CN108596169B - Block signal conversion and target detection method and device based on video stream image - Google Patents
Block signal conversion and target detection method and device based on video stream image Download PDFInfo
- Publication number
- CN108596169B CN108596169B CN201810201588.5A CN201810201588A CN108596169B CN 108596169 B CN108596169 B CN 108596169B CN 201810201588 A CN201810201588 A CN 201810201588A CN 108596169 B CN108596169 B CN 108596169B
- Authority
- CN
- China
- Prior art keywords
- image
- video
- global
- sub
- video stream
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 71
- 238000006243 chemical reaction Methods 0.000 title claims abstract description 46
- 238000012545 processing Methods 0.000 claims abstract description 76
- 238000000034 method Methods 0.000 claims abstract description 59
- 230000009467 reduction Effects 0.000 claims abstract description 7
- 238000009499 grossing Methods 0.000 claims abstract description 6
- 238000001914 filtration Methods 0.000 claims description 42
- 238000012544 monitoring process Methods 0.000 claims description 27
- 238000005070 sampling Methods 0.000 claims description 15
- 230000011218 segmentation Effects 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000007781 pre-processing Methods 0.000 claims description 5
- 238000003672 processing method Methods 0.000 claims description 5
- 230000008859 change Effects 0.000 abstract description 6
- 230000002159 abnormal effect Effects 0.000 abstract description 4
- 230000008569 process Effects 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 238000000605 extraction Methods 0.000 description 6
- 238000005286 illumination Methods 0.000 description 5
- 230000009545 invasion Effects 0.000 description 4
- 230000033001 locomotion Effects 0.000 description 4
- 230000000903 blocking effect Effects 0.000 description 3
- 238000010276 construction Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 241000124033 Salix Species 0.000 description 2
- ZZUFCTLCJUWOSV-UHFFFAOYSA-N furosemide Chemical compound C1=C(Cl)C(S(=O)(=O)N)=CC(C(O)=O)=C1NCC1=CC=CO1 ZZUFCTLCJUWOSV-UHFFFAOYSA-N 0.000 description 2
- 238000003709 image segmentation Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/255—Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Signal Processing (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
本发明公开一种基于视频流图像的分块信号转换与目标检测方法及装置，将视频图像进行网格分块，对分块后的二维子图像，利用分块区域内所有像素值的方差值，表示该区域在视频监控时刻的信号波动值，完成二维图像到一维信号的数据降维度处理。对于整体视频图像，转化为若干个一维信号波形进行检测，对于图像各区域的分块子图像，得到相对应的一维信号数据，从而通过实时波形的方法显示，利用数字信号处理中平滑滤波的方法降噪处理，排除子区域中背景的微弱变化，当前景目标或异常状况出现时，一维信号出现强烈的波动变化，可以明显的判断出视频流当中存在的异常波动，实现对视频检测区域中前景目标出现与否的判断。
The invention discloses a method and device for block signal conversion and target detection based on video stream images. The difference value represents the signal fluctuation value of the area at the moment of video surveillance, and completes the data dimension reduction processing from two-dimensional images to one-dimensional signals. For the overall video image, it is converted into several one-dimensional signal waveforms for detection. For the sub-images in each area of the image, the corresponding one-dimensional signal data is obtained, which is displayed by the method of real-time waveform, and the smoothing filter in digital signal processing is used. The method of noise reduction processing, eliminates the weak changes in the background in the sub-area, when the foreground target or abnormal situation occurs, the one-dimensional signal has a strong fluctuation change, which can clearly determine the abnormal fluctuation in the video stream, and realize the detection of the video. Judgment of the presence or absence of foreground objects in the area.
Description
Technical Field
The invention relates to the technical field of video image processing, in particular to a method and a device for block signal conversion and target detection based on video stream images.
