JP2007180970A - Video processor and monitoring camera system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device capable of easily confirming a change after recording by stably detecting an abnormal state without requiring new processing and a device different from encoding processing in a monitoring camera system provided with a compression coding means including an inter-frame predictive encoding method. <P>SOLUTION: This video processor and monitoring camera system for filing encoded video data and recording the video data in a randomly accessible recording medium 120 is provided with a compression encoding means 104 including motion compensation processing using a motion vector detected by a motion vector detecting means 105 and quantization processing, determines a video change point (abnormal state) from either a maximum value of the size of a motion vector of a corresponding encoded frame or a variation in an average value and a variation in a quantization coefficient of the encoded frame, associates the encoded data with the filed video file, and records the encoded data as meta data. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a video processing apparatus and a surveillance camera system that detect video change points simultaneously with compression encoding using motion vector information detected when video is compression-encoded.

  In recent years, in the field of video encoding technology, particularly in fields that require long-time recording, that is, recording at a low bit rate, apparatuses that perform inter-frame predictive encoding using correlation information between frames are generally used. It has become. In inter-frame predictive coding, generally, motion compensation processing is performed for efficiently performing compression coding by detecting a motion vector from the correlation between frames.

  On the other hand, with the advance of digital video signal processing, compression encoding technology, and network technology, video equipment has been digitized, and in the surveillance camera system field, a system that records and compresses digital video signals Has been developed.

  In surveillance camera systems, events that occur occasionally (for example, people have entered a place where people usually do not enter) can be efficiently discovered and recorded only, and the video can be easily viewed during playback. If it is possible to search, the convenience will be improved dramatically. That is, a change point detection technique that detects a change in video and uses the position as a change point for recording and reproduction is important. In view of this, there has been proposed a technique for detecting a change point using motion vector detection used in inter-frame predictive encoding in a surveillance camera system that records video after compression encoding.

  For example, when a change from a small state to a large state is detected between frames of the input video, it is recorded / transmitted from that time to a time when the change again changes from a large state to a small state, and the change is small. A video processing apparatus that does not record / transmit video for a period is disclosed as useless video for monitoring (see, for example, Patent Document 1). In this conventional video processing apparatus, a change between frames is detected using both a difference in pixel values at the same position between frames and the magnitude of a motion vector.

  Normally, only the background video is input from the camera, but if an intruder appears in the video at some time, the intruder is recognized as a difference to the background video, so this should be detected as a scene change. Can do. Therefore, a video processing apparatus is disclosed that stores information representing scene changes as additional information in a frame that is recognized as an intruder, and displays a list of scene change portions by searching for the additional information later. (For example, refer to Patent Document 2). In the scene change detection in this conventional video processing apparatus, the amount of screen change is obtained from the magnitude and distribution of the motion vector, and a scene change is determined where there has been a change beyond a certain value.

These conventional video processing apparatuses detect a change or abnormality in a video by combining a motion vector detected at the time of compression encoding and a technique different from the compression encoding.
JP 2001-309354 A JP 2000-287165 A

  However, the motion vector detected at the time of video inter-frame predictive encoding is detected in order to efficiently encode video data, and is not detected for detecting a change point. . For this reason, in conventional video processing devices, video features and motion vector detection results do not necessarily match, and therefore, change point detection based only on motion vector detection results often results in false detection of change points. There was a problem. For example, in an outdoor surveillance camera system, if trees are reflected in the background and the wind is blowing, the average value of the motion vector of the entire screen will be high, and it may be detected as a change point. is there. In order to improve this detection accuracy, as disclosed in Patent Document 1, apart from compression coding, a new process or device for detecting a change point is necessary, resulting in a large-scale system. It is difficult to save power and reduce prices.

  In view of the above problems, an object of the present invention is to provide a video processing apparatus capable of performing efficient change point detection while suppressing an increase in system scale.

  In order to solve the above-described problems, a video processing apparatus according to the present invention is a video processing apparatus that records video data encoded in a randomly accessible recording medium as a file, and detects change points from the video data (hereinafter, referred to as “change point detection”). In order to perform an appropriate detection), at least a motion compensation process using a motion vector detected by the motion vector detection means, and an efficient quantization by adaptively changing the quantization coefficient Video coding system including quantization processing, the maximum value of the motion vector of the corresponding encoded frame, or the amount of change in the average value of the motion vector between frames, and the quantization coefficient between frames When both of the change amounts exceed the respective threshold values, it is determined that “abnormality exists”. The detected abnormality detection information is recorded as metadata in association with an offset frame value from the top of the video file and data obtained by encoding the video in association with a file to be recorded on the recording medium.

