CN118839031A - Multi-level file index management method for video data - Google Patents

Abstract

The present invention discloses a multi-level file index management method for video data, and relates to the technical field of file index management. The method comprises the following steps: obtaining video data corresponding to a plurality of video sources, performing frame cutting on all the video data in sequence, and then dividing the video data corresponding to each video source into a plurality of video frames corresponding to each of the video sources, performing feature processing on the plurality of video frames corresponding to the video data of each video source, and then obtaining feature frame information corresponding to each video data, constructing a multi-level index file, mapping the video data corresponding to all the video sources into the multi-level index file, and then constructing a hierarchical retrieval index corresponding to each video data, configuring a query platform, importing the hierarchical retrieval index of all the video data into a terminal database corresponding to the query platform, inputting retrieval information of the video data to be retrieved into the query platform, performing index matching based on the retrieval information and the hierarchical retrieval index, and then locating the target video data.

Description

A multi-level file index management method for video data

Technical Field

The invention relates to the technical field of file index management, and in particular to a multi-level file index management method for video data.

Background Art

Multi-level file index management refers to a method of organizing and managing file index structures at different levels. In this management method, the index is not just a simple first-level index, that is, a basic file name or attribute index, but is constructed and managed in a hierarchical manner based on the characteristics and attributes of the file.

With the massive generation and application of digital video data, how to efficiently manage and retrieve video files has become one of the key contents of file index management technology research. Traditional file systems and database management methods generally have problems such as low retrieval efficiency and low retrieval accuracy when facing large-scale video data. Therefore, a management method that combines multi-level file index management technology is needed to achieve more efficient and accurate retrieval of video data.

Summary of the invention

In order to solve the above problems, the object of the present invention is to provide a multi-level file index management method for video data.

The object of the present invention can be achieved by the following technical solution: A multi-level file index management method for video data, comprising the following steps:

Step S1: obtaining video data corresponding to a plurality of video sources, performing frame cutting on all the video data in sequence, and then dividing the video data corresponding to each video source into a plurality of video frames corresponding to each of the video sources, performing feature processing on a plurality of video frames corresponding to the video data of each video source, and then obtaining feature frame information corresponding to each video data;

Step S2: construct a multi-level index file, map the video data corresponding to all video sources into the multi-level index file, and then construct a hierarchical retrieval index corresponding to each video data;

Step S3: Configure the query platform, import the hierarchical retrieval index of all video data into the terminal database corresponding to the query platform, input the retrieval information of the video data to be retrieved into the query platform, perform index matching based on the retrieval information and the hierarchical retrieval index, and then locate the target video data.

Furthermore, the process of obtaining video data corresponding to a plurality of video sources, performing frame segmentation on all the video data in sequence, and then dividing the video data corresponding to each video source into a plurality of video frames corresponding to each of the video sources includes:

A number of video sources are numbered and recorded as i, i=1, 2, 3, ..., n, n is a natural number greater than 0, the video source numbered i includes a number of video data, all the video data included in the video source numbered i are constructed into a video set and recorded as Ω{i}, all the video data included in Ω{i} are numbered, each video data is recorded as VD[i][1], VD[i][2], VD[i][3], ..., VD[i][m], m is a natural number greater than 0, all the video data corresponding to all the video sources are obtained, starting from the video source i=1, VD[1][1], VD[1][2], ..., VD[1][m] included in the video source are frame-cut in turn, and the above operation is repeated until the frame cutting of all the video data corresponding to the video source i=n is completed;

The video data included in each video source is divided into several corresponding video frames through frame cutting, and the number of the video source where the video frame is located, the number of the video data where the video frame is located, and the corresponding timestamp on the video data where the video frame is located are combined as the frame identifier of the current video frame.