Background
With the rapid development of computer vision technology in the information age, the fields of human production and life and the like, the application of video image acquisition and processing presents a burst growth state, and the foreground target detection method is based on a background image, performs differential calculation through a global image and filters background noise by using various filtering methods in the common video monitoring from the air satellite remote sensing image to the ground unmanned surface, so as to realize the judgment of an interested target. Or detecting through the speed characteristic of the moving object by using a two-dimensional optical flow field calculation method of the video image. Or the foreground target is detected and judged by using a method based on image recognition, target classification and deep learning. Due to the environmental influences such as changes in illumination, changes in background images, shaking of a camera platform, shaking of leaves in a natural scene and the like, the traditional video image detection methods need to adopt algorithms such as background updating, image shaking removal, sample image learning, a deep neural network and the like, and are completed through large-data-volume operation. The method has large calculated amount, needs to process the image global pixel points in real-time detection, has complex data calculation process, has large calculated amount on each frame of image in real-time image processing, and has low timeliness of on-line monitoring.
For example, royal, et al propose a detection alarm system for foreign body invasion in railway construction departments (royal. jinghu high-speed railway crossing or parallel existing railway construction foreign body invasion alarm technology research [ D ]. beijing: beijing university of transportation, 2010). The system is used for installing two-dimensional laser sensors at different angles on a railway construction site to form a laser light curtain, and the limit of foreign matter invasion in different directions can be detected. The author provides a foreign matter identification algorithm based on the least square method principle, and foreign matter identification is carried out according to the outer contour straight line characteristic parameters and the motion conditions of the invaded object, so that the normally running train and the invasion limit foreign matter are distinguished. On the other hand, the system comprises a wireless network based on ZigBee and GPRS, and remote configuration of field devices and remote access to alarm results are achieved. The method utilizes a two-dimensional laser scanning sensor and a scanning area point cloud distance mode to carry out detection and alarm on the intrusion of foreground targets and foreign matters in a monitoring area, and the method does not have the function of video playback and has a small detection area. And a single two-dimensional laser scanning sensor only detects the foreign matter condition of a sector area, and if a spatial area is monitored, a plurality of two-dimensional laser scanning sensors are required to be spliced into a three-dimensional area, so the cost is high.
For another example, the university of China and south has learned and proposed a railway intrusion detection method based on an intelligent video technology (learning can. railway intrusion detection [ D ] based on the intelligent video technology, Changsha: the university of China, 2010.), and the university of Beijing traffic, salix asahi, etc., and proposed a key algorithm of a train remote lookout system (the key algorithm research [ D ] of the train remote lookout system of salix asahi [ Beijing: the university of Beijing traffic, 2016). Most of the target detection and foreign object intrusion methods based on the video image processing technology are based on the traditional image processing algorithm, and the foreground target is detected by methods such as image difference or background image modeling. The intrusion detection algorithm of the systems comprises a plurality of steps of preprocessing, background extraction, automatic extraction of key monitoring areas, moving object detection, moving object intrusion detection, template matching and the like. The preprocessing includes lens distortion correction, image enhancement, and anti-shake processing. The background extraction is realized based on an optimized mathematical model, and the moving target detection comprises target feature extraction and moving target tracking. Finally, the detection of the target in the monitoring area and the extraction of the key video image can be realized. The method mainly utilizes the traditional image processing video processing mode to process more steps of images of each frame of image in the video stream, and has large data volume, large calculated amount and poor real-time video processing effectiveness. The technology needs to establish a dynamic background model, perform dynamic background updating according to the illumination change condition, and detect a foreground target through an image difference or other feature extraction methods, so the technology is sensitive to the external environment change, is easy to generate more noise interference, and affects the detection precision. Therefore, a new technical solution is needed to solve the problems in the prior art.
Disclosure of Invention
The invention aims to provide a method and a device for block signal conversion and target detection based on a video stream image. The invention uses the variance of the local image pixel value to represent the overall change condition of the video image data in the region. For the whole video image, the signal waveform can be converted into a plurality of one-dimensional signal waveforms for detection. The corresponding one-dimensional signal data is obtained for the block sub-images of each area of the image, so that the data is displayed by a real-time waveform method, noise reduction processing is performed by a smooth filtering method in digital signal processing, weak changes of the background in the sub-areas are eliminated, when a foreground target or an abnormal condition occurs, strong fluctuation changes of the one-dimensional signal occur, abnormal fluctuation existing in a video stream can be obviously judged, and the judgment of whether the foreground target occurs in the video detection area is realized.