  At the time of reproduction, reproduction is started from a frame corresponding to the Offset value based on the abnormality detection information recorded as metadata.

  According to the video processing apparatus of the present invention, in the video processing apparatus and the monitoring camera system for compressing and encoding video data and recording the compressed encoded data as a file on a randomly accessible recording medium, the confirmation after recording is efficiently performed. Therefore, it is possible to perform abnormality detection more stably.

  Also, the processing necessary to detect anomalies is necessary when performing compression encoding processing to record video, such as motion vectors necessary for motion compensation prediction processing and quantization coefficients resulting from quantization processing. Therefore, no new process or device different from the compression encoding process is required. Therefore, a low-priced and low-power-consumption video processing device can be realized, and can be easily applied to a portable video processing device.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 shows a configuration of a surveillance camera system according to an embodiment of the present invention. In this embodiment, an example of a surveillance camera system having a camera unit that captures an image of a subject and a recording unit that records video captured by the camera unit as compressed video data that has been compression-encoded on a recording medium will be described. To do. Such a monitoring camera system is used as, for example, a nighttime monitoring camera system in an office building or a monitoring camera system at an entrance / exit.

  As shown in FIG. 1, the surveillance camera system performs a pre-encoding filter process on the video data from the camera unit 101 that supplies video data to be recorded on the recording medium 120 and the video data. A video processing unit 102; a video encoding unit 103 that compresses and encodes video data to generate compressed encoded data; a video decoding unit 106 that decodes compressed encoded data recorded on the recording medium 120; A video output unit 107 for outputting the converted video data, a system control unit 110 for controlling the entire system, an I / O bus 130 for interconnecting the system control unit 110 to control each unit, and confirmation An input unit 131 that receives an input from the user such as an operation, and a display device for displaying list information of a clip (described later) necessary for the user operation, a still image, And a 32.

  Furthermore, the video encoding unit 103 includes a video encoding unit 104 that compresses and encodes video data, and a motion vector detection unit 105 that detects a motion vector between frames. The system control unit 110 compresses the video. Video file recording unit 111 that records the encoded data as a file on a recording medium, image information management unit 112 that acquires and manages statistical information about an image, video change analysis unit 113 that detects an abnormality, and that an abnormality has been detected Metadata recording means 114 that records abnormality detection information that is information associated with the video file, for example, display control means 115 that performs display control for displaying clip list information and thumbnail images, and recording on recording medium 120 Video file playback means 116 for controlling the playback of the compressed encoded data, and the recorded media It is configured to include a metadata playing section 117 to manage reading data.

  In the present embodiment, an example is given in which the system control unit 110 is realized by, for example, a microcomputer and a memory (not shown), and each processing unit included in the system control unit 110 is stored in the memory by the microcomputer. This is realized by executing various programs.

  On the other hand, the video data supplied to the video encoding unit 103 is, for example, the MPEG-2 system widely used as a video compression system, or MPEG-4 AVC / H. It is compression-encoded based on the H.264 standard. The present invention exerts the same effect as long as it is applied to a compression coding system to which inter-frame predictive coding is applied, but in this embodiment, MPEG-4 AVC / H. An example of compressing and encoding in the H.264 system (hereinafter abbreviated as H.264) will be described.

  In the surveillance camera system configured as described above, for example, when the user instructs to start monitoring by the input unit 131, the camera unit 101 performs an imaging process and outputs digitized video data. Next, this signal is supplied to the video processing unit 102, and the supplied video data is subjected to, for example, band limitation by a filter and noise removal. The processed video data is provided to the video encoding unit 103.

  The video encoding unit 103 includes a video encoding unit 104 and a motion vector detection unit 105. Although details will be described later, the video data is H.264. The data is compressed and encoded according to the H.264 system. The compressed encoded data compressed and encoded by the video encoding unit 103 is recorded on the recording medium 120 as a file by the video file recording unit 111 of the system control unit 110.

