KR100977417B1 - High-speed searching method for stored video - Google Patents

Abstract

PURPOSE: A rapid search method of the stored video by using a condition filtering is provided to search all video which are not indexed at high speed by performing a conditional filtering and remove the unnecessary region. CONSTITUTION: A demultiplexer and a decoder are loaded after the analysis of a format of a video file. The image surface process according to the resolution of the moving picture is prepared. The interest region and search condition filter are set in one part of the moving picture. A search and matching process is operated about the image of the region of interest according to the search condition filter. The image which is expelled from the region of interest of the moving picture through the search condition filter is outputted to the regular time interval unit to the storage.

Description

High-speed search of saved videos {HIGH-SPEED SEARCHING METHOD FOR STORED VIDEO}

The present invention relates to a fast search method for a stored video, and more particularly, to a fast search method for a stored video capable of performing a fast search for a specific area when searching a video stored in various formats.

In recent years, compressed and stored videos play a very important role in security and surveillance systems that use CCD (Charge Coupled Device) cameras to monitor specific areas or objects.In particular, crimes occur in the area or objects. If you do, it is an important key to provide a clue to the resolution of the case by searching the video stored in the time zone.

In order to search for such a video, a search method for a stored video has been proposed in the past. The method is as follows.

In the conventional method of retrieving a stored video, information of an item (all elements such as object existence, motion, color, size, etc.) appearing in the stored video is first stored as an information-index, and each item is displayed on the screen. Store the start and end positions of the displayed intervals as Lifetime-Index. By linking the information-index and the life-index so that they can be cross-referenced, the information of the items that appear according to the change of the screen is displayed in the same image or in a separate method when the video is played. Use this function to search for a scene or section of a video.

The conventional method for retrieving a stored video stores information of a predetermined item in real time during shooting, so that when an event occurs in the future, the video stored at the corresponding time period can be searched with indexed keywords to quickly confirm what happened in the video.

However, in the above-described conventional method, it is possible to quickly search only the image of the corresponding item part by setting an item to be information-indexed in advance for all images, and it is impossible to perform a fast search for other items or regions. There was this.

In addition, since most surveillance cameras are an old model and rarely store images by information-indexing in real time as in the conventional method described above, the existing video that is not applied to the aforementioned conventional technology cannot be searched. there was.

In addition, the conventional method simultaneously stores the information-index of the corresponding item portion in the video stored in real time, which increases the load on the processor operation, resulting in a decrease in system performance, and the information-index in the stored video. There was a problem that the capacity of the video itself is increased because the index is further included.

SUMMARY OF THE INVENTION An object of the present invention for solving the above problems is not limited to preset items and areas in all existing non-indexed videos, but stored at high speed for retrieving all items and areas in the video. It is to provide a fast search method of a video.

It is another object of the present invention to prevent performance degradation of a system for storing real-time video by excluding a separate process of indexing information during real-time storage, and to provide a high-speed search method of a stored video that can prevent an increase in the capacity of a stored video. To provide.

According to another aspect of the present invention, there is provided a method of fast searching a stored video, comprising: a first step of analyzing a video file format and loading a demuxer and a decoder; and the resolution of the video. And a third step of setting an ROI and a search condition filter for a portion of the video; and a search condition filter only for the image of the ROI set in the third step. And a fifth step of performing a search and matching operation in accordance with the present invention; and a fifth step of storing and outputting the image extracted through a search condition filter in a region of interest of the video at a predetermined time unit. It features.

According to a preferred embodiment, in the first step, only 31767 bytes of a portion of a video file are processed to analyze a video format.

According to an embodiment of the present disclosure, the second step may include a memory preparation step of preparing a memory for receiving an image according to the output resolution of the decoder and a memory for outputting an image of a fast postprocessor (FP); Characterized in that it comprises a video information delivery step for delivering the format information.

According to a preferred embodiment, in the third step, the region of interest is designated by dragging a certain region of the video with a cursor.

According to a preferred embodiment, the fourth step includes a frame extraction step of extracting a key frame after detecting a shot boundary frame of the ROI, and comparing the key frame with a feature value of a search condition filter. Characterized in that it comprises a feature comparison step.

According to an embodiment of the present disclosure, in the fourth step, a search and matching operation may be performed by converting the color region of the video to a black and white (Gray) format.