In order to achieve the purpose, the technical scheme of the invention is as follows: a method for block signal conversion and target detection based on video stream images, the method comprising:
video data acquisition: collecting and storing video stream image data of a monitoring area by using video image data collecting equipment;
and (3) global video processing: carrying out graying processing on the video stream image, and carrying out global downsampling on the video stream image subjected to graying processing;
local signal conversion: performing video region segmentation on the video stream global image to form a sub-region, performing region mean filtering on the sub-region, and calculating the variance of the sub-region of the video stream image;
one-dimensional signal processing: and drawing the variance value of the sub-region image into a waveform according to the time lapse, carrying out signal mean filtering, calculating the difference value of the one-dimensional signal waveform, obtaining the reference value of the background image and realizing the detection of the foreground target video image.
The above method for block signal conversion and target detection based on video stream images, the global video processing method includes:
video stream image preprocessing, namely installing a monitoring camera on a fixed platform, receiving a video image acquired by the camera in real time by using a computer, and defining the video image as I (x, y, t), wherein the I (x, y) represents data of each frame of image, the I (x, y, t) is a video stream image corresponding to time t, the image width of the video stream image is W, and the image height of the video stream image is H;
performing image graying treatment, namely converting a three-channel RGB (red, green and blue) color digital image collected by a monitoring camera into a single-channel 8-bit grayscale image, namely recording the pixel grayscale value between 0 and 255 as: b (x, y, t);
the method comprises the following steps of video image global downsampling, dimensionality reduction and downsampling calculation are carried out on a global image, the scaling coefficient of the global image in the width direction and the height direction is w and h, a new downsampled image is L (x, y and t), the length and the height of the new downsampled image are M, N, and the formula is met:
M＝W/w,N＝H/h
performing Gaussian blur processing once on the image after the down sampling, finishing primary image filtering, obtaining an image G (x, y, t), wherein the gray value of a pixel in a new image G (x, y, t) and the early-stage down sampling image L (x, y, t) meet the formula:
the method for block signal conversion and target detection based on video stream images as described above, the local signal conversion method includes:
obtaining an image G (x, y, t) after Gaussian blur processing, dividing a global image G (x, y, t) into small windows with the same size by adopting a region grid dividing method, defining an image dividing coefficient as k, and dividing the global image into k multiplied by k subregions Rn(x, y, t), where n is 1, 2.., k × k, and the size of the partial image is Wn,HnSize of partial image Wn,Hn；
For sub-region R after grid blockingn(x, y, t), adopting a j multiplied by j mean filtering template to carry out noise filtering and smoothing processing on the sub-region image, and taking f (x, y, t) as the sub-region image RnMoving the smoothed window template in (x, y, t) to obtain a sub-region filtered image Fn(x,y,t)：
Calculating the average value Ave of all pixels in the image subarea at the moment tn(t)：
Calculating the variance D at the t moment by utilizing the gray values and the average values of all pixel points of the subregion imagen(t)：
In the method for block signal conversion and target detection based on video stream images, in the one-dimensional signal processing process:
carrying out signal mean filtering processing to obtain one-dimensional signal of video image which is divided into blocks according to grids
Wherein λnThe value is an odd number within 3 to 10;
by formula San(t)＝|Sn(t)-Sn(t-1) | calculating the difference of the one-dimensional signal waveforms;
by the formula
Obtaining a reference value, Sa, for the background imagen(T) is the signal value at the current time T, and when the signal value is greater than alpha times TnWhen the threshold value is reached, the foreground object is judged to appear in the video monitoring area, and the automatic trigger switch trigger for generating and storing data of the video image is used for triggering the switch triggernSatisfy the formula
Then, automatically collecting the image of the current scene, and storing the important view when the foreground target appearsAnd (4) frequency image.