  Furthermore, the calculation result and detection result in the video encoding unit 103 are acquired by the image information management unit 112 of the system control unit 110. For example, an average value or maximum value of motion vectors detected by the motion vector detection unit 105 at the time of compression encoding, a quantization parameter calculated by the video encoding unit 104, a block size distribution of motion compensation processing, a reference frame Image statistical information of each frame such as the minimum frame length is managed in a state in which it is synchronized with the compression-encoded data of each frame. The video change analysis unit 113 acquires the image statistical information from the image information management unit 112, performs an abnormality detection process based on the information, and detects two abnormality detections “abnormal” and “abnormal” for each frame. Information is determined. If it is determined that there is “abnormal”, this is notified to the metadata recording means 114. The metadata recording unit 114 records the abnormality detection information “abnormal” in a separate file as metadata in association with the file in which the video is recorded and the frame number determined as “abnormal”. Details of the abnormality detection method in the video change analysis unit 113 and the method of recording the abnormality detection information in the metadata recording unit 114 as metadata will be described later.

  When the user instructs the end of monitoring by using the input means 131, the monitoring camera system ends the recording operation.

  A file group formed by one continuous recording as described above is called a clip. The clip includes at least a data file obtained by compressing and encoding a video, and may include, for example, an audio file and a file that records metadata indicating a shooting location and time. Also, when it is necessary to divide the video file due to the capacity of the recording medium or management restrictions on the file system, the video data from the start of monitoring to the end of monitoring is divided into two or more clips and recorded. Record on media. In this embodiment, since it is a system for monitoring use, the order of recorded clips can be easily determined by, for example, recording time. Therefore, in the present embodiment, the description that defines the continuity of the clip is not recorded on the recording medium, but the description may be recorded on the recording medium.

  When reproducing the compressed and encoded data recorded as a file on the recording medium 120, the metadata reproducing means 117 reads the metadata file from the recording medium 120, and further reads the abnormality detection information in the metadata. Then, in response to a playback instruction from the user through the input unit 131, the video file playback unit 116 reads out the compressed encoded data of the frame determined to be “abnormal” and provides it to the video decoding unit 106. The video decoding unit 106 performs the H.264 decoding. The decoding processing conforming to the H.264 standard is performed, the decoded video data is provided to the video output unit 107, and the video output unit 107 displays the video. Details of the processing of the metadata reproducing means 117 will be described later.

  Next, the operation of the video encoding unit 103 will be described in detail. A detailed block diagram of the video encoding unit 103 is shown in FIG. FIG. 2 shows a general H.264 standard defined by the standard. The motion vector detection unit 105 in FIG. 1 corresponds to the motion vector detection unit 208, and the video encoding unit 104 includes a DCT / quantization unit 201 to an inter-frame prediction processing unit. 207 corresponds to each block.

  First, when encoding is started in accordance with a recording instruction from the input unit 131, input video data is provided to the DCT / quantization unit 201, subjected to discrete cosine transform (DCT) and quantization processing, and then entropy encoded. The unit 202 performs variable length coding and arithmetic coding, and outputs the compression coded data as a bit stream. At the same time, the inverse quantization / IDCT unit 203 performs an inverse quantization process and an inverse DCT (IDCT) process on the converted data provided from the DCT / quantization unit 201, and the image inversely converted to the filter unit 204. Provide data. This video data is subjected to deblocking filter processing in the filter unit 204 and then stored in the frame memory 205 and used in intra-frame prediction processing and inter-frame prediction processing. In the intra-frame prediction processing unit 206, Prediction is performed within a frame that is also one of the H.264 features. The prediction result is used by the DCT / quantization unit 201 when DCT / quantization of the input image of the next frame.

  On the other hand, the input video data is also input to the motion vector detection unit 208 to detect horizontal and vertical motion vectors from the reference frame stored in the frame memory 205, and the inter-frame prediction processing unit 207 Block division processing, motion compensation processing, and motion vector encoding are performed using variable MC (Motion Compensation). The prediction results in the intra-frame prediction processing unit 206 and the inter-frame prediction processing unit 207 are provided to the DCT / quantization unit 201 after selecting either one for each frame in the intra-frame / inter-frame prediction switching unit 210. The

  Further, the motion vector detection unit 208 first divides one frame of video data to be encoded into rectangular areas (for example, 16 × 16 pixels, 8 × 8 pixels, etc.) having a predetermined size, and the frame memory 205 An appropriate block is searched as a predicted value of each block from the reference frame stored therein. In general, the position of a block is moved little by little within a preset search range in a reference frame to determine whether or not the block at that position is appropriate as a predicted value. As a criterion for determining appropriateness, a sum of squares of differences of pixel values in a block, a sum of absolute differences, and the like are used. In this manner, for each block of one frame of video data to be encoded, a shift in the position of the block searched as a predicted value of the block in the reference screen is detected as a motion vector. Therefore, it is possible to grasp the outline of the motion feature of the video using the detected motion vector.

