the main target according to the time sequence to obtain the main target trajectory; it is used to reorganize the target analysis frame with the associated target according to the time sequence to obtain the associated target trajectory; the information storage module is used to store the main target trajectory, the associated target trajectory and the target associated trajectory.

[0020] According to the reorganized video, the target position association characteristics are analyzed, and the associated target is selected and set. The time series corresponding to the main target trajectory is marked as the target time series. After the associated target trajectory of the set associated target is obtained, the associated target trajectory is intercepted according to the target time series, and then the intercepted associated target trajectory and the main target trajectory are reorganized through video. After obtaining the target associated trajectory, the intersection, overlap, inclusion and separation states of the edge coordinates of the secondary area of ​​interest corresponding to the set associated target and the edge coordinates of the main area of ​​interest corresponding to the main target are judged. The judgment result is the target position association characteristic. Embodiment 1

[0021] A video image intelligent analysis method based on big data technology specifically includes the following steps: S1. Framework building: Determine the target to be analyzed as the main target, and build a main target recognition model based on the image recognition model according to the relevant structural characteristics of the main target; S2, time sequence arrangement: read the video data frames, mark the frames in time sequence, and obtain the original frame data corresponding to the time mark; S3, target screening: taking the main target as the main region of interest, the main target recognition model identifies the main region of interest for the frame, obtains the edge coordinates of the main region of interest, adds the main target mark to the frame image with the main region of interest, screens the frame image with the main target mark, and removes the background image as the target analysis frame; S4. Association analysis: reorganize the target analysis frames with the main target in time sequence to obtain the main target trajectory.

[0022] This embodiment provides support for continuous tracking of the target. Embodiment 2

[0023] A video image intelligent analysis method based on big data technology specifically includes the following steps: S1. Framework construction: determine the target to be analyzed as the main target, determine the target related to the target as the associated target, establish a distributed processing framework for the main target and several associated targets, build image recognition models for the main target and several associated targets based on the relevant structural features of the main target and several associated targets, and build a main target recognition model and several associated target recognition models based on the image recognition model as a big data processing framework; S2, time sequence arrangement: read the video data frames, mark the frames in time sequence, and obtain the original frame data corresponding to the time mark; S3, target screening: after the original frame data is segmented, the main target is taken as the main region of interest, and the associated target is taken as the secondary region of interest. After the main target recognition model performs the main region of interest on the frame, the associated target recognition model performs the secondary region of interest on the frame, obtains the edge coordinates of the main region of interest and the secondary region of interest, and adds the main target mark to the frame image with the main region of interest, adds the secondary target mark to the frame image with the secondary region of interest, screens the frame image with the main target mark and the frame image with the secondary target mark, and removes the background image as the target analysis frame, wherein the original frame data imported into the target recognition model is unsegmented data; S4, association analysis: reorganize the target analysis frames with the main target according to the time sequence to obtain the main target trajectory; reorganize the target analysis frames with the associated targets according to the time sequence to obtain the associated target trajectory; after the main target trajectory and several associated target trajectories are stored in the information storage module, select and set the associated target, mark the time series corresponding to the main target trajectory as the target time series, and after obtaining the associated target trajectory of the set associated target, intercept the associated target trajectory according to the target time series, and then reorganize the intercepted associated target trajectory with the main target trajectory to obtain the target associated trajectory, and judge according to the intersection, overlap, inclusion and separation status of the edge coordinates of the secondary area of ​​interest corresponding to the set associated target and the edge coordinates of the main area of ​​interest corresponding to the main target, and the judgment result is the target position association characteristic.

[0024] In this embodiment, analysis of associated targets may be performed during the continuous tracking of targets.

[0025] The target position correlation characteristic can also be analyzed by the distance between the associated target and the main target.

[0026] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0027] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A video image intelligent analysis method based on big data technology, characterized by: The specific steps include: S1. Framework building: determine the target to be analyzed as the main target, determine the targets related to the target as the associated targets, and establish a distributed processing framework of the main target and several associated targets as the big data processing framework; S2, time sequence arrangement: read the video data frames, mark the frames in time sequence, and obtain the original frame data corresponding to the time mark; S3, target screening: the original frame data is segmented and imported into the big data processing framework, and several frame images with main targets and associated targets are screened as target analysis frames; S4. Association analysis: reorganize the target analysis frames according to the time sequence, and analyze the target position association characteristics based on the reorganized video.