Furthermore, the process of performing feature processing on a plurality of video frames of video data corresponding to each video source, and then obtaining feature frame information corresponding to each video data includes:

Feature processing of all video frames includes feature extraction and feature vector construction;

Obtain the color features, edge features, and texture features of each video frame through feature extraction;

The feature vector is constructed as follows: a vector construction model for vectorization processing is constructed by using convolutional neural network technology, and the video frames with the highest color feature ratio, the highest edge feature ratio, and the highest texture feature ratio in the video frames included in each video data are used as the first vector parameter, the second vector parameter, and the third vector parameter of the vector construction model respectively;

Inputting the first vector parameter, the second vector parameter and the third vector parameter into the vector construction model, thereby constructing a conventional feature vector corresponding to the color feature, a conventional feature vector corresponding to the edge feature and a conventional feature vector corresponding to the texture feature, and setting a vector distinction angle range of each conventional feature vector, wherein the vector distinction angle range includes a first positive interval range, a second positive interval range, a first negative interval range and a second negative interval range;

Input the color features, edge features and texture features corresponding to all video frames into the vector construction model, and then obtain the color feature, edge feature and texture feature feature vectors corresponding to each video frame, obtain the angle between the feature vector of the color feature corresponding to each video frame and the corresponding conventional feature vector, the angle between the feature vector of the edge feature corresponding to each video frame and the corresponding conventional feature vector, and the angle between the feature vector of the texture feature corresponding to each video frame and the corresponding conventional feature vector;

When the angle degree is in the first positive interval range, the corresponding feature vector is marked as a Class A feature vector; when the angle degree is in the second positive interval range, the corresponding feature vector is marked as a Class B feature vector; when the angle degree is in the first negative interval range, the corresponding feature vector is marked as a Class C feature vector; when the angle degree is in the second negative interval range, the corresponding feature vector is marked as a Class D feature vector. All the marked feature vectors of the video frames corresponding to each video data are used as the feature frame information of the corresponding video data.

Furthermore, the process of building a multi-level index file includes:

Obtain source address information corresponding to each video source, video information summaries corresponding to all video data included in each video source, and frame information summaries of video frames included in each video data;

The source address information includes address IP, address gateway and address digital sequence;

The video information summary is used to record the video compression information corresponding to each video data;

The frame information summary includes frame length, frame resolution and frame rate;

A top-level index structure is constructed according to the source address information corresponding to the video source, a middle-level index structure equal to the number of current video data under the top-level index structure is constructed according to the video information summary corresponding to each video data under the video source, and a bottom-level index structure equal to the number of current video frames under the middle-level index structure is constructed according to the frame information summary corresponding to each video frame under the video data;

A corresponding number of blank files are allocated according to the number of top-level index structures, and the middle-level index structure and the bottom-level index structure corresponding to the top-level index structure are used as record items of the blank files, thereby constructing a general index file corresponding to each video source, setting an index traversal starting point, and linking the general index file corresponding to each video source to the index traversal starting point, thereby constructing a multi-level index file for traversing all video sources, all video data included in the video sources, and all video frames included in the video data.

Furthermore, the process of mapping the video data corresponding to all video sources into a multi-level index file and then constructing a hierarchical retrieval index corresponding to each video data includes:

Map each video source to the top-level index structure in the multi-level index file, and use the number of each video source as the top-level index; map all video data included in each video source to the middle-level index structure in the multi-level index file, and use the number of each video data as the middle-level index; map all video frames included in each video data to the bottom-level index structure in the multi-level index file, and use the timestamp of each video frame as the bottom-level index; integrate the top-level index, middle-level index and bottom-level index to construct a hierarchical retrieval index corresponding to each video data.

Furthermore, the process of configuring the query platform includes:

The query platform sends a connection request to the preset server, and the server accepts the connection request and determines whether the IP address of the connection request is in the address table set by the server;

If yes, the query platform is linked to the server;

If not, it is determined that there is a system risk in the current query platform, and the platform log that records the current system risk is obtained, and several log record items included in the platform log are input into the preset risk library. The risk library matches the specific risk behavior and the corresponding risk details information according to the content of each log record item, and simultaneously matches the risk repair parameters corresponding to each risk detail information, and repairs each specific risk behavior through the corresponding risk repair parameters, thereby repairing the system risk existing in the current query platform, and continuing the link between the current query platform and the server after the system risk repair is completed;

When the query platform and the server are linked successfully, the query platform configuration is successful, otherwise, the repair of specific risk behaviors and the link between the query platform and the server are continued until the query platform and the server are linked successfully.

