Search

CN-122024145-A - Feature data feedback-based monitoring video tampering detection method and system

CN122024145ACN 122024145 ACN122024145 ACN 122024145ACN-122024145-A

Abstract

The invention relates to the technical field of video tampering detection, and discloses a monitoring video tampering detection method and a monitoring video tampering detection system based on characteristic data feedback, wherein the monitoring video tampering detection method based on characteristic data feedback comprises the following steps of S101, calculating a normalized brightness frame; step S102, generating a block boundary activation field, step S103, generating a multi-scale grid energy score, step S104, calculating multi-scale spectrum kurtosis, step S105, screening to form a hidden dominant candidate set, and step S106, judging classification labels of non-malicious error concealment and malicious frame deletion. The invention firstly carries out normalization processing on the brightness component of the video frame, reduces the interference of non-tampering factors such as exposure change, captures the structural trace of error concealment through a time derivative field and a block boundary activation field, strengthens the characteristic difference through frequency domain transformation and multi-scale energy analysis, and finally combines presentation time stamps to construct a decision index to finish classification.

Inventors

  • LIU WEI
  • LI RONGGUO

Assignees

  • 莱芜职业技术学院

Dates

Publication Date
20260512
Application Date
20260414

Claims (10)

  1. 1. The monitoring video tampering detection method based on characteristic data feedback is characterized by comprising the following steps of: step S101, decoding the video stream to extract the brightness component of each frame, calculating the mean value and standard deviation in the frame, and executing the mean value removal and normalization operation on the brightness component to obtain a normalized brightness frame; Step S102, carrying out differential calculation on adjacent normalized brightness frames to obtain time derivative fields, respectively calculating differential absolute values of the time derivative fields in horizontal and vertical dimensions, and superposing the time derivative fields to generate a block boundary activation field; Step S103, transforming the block boundary activation field into a two-dimensional frequency domain, extracting grid harmonic energy duty ratio of a corresponding frequency band according to a preset coding block scale, and generating multi-scale grid energy scores corresponding to each block scale; Step S104, calculating a distribution kurtosis value of the multi-scale grid energy score, and multiplying the distribution kurtosis value by the maximum value in the multi-scale grid energy score to obtain multi-scale spectrum kurtosis; Step S105, counting the absolute deviation between the median and the median of the kurtosis of the full-sequence multi-scale spectrum, calculating the standardized abnormal degree according to the absolute deviation, and screening out video frames with the abnormal degree exceeding a preset threshold to form a hidden dominant candidate set; Step S106, extracting a video presentation time stamp sequence, calculating a normalized time notch strength, constructing a judgment index by combining the multi-scale spectral kurtosis, and judging according to the judgment index to obtain a classification label of the non-malicious error concealment and the malicious frame deletion of the concealment master candidate set.
  2. 2. The method for detecting tampering of surveillance video based on feedback of feature data according to claim 1, wherein the video stream is decoded frame by frame to extract brightness values of pixels of a single frame image; Accumulating the brightness values of all pixel points of the single-frame image and dividing the brightness values by the total number of pixels to obtain an intra-frame average value; Accumulating the squares of the difference between the brightness value of each pixel point of the single frame image and the average value in the frame, dividing the accumulated result by the total number of pixels and squaring to obtain the standard deviation in the frame; Setting a non-zero minimum constant, subtracting the intra-frame average value from the brightness value of each pixel point of the single-frame image, and dividing the sum of the intra-frame standard deviation and the non-zero minimum constant to generate a normalized brightness frame.
  3. 3. The method for detecting tampering of surveillance video based on feedback of characteristic data according to claim 1, wherein a current normalized luminance frame and a last normalized luminance frame are extracted, and a time derivative field is obtained by subtracting the value of each pixel point of the current normalized luminance frame from the value of the pixel point of the same coordinate position of the last normalized luminance frame; calculating the absolute value of the difference between the numerical value of each pixel point in the time derivative field and the numerical value of the adjacent pixel point in the horizontal direction to obtain a horizontal differential absolute value; calculating the absolute value of the difference between the numerical value of each pixel point in the time derivative field and the numerical value of the adjacent pixel point in the vertical direction to obtain a vertical differential absolute value; the horizontal differential absolute value is added to the vertical differential absolute value to generate a block boundary activation field.
  4. 4. The method for detecting the tampering of the surveillance video based on the feedback of the characteristic data according to claim 1, wherein the two-dimensional discrete Fourier transform is implemented on the block boundary activation field to obtain the two-dimensional frequency domain amplitude; And respectively calculating the quotient of the frame width and the coding block scale and the quotient of the frame height and the coding block scale aiming at any coding block scale in the coding block scale set, multiplying the two quotient by the harmonic frequency of one, two and three and rounding to obtain a horizontal direction frequency point set and a vertical direction frequency point set.
  5. 5. The method for detecting tampering with surveillance video based on feedback of feature data according to claim 4, wherein frequency domain coordinates belonging to the set of frequency points in the horizontal direction or the set of frequency points in the vertical direction are determined as corresponding frequency bands; Setting a non-zero minimum constant, calculating the square sum of the two-dimensional frequency domain amplitude in the corresponding frequency band, taking the square sum as a numerator, calculating the sum of the square sum of the two-dimensional frequency domain amplitude in the full frequency band and the non-zero minimum constant, taking the sum as a denominator, dividing the numerator by the denominator to obtain a grid harmonic energy duty ratio, and combining the grid harmonic energy duty ratios corresponding to the scales of all the encoding blocks to generate a multi-scale grid energy fraction.