2. The method for intelligent video image analysis based on big data technology according to claim 1, characterized in that: The establishment method of the big data processing framework in S1 includes: According to the relevant structural features of the main target and several related targets, image recognition models of the main target and several related targets are constructed, and a main target recognition model and several related target recognition models based on the image recognition model are built.

3. The method for intelligent video image analysis based on big data technology according to claim 1, characterized in that: The method for acquiring the target analysis frame in S3 includes: The main target is taken as the main region of interest, and the associated target is taken as the secondary region of interest. After the main target recognition model performs the main region of interest on the frame, the associated target recognition model performs the secondary region of interest on the frame. Obtain edge coordinates of the primary region of interest and the secondary region of interest, and add a primary target mark to the frame image with the primary region of interest, and add a secondary target mark to the frame image with the secondary region of interest; Frame images with primary target markers and frame images with secondary target markers are screened, and background images are removed to serve as target analysis frames.

4. The method for intelligent video image analysis based on big data technology according to claim 1, characterized in that: The reorganization method in S4 includes: Reconstruct the target analysis frames with the main target in time sequence to obtain the main target trajectory; The target analysis frames with associated targets are reassembled according to the time series to obtain the associated target trajectory.

5. The method for intelligent video image analysis based on big data technology according to claim 4 is characterized in that: The target position correlation characteristic analysis method in S4 includes: After selecting and setting the associated target and obtaining the target associated trajectory, a judgment is made based on the intersection, overlap, inclusion and separation states of the edge coordinates of the secondary region of interest corresponding to the associated target and the edge coordinates of the primary region of interest corresponding to the primary target. The judgment result is the target position association characteristic.

6. The method and system for intelligent video image analysis based on big data technology according to claim 5, characterized in that: The method of obtaining the target associated trajectory includes: The time series mark corresponding to the main target trajectory is used as the target time series. After obtaining the associated target trajectory of the set associated target, the associated target trajectory is intercepted according to the target time series. The intercepted associated target trajectory is then reassembled with the main target trajectory to obtain the target associated trajectory.

7. A video image intelligent analysis system based on big data technology, characterized by: It includes a data input module, a big data processing layer and an output adjustment layer, wherein the data input module is connected to the big data processing layer, and the big data processing layer is connected to the output adjustment layer; The data input module is used to read the video data frames, mark the frames in time series, and obtain the original frame data corresponding to the time mark; The big data processing layer is used to build a main target recognition model and several related target recognition models based on the image recognition model according to the relevant structural features of the main target and several related targets, and perform target recognition on the original frame data to obtain a target analysis frame; The output adjustment layer is used to reorganize the target analysis frames according to the time sequence, and perform target position correlation characteristic analysis based on the reorganized video.

8. The video image intelligent analysis system based on big data technology according to claim 7 is characterized by: The big data processing layer includes a main target recognition module and several associated target recognition modules; The main target recognition module is used as a main target recognition model to recognize and extract the corresponding main region of interest of the frame; The plurality of associated target recognition modules are all used as associated target recognition models, and are respectively used to recognize and extract secondary regions of interest corresponding to different associated targets in the frame.

9. The video image intelligent analysis system based on big data technology according to claim 7 is characterized in that: The output adjustment layer includes a reorganization analysis module and an information storage module; The reorganization analysis module is used to reorganize the target analysis frames with the main target according to the time sequence to obtain the main target trajectory; It is used to reorganize the target analysis frames with associated targets in time sequence to obtain the associated target trajectory; It is also used to mark the time series corresponding to the main target trajectory as the target time series, and after obtaining the associated target trajectory of the set associated target, intercept the associated target trajectory according to the target time series, and then reassemble the intercepted associated target trajectory with the main target trajectory to obtain the target associated trajectory; The information storage module is used to store the main target trajectory, the associated target trajectory and the target associated trajectory.

Images