The invention also provides a device for block signal conversion and target detection based on video stream images, which comprises:
the video data acquisition unit acquires and stores video stream image data of the monitoring area by using video image data acquisition equipment;
the video image data acquisition device comprises: the camera is used for video acquisition; a computer: the system is used for collecting and storing video image data and executing the algorithm designed by the invention; a display: the device is used for storing and displaying the video image and the detection result;
the global video processing unit is used for carrying out graying processing on the video stream image and carrying out global downsampling on the video stream image after the graying processing;
the local signal conversion unit is used for carrying out video area segmentation on the video stream global image to form a sub-area, carrying out area mean filtering on the sub-area and calculating the variance of the video stream image sub-area;
and the one-dimensional signal processing unit is used for drawing the variance value of the sub-region image into a waveform according to the time lapse, carrying out signal mean value filtering, calculating the difference value of the one-dimensional signal waveform, obtaining the reference value of the background image and realizing the detection of the foreground target video image.
The device for block signal conversion and target detection based on video stream images as described above, the global video processing unit includes a global down-sampling module, the global down-sampling module is configured to down-sample the video images globally, and the global down-sampling satisfies the formula:
M＝W/w,N＝H/h
in the above equation, the scaling factor in the image width direction and the same scaling factor in the height direction are W and H, the image width is W and the image height is H, and the new downsampled image length and height are M, N.
The apparatus for block signal conversion and target detection based on video stream image as described above, the local signal conversion unit includes a video region segmentation module for segmenting the global video image into a plurality of sub-regions, defining a mapThe image segmentation coefficient is k, the global video image is divided into k multiplied by k sub-region local images R by the video region segmentation modulen(x,y,t)。
The apparatus for block signal conversion and target detection based on video stream images as described above, the local signal conversion unit includes a region mean filtering module, the region mean filtering module is configured to perform region mean filtering on the sub-region, and the region mean filtering formula is
j × j represents a mean filtering template, Fn(x, y, t) represents a subregion filtered image.
In the apparatus for block signal conversion and target detection based on video stream images, the one-dimensional signal processing unit includes a foreground target detection module, and the foreground target detection module is configured to obtain a reference value of a background image to achieve foreground target video image detection, and use a formula to achieve foreground target video image detection
Obtaining a reference value, Sa, for the background imagen(T) is the signal value at the current time T, and when the signal value is greater than alpha times TnWhen the threshold value is reached, the foreground object is judged to appear in the video monitoring area, and the automatic trigger switch trigger for generating and storing data of the video image is used for triggering the switch triggernSatisfy the formula
And then, automatically acquiring the image of the current scene, and storing the important video image when the foreground target appears.
The invention has the following advantages: according to the invention, the global video image area monitored by the camera is divided into a plurality of sub-areas, and the two-dimensional image of the sub-area window is converted into a one-dimensional signal waveform, so that the data volume of real-time image processing can be simplified, and the detection speed of a video frame is improved; the invention divides the global video image into the sub-area windows to process locally, the blocking processing mode of the video stream image can provide a basis for the parallel processing of a plurality of subsequent sub-images, and the GPU can be used for processing the sub-images in parallel, thereby obviously reducing the processing time of the video stream image; the method aims at the image data of a video monitoring area, converts the image data into signal waveforms by using the mean value and the variance of the pixel values of the sub-areas, judges whether a foreground target appears in the image data or not through the fluctuation range of the pixel values in the area, and can eliminate a large amount of noise interference caused by a frame difference method or a background image difference method in the traditional image processing; the method for converting the two-dimensional image into the one-dimensional signal has better smoothing effect on obvious noise caused by illumination, micro motion and camera platform shake in the background image, and is simpler than the traditional image processing method; the invention uses the concept and method of signal processing to process the video stream image in blocks and to trigger and judge the corresponding waveform of the video stream image by the foreground object, so as to automatically detect the time of the object, automatically extract and store the video stream image when the object appears, and can be used for data playback.