  6. 6. The method for detecting tampering of surveillance video based on feedback of characteristic data according to claim 1, wherein the method is characterized in that the energy duty ratio of each grid harmonic contained in the multi-scale grid energy fraction is accumulated, and the accumulated result is divided by the number of elements of the scale set of the encoding block to obtain a mean value; Calculating the square of the difference between each grid harmonic energy duty ratio and the mean value, accumulating all square results and dividing the square results by the number of the elements to obtain a second-order central moment; and calculating the fourth power of the difference between the energy ratio of each grid harmonic and the mean value, and accumulating all the fourth power results and dividing the result by the number of the elements to obtain a fourth-order central moment.
  7. 7. The method for detecting tampering with surveillance video based on feedback of feature data according to claim 6, wherein a non-zero minimum constant is set, the fourth-order central moment is used as a numerator, the sum of the square of the second-order central moment and the non-zero minimum constant is used as a denominator, and a quotient of dividing the numerator by the denominator is calculated to obtain a distribution kurtosis value; Screening the grid harmonic energy duty ratio with the largest numerical value in the multi-scale grid energy fraction, and taking the grid harmonic energy duty ratio as the largest fraction; and calculating the product of the distribution kurtosis value and the maximum score to generate the multi-scale spectral kurtosis.
  8. 8. The method for detecting tampering of surveillance video based on characteristic data feedback according to claim 1, wherein a multi-scale spectrum kurtosis sequence covering the whole video is extracted, a median of the multi-scale spectrum kurtosis sequence is calculated, an absolute value of a difference between each multi-scale spectrum kurtosis value in the multi-scale spectrum kurtosis sequence and the median is calculated, and an absolute deviation sequence is obtained; Calculating the median of the absolute deviation sequence as a median absolute deviation; Setting a non-zero minimum constant and a preset threshold, subtracting the median from each multi-scale spectral kurtosis value in the multi-scale spectral kurtosis sequence to be used as a numerator, taking the sum of the absolute deviation of the median and the non-zero minimum constant as a denominator, calculating the quotient of dividing the numerator by the denominator to obtain a standardized abnormal degree, selecting a video frame with the standardized abnormal degree larger than the preset threshold, and establishing a hidden dominant candidate set.
  9. 9. The method for detecting tampering with surveillance video based on feedback of feature data according to claim 1, wherein a presentation time stamp sequence is read from video data, and a median of difference values of two adjacent presentation time stamps in the presentation time stamp sequence is calculated to obtain a typical frame interval; Calculating the difference between the current presentation time stamp and the previous presentation time stamp, dividing the difference by the typical frame interval, subtracting a value one from the obtained result, and selecting the larger value of the result subtracted by the value one and the value zero to obtain the normalized time notch strength; Setting a non-zero minimum constant, obtaining multi-scale spectral kurtosis corresponding to the current presentation time stamp, taking the multi-scale spectral kurtosis as a molecule, taking the sum of the multi-scale spectral kurtosis, the normalized time notch strength and the non-zero minimum constant as a denominator, calculating the quotient of dividing the molecule by the denominator, and generating a decision index; For video frames contained in the hidden dominant candidate set, if the decision index is greater than or equal to a value of 0.5, marking the video frames as non-malicious error concealment, and if the decision index is less than the value of 0.5, marking the video frames as malicious frame deletion.
  10. 10. A monitored video tampering detection system based on characteristic data feedback, wherein a monitored video tampering detection method based on characteristic data feedback as defined in any one of claims 1 to 9 is performed, comprising: the normalized brightness frame module decodes the video stream to extract the brightness component of each frame, calculates the mean value and standard deviation in the frame, and executes the mean value removal and normalization operation on the brightness component to obtain a normalized brightness frame; The block boundary activation field generation module is used for carrying out differential calculation on adjacent normalized brightness frames to obtain time derivative fields, respectively calculating differential absolute values of the time derivative fields in horizontal and vertical dimensions and superposing the time derivative fields to generate a block boundary activation field; the multi-scale grid energy score generation module is used for transforming the block boundary activation field into a two-dimensional frequency domain, extracting the grid harmonic energy duty ratio of a corresponding frequency band according to a preset coding block scale, and generating multi-scale grid energy scores corresponding to each block scale; The multi-scale spectrum kurtosis calculation module is used for calculating a distribution kurtosis value of the multi-scale grid energy score, and multiplying the distribution kurtosis value by the maximum value in the multi-scale grid energy score to obtain multi-scale spectrum kurtosis; The hidden leading candidate set screening module is used for counting the absolute deviation between the middles of the full-sequence multi-scale spectrum kurtosis, calculating the standardized abnormal degree according to the absolute deviation, and screening out video frames with the abnormal degree exceeding a preset threshold to form a hidden leading candidate set; The video tampering judging module extracts a video presentation time stamp sequence to calculate normalized time notch intensity, combines the multi-scale spectrum kurtosis to construct a judgment index, and judges and obtains a classification label of the non-malicious error concealment and the malicious frame deletion of the concealment master candidate set according to the judgment index.