Furthermore, the process of importing the hierarchical search index of all video data into the terminal database corresponding to the query platform includes:

The query platform is provided with a corresponding terminal database, and the hierarchical retrieval index of the video data corresponding to all video sources is imported into the terminal database, and an upload interface corresponding to the query platform is allocated to each video source, and each upload interface is correspondingly associated with an interface sequence, and the hierarchical retrieval index of all video data included in each video source is imported into the terminal database through the corresponding upload interface;

Before importing, determine whether the interface sequence corresponding to each upload interface has been tampered with. If so, encapsulate the hierarchical retrieval index of all video data of the corresponding upload interface into a blank file, mark the blank file as an abnormal file, and import the abnormal file into the data security area set corresponding to the terminal database. If not, directly import the hierarchical retrieval index of all video data of the upload interface into the data storage area set by the terminal database for storage;

In the data security area, the layered retrieval index of each video data included in the abnormal file is traversed, and each layered retrieval index is determined in turn whether it has been changed. If so, it means that there is an error in the layered retrieval index, and the current layered retrieval index is updated to the correct layered retrieval index before the change. If not, no operation is performed on the current layered retrieval index until the update of all the layered retrieval indexes with errors in the abnormal file is completed. When all the layered retrieval indexes with errors in the abnormal file are updated, the abnormal file is marked as a safe file, and all the layered retrieval indexes included in the safe file are transferred and stored in the data storage area.

Furthermore, the process of inputting search information of the video data to be searched into the query platform, performing index matching with the hierarchical search index according to the search information, and locating the target video data includes:

The retrieval information includes first-class information and second-class information. The first-class information is used to locate the target video data. The first-class information includes the top-level index and the middle-level index of the currently required retrieved video data. The second-class information is used to locate the single point of several video frames included in the target video data. The second-class information includes the bottom-level index of all video frames corresponding to the current target video data.

The edited retrieval information is input into the query platform, and the index is matched with the hierarchical retrieval index of the query platform. The hierarchical retrieval index that satisfies the top-level index and the middle-level index corresponding to the first category of information in the terminal database is used as the target index, and the video data corresponding to the target index is located as the target video data. After the target video data is located, several corresponding video frames corresponding to the current target video data are searched according to the bottom-level index corresponding to the second category of information.

Compared with the prior art, the beneficial effects of the present invention are: obtaining video data corresponding to several video sources and performing frame cutting in sequence, dividing the video data corresponding to each video source into several corresponding video frames, performing feature processing on all video frames of the video data corresponding to each video source, and then obtaining feature frame information corresponding to each video data, constructing a multi-level index file, mapping the video data corresponding to all video sources into the multi-level index file, and then constructing a hierarchical retrieval index corresponding to each video data, importing the hierarchical retrieval index of all video data into a terminal database set up by a configured query platform, inputting retrieval information of the video data to be retrieved into the query platform, performing index matching based on the retrieval information and the hierarchical retrieval index, and then locating the target video data, thereby realizing efficient and accurate retrieval of video data to a certain extent.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a flow chart of the present invention.

DETAILED DESCRIPTION

As shown in FIG1 , a multi-level file index management method for video data includes the following steps:

Step S1: obtaining video data corresponding to a plurality of video sources, performing frame cutting on all the video data in sequence, and then dividing the video data corresponding to each video source into a plurality of video frames corresponding to each of the video sources, performing feature processing on a plurality of video frames corresponding to the video data of each video source, and then obtaining feature frame information corresponding to each video data;

Step S2: construct a multi-level index file, map the video data corresponding to all video sources into the multi-level index file, and then construct a hierarchical retrieval index corresponding to each video data;

Step S3: Configure the query platform, import the hierarchical retrieval index of all video data into the terminal database corresponding to the query platform, input the retrieval information of the video data to be retrieved into the query platform, perform index matching based on the retrieval information and the hierarchical retrieval index, and then locate the target video data.