According to a preferred embodiment, in the fourth step, a search and matching operation is performed by reducing a pixel block of a predetermined size by ½.

According to a preferred embodiment, in the fifth step, the image extracted through the search condition filter in the ROI is captured by a predetermined time unit, converted into a still image file, and the index information of each file is recorded. .

The fast search method for the stored video of the present invention constructed and operated as described above performs a conditional filtering by removing an unnecessary area except the ROI of the video, so that all existing videos that are not information-indexed can be searched at high speed. In addition, since it is possible to perform conditional filtering by designating all parts of the video as a region of interest, there is an advantage of searching for items or regions that are not pre-indexed in advance.

In addition, the real-time storage of the video has the advantage of preventing the performance degradation of the system for storing the image in real time by excluding the separate process of information-indexing as in the prior art, and to increase the capacity of the stored video.

1 is an exemplary view illustrating a situation in which a fast search method of a stored video according to an embodiment of the present invention may be used.
2 is a flowchart illustrating a method of rapidly searching a stored video according to an embodiment of the present invention step by step.
3 is an exemplary view illustrating a first step of a fast search method of a stored video according to an exemplary embodiment of the present invention.
4 is an exemplary view illustrating a third step of a fast search method of a stored video according to an embodiment of the present invention.
5 is an exemplary view showing a region of interest set in a video and a Cartesian coordinate system according to an embodiment of the present invention.
6 is an exemplary view illustrating a fourth step of a fast search method of a stored video according to an embodiment of the present invention.
7 is an exemplary diagram illustrating that a pixel block of a video is reduced by a ½ multiple according to an embodiment of the present invention.
8 is an exemplary view illustrating a fifth step of a fast search method of a stored video according to an embodiment of the present invention.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. The features and advantages of the present invention will become more apparent from the following detailed description of preferred embodiments based on the accompanying drawings.

Prior to this, the terms or words used in the present specification and claims are defined in the technical spirit of the present invention on the basis of the principle that the inventor can appropriately define the concept of the term in order to explain his invention in the best way. It must be interpreted to mean meanings and concepts.

In addition, it will be apparent to those skilled in the art that the present invention may be practiced without specific details such as specific components in the following description. Therefore, a detailed description of the general configuration for the governor well known in the art will be omitted.

1 is an exemplary view illustrating a situation in which a fast search method of a stored video according to an exemplary embodiment of the present invention may be used.

As can be seen with reference to Figure 1, the present invention is used as a method for retrieving a large amount of video, and can be used to quickly search for a specific video portion that can be a clue, such as an accident or a crime in a video taken by digital CCTV, etc. Can be.

In other words, in the case of digital CCTV monitoring a certain place, a large amount of video is stored for each period. If an accident or a crime occurs at the monitoring place during a certain period, the incident location is designated as an area of interest. Fast search will be performed for each condition.

As a result, searchers can skip the process of searching for a specific video over a long period of time, thereby reducing labor costs and reducing wage costs, and can search for a specific video that can be a clue in a short time. The investigation will be able to proceed quickly.

On the basis of the accompanying drawings for the present invention having these advantages will be described in more detail below.

2 is a flowchart illustrating a fast searching method of a stored video according to an exemplary embodiment of the present invention step by step.

Referring to FIG. 2, the method of fast searching a stored video according to an exemplary embodiment of the present invention may include a first step of analyzing a format of a video file and loading a demuxer and a decoder, and the resolution of the video. A second step of preparing an image surface processing according to the third embodiment; a third step of setting a region of interest and a search condition filter for a portion of the video; and only an image of the region of interest set in the third step to match the search condition filter. And a fifth step of performing a search and matching operation, and a fifth step of storing and outputting an image extracted through a search condition filter in a region of interest of the video at a predetermined time unit.

3 is an exemplary view illustrating a first step of a fast search method of a stored video according to an exemplary embodiment of the present invention.

As shown in FIG. 3, a video file has a structure of multiple streams based on a header. A module which accurately grasps the structure and reads each stream information and data for each stream is decoded. A decoder is a module that decompresses each stream data and converts it into original video and audio sample data.

Loading the demux and decoder in the first step means selecting a demux suitable for the structure through the video file header and selecting a decoder suitable for the codec of the video file format.