Description

Feature data feedback-based monitoring video tampering detection method and system Technical Field The invention relates to the technical field of video tampering detection, in particular to a monitoring video tampering detection method and system based on characteristic data feedback. Background The monitoring video has irreplaceable functions in the fields of security protection, evidence obtaining, judicial and the like, and the data integrity of the monitoring video is directly related to true phase tracing and responsibility identification of an event. In practical application, the monitoring video usually faces two kinds of frame deletion related problems, one is non-malicious frame loss caused by network fluctuation and equipment failure in the transmission and storage processes, a decoding end fills the deleted content according to a coding block through an error concealment mechanism, and the other is artificial malicious frame deletion editing to cover key information. Both of these two cases can appear as inter-frame timeline discontinuities, forming isomorphism with similar appearance, presenting a significant challenge to the precise differentiation of tamper types. The existing video tampering detection technology has obvious limitation, most methods can only detect frame deletion or data integrity damage, and can not effectively distinguish non-malicious error concealment from malicious frame deletion. This problem arises from a number of factors, error concealment performs padding per coded block, leaving structural traces, while malicious erasure frames do not have such features, but both appear as inter-frame data discontinuities, which are difficult to capture by conventional methods. Meanwhile, the monitoring video is often interfered by non-tampering factors such as exposure change, automatic gain adjustment, natural texture and the like, and the interference can mask structural characteristics of error concealment, so that misjudgment is easy to occur in the traditional detection method. In addition, the prior art is multi-dependent on inter-frame consistency check or complex feature stacking, not only is the robustness insufficient, but also the block structure difference of different coding standards such as H264, HEVC and the like is difficult to adapt, the distinguishing difficulty of two types of frame missing phenomena is further increased, the requirement of accurate judgment of tampering types in practical application cannot be met, and the effectiveness and reliability of the monitoring video as evidence are affected. Disclosure of Invention The invention provides a monitoring video tampering detection method and system based on characteristic data feedback, which solve the technical problems in the background technology. The invention provides a monitoring video tampering detection method based on characteristic data feedback, which comprises the following steps: step S101, decoding the video stream to extract the brightness component of each frame, calculating the mean value and standard deviation in the frame, and executing the mean value removal and normalization operation on the brightness component to obtain a normalized brightness frame; Step S102, carrying out differential calculation on adjacent normalized brightness frames to obtain time derivative fields, respectively calculating differential absolute values of the time derivative fields in horizontal and vertical dimensions, and superposing the time derivative fields to generate a block boundary activation field; Step S103, transforming the block boundary activation field into a two-dimensional frequency domain, extracting grid harmonic energy duty ratio of a corresponding frequency band according to a preset coding block scale, and generating multi-scale grid energy scores corresponding to each block scale; Step S104, calculating a distribution kurtosis value of the multi-scale grid energy score, and multiplying the distribution kurtosis value by the maximum value in the multi-scale grid energy score to obtain multi-scale spectrum kurtosis; Step S105, counting the absolute deviation between the median and the median of the kurtosis of the full-sequence multi-scale spectrum, calculating the standardized abnormal degree according to the absolute deviation, and screening out video frames with the abnormal degree exceeding a preset threshold to form a hidden dominant candidate set; Step S106, extracting a video presentation time stamp sequence, calculating a normalized time notch strength, constructing a judgment index by combining the multi-scale spectral kurtosis, and judging according to the judgment index to obtain a classification label of the non-malicious error concealment and the malicious frame deletion of the concealment master candidate set. The invention provides a monitoring video tampering detection system based on characteristic data feedback, which comprises: the normalized brightness frame module deco