It should be further explained that, in a specific implementation process, the process of obtaining video data corresponding to a plurality of video sources, performing frame cutting on all the video data in sequence, and then dividing the video data corresponding to each video source into a plurality of corresponding video frames includes:

A number of video sources are numbered and recorded as i, then i=1, 2, 3, ..., n, where n is a natural number greater than 0, and the video source numbered i includes a number of video data, and all the video data corresponding to the video source numbered i are constructed into a corresponding video set, which is recorded as Ω{i};

Number all video data included in the video set Ω{i}, each video data is recorded as VD[i][1], VD[i][2], VD[i][3], ..., VD[i][m], where m is a natural number greater than 0, obtain all video data corresponding to all video sources, start from the video source i=1, and sequentially perform frame cutting on VD[1][1], VD[1][2], ..., VD[1][m] included in the video source. After all video data corresponding to the current video source are processed, continue to perform frame cutting on all video data corresponding to the video source i=2, that is, perform frame cutting on video data VD[2][1], VD[2][2], ..., VD[2][m]. Repeat the above operation to perform frame cutting on all video data corresponding to the video sources i=3-n, wherein the video data corresponding to the video source i=n includes VD[n][1], VD[n][2], VD[n][3], ..., VD[n][m].

Through frame cutting, the video data included in each video source is divided into several corresponding video frames. The number of the video source where the video frame is located, the number of the video data where the video frame is located, and the corresponding timestamp on the video data where the video frame is located are combined as the frame identifier of the current video frame. The frame identifier is denoted as Z, and the timestamp is denoted as T, then Z=i-m-T.

It should be noted that Z=i-m-T means that the m-th video data in the ith video source corresponds to a video frame with a timestamp of T, the video data includes several video frames, the length of the video frame is in seconds: S, the length of the video frame is set to 2S, and the start time of each video frame is used as the timestamp of the corresponding video frame.

It should be further explained that, in a specific implementation process, the process of performing feature processing on a plurality of video frames of video data corresponding to each video source and then obtaining feature frame information corresponding to each video data includes:

Feature processing of all video frames includes feature extraction and feature vector construction;

The color feature, edge feature and texture feature of each video frame are obtained by feature extraction, wherein the color feature is obtained by calculating the color histogram corresponding to the video frame, the edge feature is obtained by using the Canny edge detection algorithm, and the texture feature is obtained by using the local binary pattern;

The content of the feature vector construction is as follows: construct a vector construction model for vectorization processing by using the convolutional neural network technology, take the color feature with the highest proportion in the video frame included in each video data as the first vector parameter corresponding to the vector construction model, take the edge feature with the highest proportion in the video frame included in each video data as the second vector parameter corresponding to the vector construction model, take the texture feature with the highest proportion in the video frame included in each video data as the third vector parameter corresponding to the vector construction model;

Inputting the first vector parameter, the second vector parameter and the third vector parameter into the vector construction model, thereby constructing a conventional feature vector corresponding to the color feature, a conventional feature vector corresponding to the edge feature and a conventional feature vector corresponding to the texture feature;

Setting a vector distinguishing angle range of a conventional feature vector corresponding to a color feature, an edge feature, and a texture feature, wherein the vector distinguishing angle range includes a first positive interval range, a second positive interval range, a first negative interval range, and a second negative interval range;

Among them, the first positive interval range is the 0°-90° angle range on the right side of the conventional eigenvector, the second positive interval range is the 91°-80° angle range on the right side of the conventional eigenvector, the first negative interval range is the 0°-90° angle range on the left side of the conventional eigenvector, and the second negative interval range is the 91°-180° angle range on the left side of the conventional eigenvector;

It should be noted that the vector discrimination angle ranges of the color features, edge features and texture features corresponding to the conventional feature vectors are all as described above;