In the first step, it is preferable to process only 31767 bytes of a portion of a video file, which is to optimize video format analysis performance.

The second step may include a memory preparation step of preparing a memory for receiving an image according to an output resolution of the decoder and a memory for outputting an image of a fast postprocessor (FP), and transferring format information of a video to the search condition filter. Includes a video information delivery step, each of these steps is a preparation step for the high-speed search of the video.

4 is an exemplary view illustrating a third step of a fast search method of a stored video according to an embodiment of the present invention.

4 (a) shows setting a region of interest for a certain portion of the video, and the region of interest is designated by dragging a certain region of the video with a cursor. As shown in the figure, the polygons can be changed. Therefore, it is possible to easily designate a region of interest for an irregular region in a video.

A method of fast searching a region of interest of such a polygon will be described with reference to FIG. 5.

5 is an exemplary view illustrating a region of interest (a) set in a video and a Cartesian coordinate system (b) according to an embodiment of the present invention.

In the fast search of a video, a search is performed by setting only the maximum and minimum positions to be searched horizontally for an area where the Y coordinate increases based on the coordinate system b. That is as follows.

Looking at only a part of the portion where the top Y coordinate of the Cartesian coordinate system (b) of FIG. 5 becomes smaller from 50 to 1, the point at which the Y coordinate is 50 exists only in the pixel at which the X coordinate is 53, and the point at which the Y coordinate is 49 is It is the area to be recognized in the part where X coordinates are 23, 51 ~ 55, and the point where Y coordinate is 48 is the point where X coordinates are 22 ~ 24, 51 ~ 55. Set it. Table 1 shows an example of a coordinate table for fast searching a polygonal ROI.

Y coordinate X coordinate X coordinate 50 53 53 49 23 23 49 51 55 48 22 24 48 49 57 47 21 25 47 47 60 46 20 26 46 45 63 45 19 27 45 43 66 … … …

The above coordinate table is set once in configuration and used as read-only without modification in the middle of program execution. For example, if the size of the full screen is 480 vertically and 640 horizontally, the search time of 480 * 640 is required.However, when searching only the area set in the coordinate table as above, only the total length of the X coordinate of the table is used. This will shorten the time.

4 (b) shows that the search condition filters are set in the region of interest specified in the video, and the search condition filters can be set in plural to detect several conditions in one search without repetition.

6 is an exemplary view illustrating a fourth step of a fast search method of a stored video according to an embodiment of the present invention.

Referring to Figure 6 the fourth step. A frame extraction step of extracting a key frame after detecting a shot boundary frame of the ROI and a feature comparison step of comparing the key frame with a feature value of a search condition filter.

The minimum unit constituting a movie is a frame, which is an individual image corresponding to one film, and a cut is a part of a scene where a scene is changed, and a small video unit representing an image taken by one camera. Shot is called a unit, and a unit composed of consecutive shots having the same logical content is called a scene. The task of dividing a video into shots is called video segmentation, and shots are used as basic units of video segmentation.

Referring to the fourth step in more detail based on the description of the above-described video configuration and FIG. 6, first, a combination of four features is used to detect a shot boundary frame. The first feature is a difference value between pixels from a previous image. It is the most basic to measure the similarity of different images and shows the change of luminance. The second feature is the chi-squared value for the histogram of the DC image, which is generally used for shot detection, and shows the change in the histogram for the previous image. The third feature is the difference in histogram variance between the previous DC image and the current DC image, and represents the change in the overall distribution of the histogram. The fourth feature is the chi-squared value of the variance of the column and row positions of each half value of the histogram of the quantized image, which is robust to the movement of the object by using the half values of the histogram, and the variance of the column and row positions of each bin value. It is insensitive to the change of color by obtaining the chi square value.

As described above, after detecting the shot boundary frame using the four feature combinations described above, the most similar key frame is output as a result image by comparing the feature values of the key frame and the feature values of the search condition.

Here, in the fourth step, it is preferable to perform a search and matching operation by converting the color region of the video to a black and white (Gray) format, which is one of the methods for fast searching of the video.