  Next, the abnormality detection method in the video change analysis unit 113 will be specifically described. First, the image information management unit 112 acquires statistical information when the video encoding unit 103 performs compression encoding in each frame. The statistical information used in the present invention includes a motion vector calculated or detected by the video encoding unit 103, a matrix corresponding to a quantization coefficient (Qstep) of each macroblock (MB), and a reference frame number of each MB. It is done. The above information can be found in H.C. This information is necessary for encoding in the H.264 format and is embedded in the compressed encoded data (see the H.264 standard for further details).

  Furthermore, the image information management unit 112, for example, the above-described frames for each frame period (Group Of Picture: GOP period in the MPEG2 encoding method) using only intra-frame predictive encoding called IDR picture or I picture. Manage statistical information.

  In addition, regarding the motion vector, the average value (AVE_ME) and the maximum value (MAX_ME) of the motion vectors of each frame are held. Regarding the quantization coefficient, the distribution of the number of MBs used for each quantization coefficient (Qstep) in each frame is held. Regarding the reference frame number, the difference from the currently encoded frame number, that is, the number of frames representing the temporal distance between the reference frame and the currently encoded frame (hereinafter referred to as the distance between the reference frame and the encoded frame). ) Is held.

  Furthermore, the video change analysis means 113 simultaneously manages the latest frame information, that is, the Offset value from the start of recording of the currently encoded frame.

  The video change analysis unit 113 performs an abnormality detection process using the information acquired by the image information management unit 112. FIG. 3 shows an example of the processing flow of the abnormality detection processing.

  FIG. 3 shows a method for detecting an abnormality from the maximum value of the motion vector, the distance between the reference frame and the encoded frame, and the distribution change of the quantization coefficient. In S001, it is determined whether or not the maximum value (MAX_ME) of the magnitude of the motion vector managed by the image information management unit 112 is greater than or equal to the threshold value TH_ME. Next, in S002, it is determined whether the distance (LEN_RF) between the reference frame and the encoded frame is less than the threshold value TH_RF. This is the This is effective for H.264 Long Term reference pictures. The Long Term reference picture is a method for improving the coding efficiency by storing the background in the conference room in a Long Term reference memory and referring to it as a Long Term reference picture, for example, in a TV conference system. Therefore, it is useful to use a Long Term reference picture in a fixed surveillance camera system with little motion, and when encoding using a Long Term reference picture, referencing other than the Long Term reference picture It is considered that a movement has occurred (that is, an abnormal state for monitoring purposes). For example, if there is no motion, ideally, if the shooting start frame is held as a Long Term reference picture, the encoded data of all frames thereafter will refer to the Long Term reference picture. Accordingly, the distance between the reference frame of the encoded data 1000 frames after the start of imaging and the encoded frame is 1000 frames. However, when there is a motion, the nearest frame such as one or two frames in which the distance between the reference frame and the encoded frame is referred to. That is, for example, if TH_RF = 3 or the like is set, LEN_RF <TH_RF and motion (abnormality) can be detected.

  In the above-described abnormality detection process, the determination in S002 is always Yes for a frame whose distance from the shooting start frame is less than TH_RF, and thus an abnormality cannot be accurately detected. However, if TH_RF is several to several tens of frames, normal detection of abnormality is sufficient while preventing erroneous detection if the abnormality is not detected (ignored) in a frame whose distance from the shooting start frame is less than TH_RF. There is no problem because it is possible.

  H. In the H.264 system, since the reference frame can be changed in units of macroblocks (MB), LEN_RF is the minimum distance between the currently encoded frame and the reference frame of each MB. Further, when the Long Term reference picture is not used, the step of S002 may be omitted.

  In S003, the change of the quantization coefficient is determined. H. In the H.264 system, the quantization coefficient (Qstep) can be changed for each MB, a distribution table of the number of MBs in each quantization coefficient is created, and the absolute difference in the number of MBs having each quantization coefficient is created. The sum of the values is defined as the quantization coefficient change amount (MV_QS). The number of MBs quantized by n Qsteps (Q1 to Qn) and having the quantization coefficient of the currently encoded frame as Qn is MV_Qn_F1, and the quantum of the previous frame in the order of input to the video encoding unit 103 Assuming that the number of MBs with the quantization coefficient Qn is MV_Qn_F2, the amount of change in the quantization coefficient (MV_QS) is expressed by the following equation.