Input the color features, edge features and texture features corresponding to all video frames into the vector construction model, and then obtain the color feature, edge feature and texture feature feature vectors corresponding to each video frame, obtain the angle between the feature vector of the color feature corresponding to each video frame and the corresponding conventional feature vector, obtain the angle between the feature vector of the edge feature corresponding to each video frame and the corresponding conventional feature vector, and obtain the angle between the feature vector of the texture feature corresponding to each video frame and the corresponding conventional feature vector;

When the angle is within the first positive interval, the corresponding eigenvector is marked as a type A eigenvector; when the angle is within the second positive interval, the corresponding eigenvector is marked as a type B eigenvector;

When the angle is within the first negative interval, the corresponding eigenvector is marked as a C-type eigenvector; when the angle is within the second negative interval, the corresponding eigenvector is marked as a D-type eigenvector;

All the marked feature vectors of the video frames corresponding to each video data are used as the feature frame information of the corresponding video data.

It should be further explained that, in the specific implementation process, the process of building a multi-level index file includes:

Obtain source address information corresponding to each video source, obtain video information summaries corresponding to all video data included in each video source, and obtain frame information summaries of video frames included in each video data;

The source address information is used to record the source of each video source, and the source address information includes the address IP, address gateway and address digital sequence;

The video information summary is used to record the video compression information corresponding to each video data, and the video compression information is used to reflect the overview content of each video data;

The frame information summary includes frame length, frame resolution and frame rate;

A top-level index structure is constructed according to the source address information corresponding to the video source, a middle-level index structure corresponding to the top-level index structure and equal to the number of current video data is constructed according to the video information summary corresponding to each video data under the video source, and a bottom-level index structure corresponding to the middle-level index structure and equal to the number of current video frames is constructed according to the frame information summary corresponding to each video frame under the video data;

A corresponding number of blank files are allocated according to the number of top-level index structures, and the middle-level index structure and the bottom-level index structure included in the top-level index structure are used as record items of the blank files, thereby constructing a total index file corresponding to each video source;

An index traversal starting point is set, and a general index file corresponding to each video source is linked to the index traversal starting point, thereby constructing a multi-level index file for traversing all video sources, all video data included in the video sources, and all video frames included in the video data.

It should be further explained that, in a specific implementation process, the process of mapping the video data corresponding to all video sources into a multi-level index file and then constructing a hierarchical retrieval index corresponding to each video data includes:

Map each video source to the top-level index structure in the multi-level index file in the order of the number from small to large, and use the number of each video source as the top-level index, map all video data included in each video source to the middle-level index structure in the multi-level index file in the order from left to right and from top to bottom, and use the number of each video data as the middle-level index, map all video frames included in each video data to the bottom-level index structure in the multi-level index file, and use the timestamp of each video frame as the bottom-level index;

The top-level index, the middle-level index and the bottom-level index are integrated to construct a hierarchical retrieval index corresponding to each video data, and the hierarchical retrieval index is used for subsequent rapid retrieval of the video data.

It should be further explained that, in the specific implementation process, the process of configuring the query platform includes:

A query platform is set up to query the video data corresponding to different video sources and the video frames corresponding to each video data. The query platform sends a link request to a preset server, and the server receives the link request and determines whether the IP address of the link request is in the address form set corresponding to the server;

If yes, the query platform is linked to the server;

If not, it is determined that there is a system risk in the current query platform, and the platform log that records the current system risk is obtained, and several log record items included in the platform log are input into the preset risk library. The risk library matches the specific risk behavior and the corresponding risk details information according to the content of each log record item, and simultaneously matches the risk repair parameters corresponding to each risk detail information, and repairs each specific risk behavior through the corresponding risk repair parameters, thereby repairing the system risk existing in the current query platform, and continuing the link between the current query platform and the server after the system risk repair is completed;

When the query platform and the server are linked successfully, the query platform configuration is successful, otherwise, the repair of specific risk behaviors and the link between the query platform and the server are continued until the query platform and the server are linked successfully.