As shown in the table below, the YUV2 format input from the video file uses U information for odd pixels and V information for even pixels, and uses the capacity of 'full area x 2'. In the case of the RGB format, three pieces of information are required for each pixel, red, green, and blue, and thus a capacity of 'full area × 3' is required. In the case of the gray format, only one contrast information is required, so only a capacity of 'full area x 1' is required. Table 2 shows the amount of information required for each pixel by video format.

YUV2 format YU YV YU YV … RGB format RGB RGB RGB RGB … Gray format G G G G …

In this way, the search time through the search condition filter is two to three times in the YUV2 format inputted from the video file and the RGB format for general screen output. Pear to fold are shortened.

In the general video, only the difference between the black and white area, not the color area, is sufficient to search through the search condition filter. Therefore, the search speed can be doubled by using only Y information in the YUV2 format that is input to the video.

In addition, in the fourth step, it is preferable to perform a search and matching operation by reducing a pixel block of a predetermined size by ½, which will be described with reference to FIG. 7 as one of methods for high-speed search of a video.

7 is an exemplary diagram illustrating that a pixel block of a video is reduced by a ½ multiple according to an exemplary embodiment of the present invention.

One of the important factors for determining whether a motion event has occurred through the search condition filter is that the object moves based on a case where the total size of the motion pixels is greater than or equal to a certain size.

In the present invention, a high-speed search is performed by reducing a pixel block corresponding to a search condition to ½, ¼, ⅛, etc. to speed up a video search. When searching by reducing the size of the pixel block to ½, the search area is horizontally and vertically. By multiplying by ½ × ½ = ¼, the time can be reduced by 4 times. The fast search method is performed by reducing the pixel block to ½ as follows.

As shown in FIG. 7, the pixel block corresponding to the ROI coordinate is skipped as much as the corresponding space according to the double speed (½, ¼, 에서). The movement is determined to have occurred.

8 is an exemplary diagram illustrating a fifth step of a fast search method of a stored video according to an exemplary embodiment of the present invention.

Referring to FIG. 8, in the fifth step, it is preferable to capture the image extracted through the search condition filter in the ROI and convert the captured image into a still image file. Index information such as time and related location is recorded so that users can search it.

The foregoing has outlined rather broadly the features and technical advantages of the present invention in order that the claims that follow may be better understood. It should be appreciated by those skilled in the art that the above-described concepts and specific embodiments of the present invention can be used immediately as a basis for designing or modifying other shapes for carrying out similar objects to the present invention.

In addition, the above-described embodiment is only one embodiment according to the present invention, and various modifications and changes may be made by those skilled in the art within the scope of the technical idea of the present invention. These various modifications and changes are also within the scope of the technical idea of the present invention, and will be included in the claims of the present invention described below.

Claims (8)

Analyzing a format of a video file and loading a demuxer and a decoder;
A second step of preparing an image surface process according to the resolution of the video;
Setting a region of interest and a search condition filter for a portion of the video;
A fourth step of searching and matching the image of the ROI set in the third step according to the search condition filter;
And a fifth step of storing and outputting the image extracted through the search condition filter in the region of interest of the video at a predetermined time unit. The method of claim 1,
In the first step, the moving picture format is analyzed by processing only 31767 bytes of a portion of the moving picture file. The method of claim 1,
The second step,
A memory preparation step of preparing a memory for receiving an image according to an output resolution of the decoder and a memory for outputting an image of a fast postprocessor (FP);
And a moving picture information transferring step of transferring the format information of the moving picture to the search condition filter. The method of claim 1,
In the third step, the region of interest is designated by dragging a predetermined region of the video with a cursor, and the dragged predetermined region is variable through a cursor manipulation. The method of claim 1,
The fourth step,
A frame extraction step of extracting a key frame after detecting a shot boundary frame of the ROI;
And a feature comparison step of comparing the key frame with a feature value of a search condition filter. The method of claim 1,
In the fourth step, a fast search method for a stored video, which is performed by performing a search and matching operation by converting the color region of the video to a gray format (Gray). The method of claim 1,
In the fourth step, a search and matching operation is performed by reducing a pixel block of a predetermined size by a factor of ½. The method of claim 1,
In the fifth step, the high speed retrieval method of the stored video, which captures the image extracted through the search condition filter in the ROI, converts it into a still image file, and records index information of each file. .

Images