MV_QS = | MV_Q1_F2-MV_Q1_F1 |
+ | MV_Q2_F2-MV_Q2_F1 |
+ ...
+ | MV_Qn_F2-MV_Qn_F1 |
It is determined whether this MV_QS is greater than or equal to a threshold value TH_QS. This is because the quantization coefficient usually has a constant value or a constant distribution if there is no change in each frame, and there is a change in the value and distribution compared to the previous frame. This means that a large change has occurred in the image of the frame currently being encoded.

  When all of the above determinations of S001, S002, and S003 are applicable, it is determined that “abnormality is present”, and in other cases, it is determined that “abnormality is not present”. If “abnormal” is detected, an offset from the recording start time of the clip managed by the image information management unit 112 to the currently encoded frame is acquired in S004, and in S005 Abnormality detection information including Offset and abnormality type is notified to the metadata recording unit 114.

  In this embodiment, TH_ME = 10, TH_RF = 3, and TH_QS = 200. This is a spatial movement distance between one frame of a video, assuming that the video is shot at 30 frames per second and the video size is 720 x 480 and is used for surveillance cameras with a long distance shooting and little motion. Is intended to be judged as “abnormal” if the movement continues for about 1/10 second (influences the set value of TH_RF) or more than about 0.1% (influences the set values of TH_ME and TH_QS) Means a combination of parameters.

  For these parameters, optimum threshold values may be selected according to the installation location of the surveillance camera, the purpose, and the number of frames taken per second. For example, in the case of a fixed surveillance camera system for entrance / exit monitoring, it is considered that there is no movement due to external factors such as wind, and the motion vector is simply detected as the movement of a person or an object. Therefore, the values of TH_ME and TH_QS may be reduced to increase sensitivity. Further, when the number of frames taken per second is small, the difference between the frames becomes large, so the values of TH_ME and TH_QS may be set large.

  Further, if the input unit 131 has a configuration in which the values of TH_ME and TH_QS can be selected, a system applicable to various uses can be realized. For example, in a normal outdoor recording mode, TH_ME = 10 and TH_QS = 200, and in an indoor recording mode, TH_ME = 5 and TH_QS = 100, and these modes can be selected by the user.

  FIG. 4 shows a process flow of another example of the abnormality detection process. In FIG. 4, in S101, the difference between the average value of the motion vectors of the immediately preceding frame (AVE_ME1) in the display order and the average value of the motion vectors of the currently encoded frame (AVE_ME2) is used. To detect abnormalities. That is, abnormality determination is performed based on whether or not AVE_ME_DIFF represented by the following formula is equal to or greater than a threshold value TH_ME2.

AVE_ME_DIFF = | AVE_ME1-AVE_ME2 |
In S102, as in S003 of FIG. 3, determination is made based on whether or not the change amount (MV_QS) of the quantization coefficient is equal to or greater than the threshold TH_QS2. If all of the above determinations in S101 and S102 apply, it is determined that “abnormality exists”, and in other cases, it is determined that “no abnormality”. If “abnormal” is detected, an offset from the recording start time of the clip managed by the image information management unit 112 to the currently encoded frame is acquired in S103, and in S104, the above-described offset is acquired. Abnormality detection information including Offset and abnormality type is notified to the metadata recording unit 114.

  As mentioned above, although two examples were shown about the abnormality detection method, the judgment of abnormality detection may change a combination according to a use.

  Next, a method for recording abnormality detection information as metadata in the metadata recording unit 114 will be described. FIG. 5 shows the contents of clip metadata, which is metadata assigned to a clip. Clip ID is a unique ID automatically assigned by each device for each clip. Video File Name is a file name of a file that records data obtained by compressing and encoding video. Frame Rate is a name of a captured video. The number of frames per second, and Duration represents the length of the recorded clip in terms of the number of frames. NumWarning is the number of detected abnormal detection results detected by the video change analysis means 113 in the clip, and represents the number of WarningInfo described below. “WarningInfo” is abnormality detection information defined in FIG. 5B, “Offset” is the number of frames from the top of the clip when the abnormality is detected, and “WarningType” is the type of abnormality detection (for example, “WARNING:“ abnormal ”). Present ", INFOMATION: reference information, etc.). WarningInfo is recorded by the number of values of NumWarning.