Drawings
FIG. 1 is a schematic diagram of a block signal conversion and target detection method based on video stream images;
FIG. 2 is a schematic diagram of high-speed periodic mechanical motion of a block signal conversion and target detection method based on video stream images;
FIG. 3 is a schematic diagram of video region segmentation based on a block signal conversion and target detection method of video stream images;
FIG. 4 is a schematic diagram of a video image subregion mean value waveform of a block signal conversion and target detection method based on a video stream image;
FIG. 5 is a diagram of a video image subregion variance waveform for a block signal conversion and target detection method based on video stream images
FIG. 6 is a video image subregion difference waveform based on block signal conversion and target detection of a video stream image;
fig. 7 is a schematic diagram of foreground target images in video image block areas based on a block signal conversion and target detection method for video stream images.
Detailed Description
The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.
As shown in fig. 1 and fig. 2, a method for block signal conversion and object detection based on video stream images includes:
video data acquisition: collecting and storing video stream image data of a monitoring area by using video image data collecting equipment;
and (3) global video processing: carrying out graying processing on the video stream image, and carrying out global downsampling on the video stream image subjected to graying processing;
local signal conversion: performing video region segmentation on the video stream global image to form a sub-region, performing region mean filtering on the sub-region, and calculating the variance of the sub-region of the video stream image;
one-dimensional signal processing: and drawing the variance value of the sub-region image into a waveform according to the time lapse, carrying out signal mean filtering, calculating the difference value of the one-dimensional signal waveform, obtaining the reference value of the background image and realizing the detection of the foreground target video image.
In an embodiment of a method for block signal conversion and target detection based on video stream images, a global video processing method includes:
video stream image preprocessing, namely installing a monitoring camera on a fixed platform, receiving a video image acquired by the camera in real time by using a computer, and defining the video image as I (x, y, t), wherein the I (x, y) represents data of each frame of image, the I (x, y, t) is a video stream image corresponding to time t, the image width of the video stream image is W, and the image height of the video stream image is H;
performing image graying treatment, namely converting a three-channel RGB (red, green and blue) color digital image collected by a monitoring camera into a single-channel 8-bit grayscale image, namely recording the pixel grayscale value between 0 and 255 as: b (x, y, t);
the images collected by video monitoring have higher resolution, namely, a large number of pixel points are involved in calculation. In the real-time image processing, the invention utilizes the whole change of pixel values in the sub-area to calculate, so that the dimensionality reduction and down-sampling calculation is firstly carried out on the global image. Since the pixel value is an integer value, the scaling factor with the same image width direction and height direction is set to be w, h, the new downsampled image is L (x, y, t), the image length and height thereof are M, N, the global downsampling of the video image is met, dimensionality reduction and downsampling calculation are performed on the global image, the scaling factor with the same image width direction and height direction is w, h, the new downsampled image is L (x, y, t), the length and height of the new downsampled image is M, N, and the formula is met:
M＝W/w,N＝H/h
where, the image length M and the height N are integers, i.e. it means that W and H can be evenly divided by W and H, respectively. Then, performing a gaussian blur processing on the down-sampled image for once to complete the filtering of the primary image, obtaining an image G (x, y, t), and reducing the interference of background noise, wherein the pixel gray value in the new image and the down-sampled image L (x, y, t) in the previous period, performing a gaussian blur processing on the down-sampled image for once to complete the filtering of the primary image, obtaining an image G (x, y, t), and the pixel gray value in the new image G (x, y, t) and the down-sampled image L (x, y, t) in the previous period, and satisfying the formula:
in one embodiment of a method for block signal conversion and object detection based on video stream images, a local signal conversion method comprises:
referring to fig. 1 again, after gaussian blurring processing, an image G (x, y, t) is obtained, a region mesh division method is adopted to divide the global image G (x, y, t) into small windows with the same size, an image division coefficient is defined as k, and the global image is divided into k × k sub-regions Rn(x, y, t), where n is 1, 2.., k × k, and the size of the partial image is Wn,HnSize of partial image Wn,Hn(ii) a Referring to fig. 3, the effect of dividing a global video image into 2 x 2 grid windows according to the present invention is shown.