It should be further explained that, in the specific implementation process, the process of importing the hierarchical retrieval index of all video data into the terminal database corresponding to the query platform includes:

The query platform is provided with a corresponding terminal database, a data import period corresponding to the terminal database is set, and the data import period is recorded as _Timport , _Timport = [ _t1 , _t2 ], wherein _t1 is the start time of the data import period, and _t2 is the end time of the data import period;

During the data import period, the hierarchical retrieval index of the video data corresponding to all video sources is imported into the terminal database, an upload interface corresponding to the query platform is allocated to each video source, each upload interface is associated with an interface sequence, and the hierarchical retrieval index of all video data included in each video source is imported into the terminal database through the corresponding upload interface;

Before importing, determine whether the interface sequence corresponding to each upload interface has been tampered with. If so, encapsulate the hierarchical retrieval index of all video data of the corresponding upload interface into a blank file, mark the blank file as an abnormal file, and import the abnormal file into the data security area set corresponding to the terminal database. If not, directly import the hierarchical retrieval index of all video data of the upload interface into the data storage area set by the terminal database for storage;

In the data security area, the hierarchical retrieval index of each video data included in the abnormal file is traversed, and each hierarchical retrieval index is judged in turn whether it has been changed. If so, it means that there is an error in the hierarchical retrieval index, and the current hierarchical retrieval index is updated to be the correct hierarchical retrieval index before the change. If not, no operation is performed on the current hierarchical retrieval index; until the update of all the hierarchical retrieval indexes with errors in the abnormal file is completed;

When all erroneous hierarchical retrieval indexes in the abnormal file are updated, the abnormal file is marked as a safe file, and all hierarchical retrieval indexes included in the safe file are transferred and stored in the data storage area.

It should be further explained that, in the specific implementation process, the process of inputting the search information of the video data to be searched into the query platform, performing index matching with the hierarchical search index according to the search information, and then locating the target video data includes:

Set a search period, and record the search period as _Tsearch , where _Tsearch =[ _tstart , _tend ], where tstart is the _start time of the search period, and _tend is the end time of the search period. At the start time corresponding to _tstart , edit the search information corresponding to the video data to be searched, wherein the search information includes the first type of information and the second type of information;

The first type of information is used to locate the target video data, and the first type of information includes the top-level index and the middle-level index of the video data currently required to be retrieved;

The second type of information is used to perform single-point positioning of a plurality of video frames included in the target video data, and the second type of information includes the bottom-level indexes of all video frames corresponding to the current target video data;

Input the edited search information into the query platform, and then the query platform performs index matching with the hierarchical search index according to the search information, and uses the hierarchical search index that satisfies the top-level index and the middle-level index corresponding to the first type of information in the terminal database as the target index, and locates the video data corresponding to the target index as the target video data;

After the target video data to be retrieved is located, several corresponding video frames corresponding to the current target video data are searched according to the underlying index corresponding to the second type of information, thereby completing the single-point positioning of each video frame.

It should be noted that the first type of information is used to quickly locate the required video data, and the located video data is used as the target video data. The second type of information is used to further locate all video frames corresponding to the target video data, thereby obtaining more detailed video features of the video data.

The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims (8)