  In the present embodiment, in order to improve convenience during search, clip metadata is recorded in a text file format as a separate file from data obtained by compression-encoding video.

  FIG. 6 shows an example of changes in the contents of the clip metadata file from the start to the end of recording. FIG. 6A shows an example of the contents of a clip metadata file immediately after the start of clip recording. In this example, Clip ID is C0001, and Video File Name is C0001. mxf and Frame Rate are 30 frames per second. Clip ID and Video File Name are determined before the start of clip recording. For example, Clip ID is a 4-digit increment counter value with “C” added at the beginning, and Video File Name is in a format with an extension added to Clip ID. Here, the video file is an MXF file format conforming to SMPTE-377M and the extension is mxf, but other file formats may be used. Also, Duration is set to 0 because it is not fixed immediately after the start of recording. Further, NumWarning is also set to 0 until “abnormal” is detected, and when NumWarning is 0 after recording of the clip, it indicates that no abnormality is detected in the clip.

  Next, FIG. 6B shows the contents of the clip metadata file after detecting an abnormality for the first time during the recording of the clip. 6A, NumWarning is set to 1, and corresponding WarningInfo (Offset: 1234, Warning Type: WARNING) is added. Also, the Duration item is changed to the clip length (1234) at that time. FIG. 6C shows the contents of the clip metadata file after clip recording is completed. In this example, three abnormalities are detected.

  Next, the details of the processing of the metadata reproduction means 117 when reproducing in order to confirm an abnormality after recording will be described. When the abnormality detection information is confirmed after recording, the abnormality search can be easily performed using the metadata indicating the abnormality detection information described above. By searching the clip metadata file of each clip, the presence or absence and number of abnormalities can be known, and the offset can be easily acquired. Therefore, the metadata reproduction means 117 creates an abnormality detection list including an abnormality type (Warning Type), a clip ID (Clip ID), and an abnormality occurrence position (Offset) as shown in FIG. Display control is performed, and the abnormality detection list is displayed on the display device 132. For example, when the user selects a desired abnormality detection position number (NO.) For the displayed abnormality detection list by the input means 131, the video file reproduction means 116 starts from the corresponding Offset of the target video file. The compressed encoded data recorded on the recording medium is read out. The read compressed encoded data is provided to the video decoding unit 106, subjected to decoding processing, and supplied to the video output unit 107 as video data. The supplied video data is displayed on the video output unit 107. In this way, it is possible to easily search for and reproduce a video when an abnormality is detected.

  In addition, although the abnormality detection list to be displayed (FIG. 7) shows an example in which the specification of the frame position is presented in Offset, for example, a time at the time of abnormality detection or a thumbnail image may be displayed. The time at the time of abnormality can be calculated from the recording start time of the clip and the offset, and the thumbnail image is created from video data before encoding at the time of abnormality detection, or within the frame called the latest I picture after recording There is a method of acquiring and creating a DC component of a frame encoded using only predictive encoding.

  Furthermore, a description will be given of a method for detecting an abnormality by searching recorded compressed encoded data without detecting an abnormality during recording. In the compressed and encoded data recorded on the recording medium, information such as motion vectors, quantization coefficients, and reference frame numbers are embedded in an encoded state because they are necessary for decoding. If this information is extracted at the time of decoding and subjected to the same processing, it is possible to detect an abnormality after recording. In addition, since a motion vector and a quantization coefficient must be extracted at the time of decoding if the decoding device is compatible with the same encoding method (for example, H.264 method) as at the time of encoding, a special device is used. If information extracted at the time of decoding is not necessary, it can be realized with the same configuration as in FIG. That is, compressed video data is supplied to the video decoding unit 106 by the video file reproduction unit 116, and information such as motion vectors and quantization coefficients extracted by the video decoding unit 106 is acquired by the image information management unit 112. If management is performed, the other blocks can be realized by the method of this embodiment described above.

  As described above, according to the surveillance camera system of the present embodiment, more stable abnormality detection can be performed using only information used at the time of compression encoding without adding a processing unit different from the compression encoding processing. For example, a low-priced and power-saving surveillance camera system such as a portable surveillance camera system can be realized.