Referring to fig. 4, waveforms for processing the video image of fig. 3 using the present invention are shown. It can be seen that the background of the sub-window has dynamic backgrounds of trees, grasslands, vegetation, illumination change and the like, and the mean waveform of the background has narrow fluctuation before a foreground target appears; but do notWhen the foreground object appears, the foreground object has obvious fluctuation. For sub-region R after grid blockingn(x, y, t), adopting a j multiplied by j mean filtering template to carry out noise filtering and smoothing processing on the sub-region image, and taking f (x, y, t) as the sub-region image RnMoving the smoothed window template in (x, y, t) to obtain a sub-region filtered image Fn(x,y,t)：
Calculating the average value Ave of all pixels in the image subarea at the moment tn(t)：
Calculating the variance D at the t moment by utilizing the gray values and the average values of all pixel points of the subregion imagen(t)：
Referring to fig. 5, according to the method, video stream image data is processed in a grid blocking manner, and a two-dimensional image in a sub-window area is converted into a one-dimensional signal through variance value calculation, which represents the fluctuation situation of the pixel value in the area of the monitored video at the time t. The variance value represents the energy fluctuation of pixel values in the area, and the fluctuation of the target appearance time can be more obvious than the average value waveform.
In an embodiment of a method for block signal conversion and target detection based on video stream images, in the process of one-dimensional signal processing:
one-dimensional signal D for a video image signaled in grid blocksn(t), due to the occurrence of noise in actual detection, a larger first derivative deviation at adjacent moments is caused, and the normal judgment of the foreground target is influenced. Using λ in the embodiments of the inventionnThree neighborhood mean filtering of 3, where λ can be chosen to be larger thanAn odd number equal to 3, but the larger the range of neighborhood mean filtering, the more smooth the signal waveform is, and the judgment on the foreground target is reduced, generally lambdanThe value is an odd number within 3 to 10; then, signal mean filtering processing is carried out to obtain one-dimensional signals of video images which are signalized according to grid blocks
By formula San(t)＝|Sn(t)-Sn(t-1) | calculating the difference of the one-dimensional signal waveforms; the one-dimensional waveform obtained by calculating the variance of the video image subareas can smooth a small amount of noise interference after filtering, and the dynamic waveform of the corresponding window is obtained. When no foreground object appears, the dynamic waveform only changes slightly; when the foreground object appears, the waveform will generate larger fluctuation.
In the difference waveform of the adjacent data of the one-dimensional signal, when the foreground object appears, a larger step signal is generated, as shown in fig. 6, for the signal with such large fluctuation, it can be determined that it has exceeded the fluctuation range of the background tiny noise interference, that is, it is defined as the appearance of the foreground object. Definition of tau in the examples of the inventionnThe average value of the signal waveform corresponding to 10 frames of video images is calculated by the formula
Obtaining a reference value T of a background imagen，San(T) is the signal value at the current time T, and when the signal value is greater than alpha times TnWhen the threshold value is reached, the foreground object is judged to appear in the video monitoring area, and the automatic trigger switch trigger for generating and storing data of the video image is used for triggering the switch triggernSatisfy the formula
And then, automatically acquiring the image of the current scene, and storing the important video image when the foreground target appears.
At this moment, the automatic trigger switch trigger for storing the video image generation datanThe current scene is automatically acquired, namelyAnd saving the important video image when the foreground object appears. Fig. 7 shows an example of a foreground object video image detected and stored by the present invention, in which the window 2 and the window 4 are sub-images of the whole monitoring area, corresponding to the mesh partition area in fig. 3. The one-dimensional signal waveform of the corresponding sub-region has obvious signal fluctuation with reference to fig. 4, 5 and 6 again, and according to the method of the present invention, the image when the foreground object appears can be accurately detected and stored.