1. A multi-level file index management method for video data, characterized in that it comprises the following steps: Step S1: obtaining video data corresponding to a plurality of video sources, performing frame cutting on all the video data in sequence, and then dividing the video data corresponding to each video source into a plurality of video frames corresponding to each of the video sources, performing feature processing on a plurality of video frames corresponding to the video data of each video source, and then obtaining feature frame information corresponding to each video data; Step S2: construct a multi-level index file, map the video data corresponding to all video sources into the multi-level index file, and then construct a hierarchical retrieval index corresponding to each video data; Step S3: Configure the query platform, import the hierarchical retrieval index of all video data into the terminal database corresponding to the query platform, input the retrieval information of the video data to be retrieved into the query platform, perform index matching based on the retrieval information and the hierarchical retrieval index, and then locate the target video data. 2. A multi-level file index management method for video data according to claim 1, characterized in that the process of obtaining video data corresponding to a plurality of video sources, sequentially performing frame cutting on all the video data, and then dividing the video data corresponding to each video source into a plurality of video frames corresponding to each of the video sources comprises: A number of video sources are numbered and recorded as i, i=1, 2, 3, ..., n, n is a natural number greater than 0, the video source numbered i includes a number of video data, all the video data included in the video source numbered i are constructed into a video set and recorded as Ω{i}, all the video data included in Ω{i} are numbered, each video data is recorded as VD[i][1], VD[i][2], VD[i][3], ..., VD[i][m], m is a natural number greater than 0, all the video data corresponding to all the video sources are obtained, starting from the video source i=1, VD[1][1], VD[1][2], ..., VD[1][m] included in the video source are frame-cut in turn, and the above operation is repeated until the frame cutting of all the video data corresponding to the video source i=n is completed; The video data included in each video source is divided into several corresponding video frames through frame cutting, and the number of the video source where the video frame is located, the number of the video data where the video frame is located, and the corresponding timestamp on the video data where the video frame is located are combined as the frame identifier of the current video frame. 3. A multi-level file index management method for video data according to claim 2, characterized in that the process of performing feature processing on a plurality of video frames of video data corresponding to each video source and then obtaining feature frame information corresponding to each video data comprises: Feature processing of all video frames includes feature extraction and feature vector construction; Obtain the color features, edge features, and texture features of each video frame through feature extraction; The feature vector is constructed as follows: a vector construction model for vectorization processing is constructed by using convolutional neural network technology, and the video frames with the highest color feature ratio, the highest edge feature ratio, and the highest texture feature ratio in the video frames included in each video data are used as the first vector parameter, the second vector parameter, and the third vector parameter of the vector construction model respectively; Inputting the first vector parameter, the second vector parameter and the third vector parameter into the vector construction model, thereby constructing a conventional feature vector corresponding to the color feature, a conventional feature vector corresponding to the edge feature and a conventional feature vector corresponding to the texture feature, and setting a vector distinction angle range of each conventional feature vector, wherein the vector distinction angle range includes a first positive interval range, a second positive interval range, a first negative interval range and a second negative interval range; Input the color features, edge features and texture features corresponding to all video frames into the vector construction model, and then obtain the color feature, edge feature and texture feature feature vectors corresponding to each video frame, obtain the angle between the feature vector of the color feature corresponding to each video frame and the corresponding conventional feature vector, the angle between the feature vector of the edge feature corresponding to each video frame and the corresponding conventional feature vector, and the angle between the feature vector of the texture feature corresponding to each video frame and the corresponding conventional feature vector; When the angle degree is in the first positive interval range, the corresponding feature vector is marked as a Class A feature vector; when the angle degree is in the second positive interval range, the corresponding feature vector is marked as a Class B feature vector; when the angle degree is in the first negative interval range, the corresponding feature vector is marked as a Class C feature vector; when the angle degree is in the second negative interval range, the corresponding feature vector is marked as a Class D feature vector. All the marked feature vectors of the video frames corresponding to each video data are used as the feature frame information of the corresponding video data. 4. A multi-level file index management method for video data according to claim 3, characterized in that the process of constructing a multi-level index file comprises: Obtain source address information corresponding to each video source, video information summaries corresponding to all video data included in each video source, and frame information summaries of video frames included in each video data; The source address information includes address IP, address gateway and address digital sequence; The video information summary is used to record the video compression information corresponding to each video data; The frame information summary includes frame length, frame resolution and frame rate; A top-level index structure is constructed according to the source address information corresponding to the video source, a middle-level index structure equal to the number of current video data under the top-level index structure is constructed according to the video information summary corresponding to each video data under the video source, and a bottom-level index structure equal to the number of current video frames under the middle-level index structure is