  In the present embodiment, the method of recording metadata as a separate file has been described. However, the same effect can be achieved by embedding the metadata in the header portion of the video recording file. In addition, the method of recording in the text format when recording the metadata has been described. For example, when describing in a markup language such as an XML (extensible Markup Language) format, or when abnormality detection information (WarningInfo) is described. The same effect is exhibited as a list structure.

  The video processing apparatus according to the present invention detects an abnormality from a video with little movement and improves searchability at the time of confirmation, and is widely applied to a night surveillance camera system, an entrance / exit surveillance camera system, or a video editing apparatus. Applicable.

Diagram showing the configuration of the surveillance camera system The figure which shows the structure of a video encoding part. The figure which shows the 1st processing flow example of an abnormality detection method The figure which shows the 2nd processing flow example of the abnormality detection method Figure showing the contents of a clip metadata file The figure which shows the example of the content in each step of a clip metadata file Diagram showing an example of anomaly detection display

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 Camera part 102 Image | video processing part 103 Image | video encoding part 104 Image | video encoding means 105 Motion vector detection means 106 Image | video decoding part 107 Image | video output part 110 System control part 111 Image | video file recording means 112 Image | video information management means 113 Image | video change analysis means 114 Metadata recording unit 115 Display control unit 116 Video file reproduction unit 117 Metadata reproduction unit 120 Recording medium 130 I / O bus 131 Input unit 132 Display device 201 DCT / quantization unit 202 Entropy encoding unit 203 Inverse quantization / IDCT Unit 204 filter unit 205 frame memory 206 intra-frame prediction processing unit 207 inter-frame prediction processing unit 208 motion vector detection unit 210 intra-frame / inter-frame prediction switching unit

Claims (8)

A video processing device that performs compression coding using correlation information between frames for an input video and detects a change in the video as a video change point,
Motion vector detecting means for detecting a motion vector between frames of the video;
Video encoding means that includes quantization processing and compresses and encodes the video using correlation information between frames based on the motion vector;
Image information management means for managing parameters including information relating to at least the quantization coefficient in the quantization processing when the video encoding means performs compression encoding;
A video processing apparatus comprising: a video change analysis unit that detects the video change point based on the motion vector and the parameter managed by the image information management unit. The video change analysis means calculates the magnitude of the motion vector in the frame immediately before the encoding target frame, of the maximum value of the size of the motion vector in the encoding target frame and the average value of the size of the motion vector in the encoding target frame. Any one of the change amounts with respect to the average value of the lengths and the change amount of the quantization coefficient in the encoding target frame with respect to the quantization coefficient in the frame immediately before the encoding target frame are set in advance. The video processing apparatus according to claim 1, wherein the video change point is detected based on a comparison result by comparing with a threshold value. The video processing apparatus according to claim 2, wherein the video change analysis unit further detects a video change point based on temporal distance information from an encoding target frame to a frame referenced by the encoding target frame during encoding. Furthermore, video file recording means for recording the compression encoded data as a file on a recording medium;
Metadata recording means for adding metadata that is information about the video to the file and recording it on the recording medium;
The video processing apparatus according to claim 1, wherein the metadata recording unit records at least a time of the video change point as metadata on a recording medium. 5. The video recording apparatus according to claim 4, wherein the metadata recording means creates a thumbnail image of a frame at the video change point and records it as metadata. A video processing device that decodes compression-encoded data obtained by compressing and encoding video using correlation information between frames and detects a change in the video as a video change point,
Video decoding means for extracting information on a motion vector and a quantization coefficient from the compressed encoded data at the time of decoding the compressed encoded data;
A video processing apparatus comprising: video change analysis means for detecting the video change point based on the motion vector and the quantization coefficient extracted from the compressed encoded data by the video decoding means. The video decoding means further extracts temporal distance information from a frame to be encoded to a frame to which the frame to be encoded is referred when encoding the compressed encoded data,
The video processing apparatus according to claim 6, wherein the video change analysis unit further detects the video change point based on the temporal distance information. A surveillance camera system that performs compression encoding on a video input from a camera unit using correlation information between frames and detects a change in the video as a video change point,
Motion vector detecting means for detecting a motion vector between frames of the video;
Video encoding means for compressing and encoding the video using correlation information between frames based on the motion vector;
Image information management means for managing parameters including information on at least quantization coefficients when compression-encoded by the video encoding means;
A surveillance camera system comprising: a video change analysis unit that detects the video change point based on the motion vector and the parameter managed by the image information management unit.

Images