The invention also provides a block signal conversion and target detection device based on video stream images, which comprises:
the video data acquisition unit acquires and stores video stream image data of the monitoring area by using video image data acquisition equipment; wherein, video image data acquisition equipment includes: the camera is used for video acquisition; a computer: the system is used for collecting and storing video image data and executing the algorithm designed by the invention; a display: the device is used for storing and displaying the video image and the detection result;
the global video processing unit is used for carrying out graying processing on the video stream image and carrying out global downsampling on the video stream image after the graying processing;
the local signal conversion unit is used for carrying out video area segmentation on the video stream global image to form a sub-area, carrying out area mean filtering on the sub-area and calculating the variance of the video stream image sub-area;
and the one-dimensional signal processing unit is used for drawing the variance value of the sub-region image into a waveform according to the time lapse, carrying out signal mean value filtering, calculating the difference value of the one-dimensional signal waveform, obtaining the reference value of the background image and realizing the detection of the foreground target video image.
As above, the global video processing unit includes a global down-sampling module, the global down-sampling module is used for global down-sampling the video image, and the global down-sampling satisfies the formula:
M＝W/w,N＝H/h
in the above equation, the scaling factor in the image width direction and the same scaling factor in the height direction are W and H, the image width is W and the image height is H, and the new downsampled image length and height are M, N.
In an embodiment of the apparatus for block signal conversion and target detection based on video stream image, the local signal conversion unit includes a video region segmentation module, the video region segmentation module is configured to segment the global video image into a plurality of sub-regions, an image segmentation coefficient is defined as k, and then the global video image is divided into k × k sub-region local images R by the video region segmentation modulen(x,y,t)。
In an embodiment of the apparatus for block signal conversion and target detection based on video stream images, the local signal conversion unit comprises a region mean filtering module, the region mean filtering module is configured to perform region mean filtering on the sub-regions, and the region mean filtering formula is
j × j represents a mean filtering template, Fn(x, y, t) represents a subregion filtered image.
In an embodiment of the apparatus for block signal conversion and target detection based on video stream images, the one-dimensional signal processing unit includes a foreground target detection module, the foreground target detection module is used for obtaining a reference value of a background image to realize foreground target video image detection, and the formula is used for detecting the foreground target video image
Obtaining a reference value, Sa, for the background imagen(T) is the signal value at the current time T, and when the signal value is greater than alpha times TnWhen the threshold value is reached, the foreground object is judged to appear in the video monitoring area, and the automatic trigger switch trigger for generating and storing data of the video image is used for triggering the switch triggernSatisfy the formula
And then, automatically acquiring the image of the current scene, and storing the important video image when the foreground target appears.
According to the invention, the global video image area monitored by the camera is divided into a plurality of sub-areas, and the two-dimensional image of the sub-area window is converted into a one-dimensional signal waveform, so that the data volume of real-time image processing can be simplified, and the detection speed of a video frame is improved; the invention divides the global video image into the sub-area windows to process locally, the blocking processing mode of the video stream image can provide a basis for the parallel processing of a plurality of subsequent sub-images, and the GPU can be used for processing the sub-images in parallel, thereby obviously reducing the processing time of the video stream image; the method aims at the image data of a video monitoring area, converts the image data into signal waveforms by using the mean value and the variance of the pixel values of the sub-areas, judges whether a foreground target appears in the image data or not through the fluctuation range of the pixel values in the area, and can eliminate a large amount of noise interference caused by a frame difference method or a background image difference method in the traditional image processing; the method for converting the two-dimensional image into the one-dimensional signal has better smoothing effect on obvious noise caused by illumination, micro motion and camera platform shake in the background image, and is simpler than the traditional image processing method; the invention uses the concept and method of signal processing to process the video stream image in blocks and to trigger and judge the corresponding waveform of the video stream image by the foreground object, so as to automatically detect the time of the object, automatically extract and store the video stream image when the object appears, and can be used for data playback.
Although the invention has been described in detail above with reference to a general description and specific examples, it will be apparent to one skilled in the art that modifications or improvements may be made thereto based on the invention. Accordingly, such modifications and improvements are intended to be within the scope of the invention as claimed.