constructed according to the frame information summary corresponding to each video frame under the video data; A corresponding number of blank files are allocated according to the number of top-level index structures, and the middle-level index structure and the bottom-level index structure corresponding to the top-level index structure are used as record items of the blank files, thereby constructing a general index file corresponding to each video source, setting an index traversal starting point, and linking the general index file corresponding to each video source to the index traversal starting point, thereby constructing a multi-level index file for traversing all video sources, all video data included in the video sources, and all video frames included in the video data. 5. A multi-level file index management method for video data according to claim 4, characterized in that the process of mapping the video data corresponding to all video sources into the multi-level index file and then constructing the hierarchical retrieval index corresponding to each video data comprises: Map each video source to the top-level index structure in the multi-level index file, and use the number of each video source as the top-level index; map all video data included in each video source to the middle-level index structure in the multi-level index file, and use the number of each video data as the middle-level index; map all video frames included in each video data to the bottom-level index structure in the multi-level index file, and use the timestamp of each video frame as the bottom-level index; integrate the top-level index, middle-level index and bottom-level index to construct a hierarchical retrieval index corresponding to each video data. 6. A multi-level file index management method for video data according to claim 5, characterized in that the process of configuring the query platform comprises: The query platform sends a connection request to the preset server, and the server accepts the connection request and determines whether the IP address of the connection request is in the address table set by the server; If yes, the query platform is linked to the server; If not, it is determined that there is a system risk in the current query platform, and the platform log that records the current system risk is obtained, and several log record items included in the platform log are input into the preset risk library. The risk library matches the specific risk behavior and the corresponding risk details information according to the content of each log record item, and simultaneously matches the risk repair parameters corresponding to each risk detail information, and repairs each specific risk behavior through the corresponding risk repair parameters, thereby repairing the system risk existing in the current query platform, and continuing the link between the current query platform and the server after the system risk repair is completed; When the query platform and the server are linked successfully, the query platform configuration is successful, otherwise, the repair of specific risk behaviors and the link between the query platform and the server are continued until the query platform and the server are linked successfully. 7. A multi-level file index management method for video data according to claim 6, characterized in that the process of importing the hierarchical retrieval index of all video data into the terminal database corresponding to the query platform comprises: The query platform is provided with a corresponding terminal database, and the hierarchical retrieval index of the video data corresponding to all video sources is imported into the terminal database, and an upload interface corresponding to the query platform is allocated to each video source, and each upload interface is correspondingly associated with an interface sequence, and the hierarchical retrieval index of all video data included in each video source is imported into the terminal database through the corresponding upload interface; Before importing, determine whether the interface sequence corresponding to each upload interface has been tampered with. If so, encapsulate the hierarchical retrieval index of all video data of the corresponding upload interface into a blank file, mark the blank file as an abnormal file, and import the abnormal file into the data security area set corresponding to the terminal database. If not, directly import the hierarchical retrieval index of all video data of the upload interface into the data storage area set by the terminal database for storage; In the data security area, the layered retrieval index of each video data included in the abnormal file is traversed, and each layered retrieval index is determined in turn whether it has been changed. If so, it means that there is an error in the layered retrieval index, and the current layered retrieval index is updated to the correct layered retrieval index before the change. If not, no operation is performed on the current layered retrieval index until the update of all the layered retrieval indexes with errors in the abnormal file is completed. When all the layered retrieval indexes with errors in the abnormal file are updated, the abnormal file is marked as a safe file, and all the layered retrieval indexes included in the safe file are transferred and stored in the data storage area. 8. A multi-level file index management method for video data according to claim 7, characterized in that the process of inputting search information of the video data to be searched into the query platform, performing index matching with the hierarchical search index according to the search information, and locating the target video data comprises: The retrieval information includes first-class information and second-class information. The first-class information is used to locate the target video data. The first-class information includes the top-level index and the middle-level index of the currently required retrieved video data. The second-class information is used to locate the single point of several video frames included in the target video data. The second-class information includes the bottom-level index of all video frames corresponding to the current target video data. The edited retrieval information is input into the query platform, and the index is matched with the hierarchical retrieval index of the query platform. The hierarchical retrieval index that satisfies the top-level index and the middle-level index corresponding to the first category of information in the terminal database is used as the target index, and the video data corresponding to the target index is located as the target video data. After the target video data is located, several corresponding video frames corresponding to the current target video data are searched according to the bottom-level index corresponding to the second category of information.