Claims (9)
1. A method for block signal conversion and target detection based on video stream images, the method comprising:
video data acquisition: collecting and storing video stream image data of a monitoring area by using video image data collecting equipment;
and (3) global video processing: carrying out graying processing on the video stream image, and carrying out global downsampling on the video stream image subjected to graying processing;
local signal conversion: performing video region segmentation on the video stream global image to form a sub-region, performing region mean filtering on the sub-region, and calculating the variance of the sub-region of the video stream image;
one-dimensional signal processing: and drawing the variance value of the sub-region image into a waveform according to the time lapse, carrying out signal mean filtering, calculating the difference value of the one-dimensional signal waveform, obtaining the reference value of the background image and realizing the detection of the foreground target video image.
2. The method of claim 1, wherein the global video processing method comprises:
video stream image preprocessing, namely installing a monitoring camera on a fixed platform, receiving a video image acquired by the camera in real time by using a computer, and defining the video image as I (x, y, t), wherein the I (x, y) represents data of each frame of image, the I (x, y, t) is a video stream image corresponding to time t, the image width of the video stream image is W, and the image height of the video stream image is H;
performing image graying treatment, namely converting a three-channel RGB (red, green and blue) color digital image collected by a monitoring camera into a single-channel 8-bit grayscale image, namely recording the pixel grayscale value between 0 and 255 as: b (x, y, t);
the method comprises the following steps of video image global downsampling, dimensionality reduction and downsampling calculation are carried out on a global image, the scaling coefficient of the global image in the width direction and the height direction is w and h, a new downsampled image is L (x, y and t), the length and the height of the new downsampled image are M, N, and the formula is met:
M＝W/w,N＝H/h
performing Gaussian blur processing once on the image after the down sampling, finishing primary image filtering, obtaining an image G (x, y, t), wherein the gray value of a pixel in a new image G (x, y, t) and the early-stage down sampling image L (x, y, t) meet the formula:
3. the method of claim 1, wherein the local signal conversion method comprises:
obtaining an image G (x, y, t) after Gaussian blur processing, dividing a global image G (x, y, t) into small windows with the same size by adopting a region grid dividing method, defining an image dividing coefficient as k, and dividing the global image into k multiplied by k subregions Rn(x, y, t), where n is 1, 2.., k × k, and the size of the partial image is Wn,Hn；
For sub-region R after grid blockingn(x, y, t), adopting a j multiplied by j mean filtering template to carry out noise filtering and smoothing processing on the sub-region image, and taking f (x, y, t) as the sub-region image RnMoving the smoothed window template in (x, y, t) to obtain a sub-region filtered image Fn(x,y,t)：
Calculating the average value Ave of all pixels in the image subarea at the moment tn(t)：
Calculating the variance D at the t moment by utilizing the gray values and the average values of all pixel points of the subregion imagen(t)：
4. The method of claim 1, wherein the one-dimensional signal processing comprises:
carry out signalMean value filtering processing to obtain one-dimensional signal of video image block-by-block signalization according to grid
Wherein λnThe value is an odd number within 3 to 10;
by formula San(t)＝|Sn(t)-Sn(t-1) | calculating the difference of the one-dimensional signal waveforms;
by the formula
Obtaining a reference value, Sa, for the background imagen(T) is the signal value at the current time T, and when the signal value is greater than alpha times TnWhen the threshold value is reached, the foreground object is judged to appear in the video monitoring area, and the automatic trigger switch trigger for generating and storing data of the video image is used for triggering the switch triggernSatisfy the formula
And then, automatically acquiring the image of the current scene, and storing the important video image when the foreground target appears.
5. Apparatus for block signal conversion and object detection based on video stream images, the apparatus comprising:
the video data acquisition unit acquires and stores video stream image data of the monitoring area by using video image data acquisition equipment; the video image data acquisition apparatus includes: the camera is used for video acquisition; a computer: the system is used for acquiring and storing video image data and executing an algorithm; a display: the device is used for storing and displaying the video image and the detection result;
the global video processing unit is used for carrying out graying processing on the video stream image and carrying out global downsampling on the video stream image after the graying processing;
the local signal conversion unit is used for carrying out video area segmentation on the video stream global image to form a sub-area, carrying out area mean filtering on the sub-area and calculating the variance of the video stream image sub-area;
and the one-dimensional signal processing unit is used for drawing the variance value of the sub-region image into a waveform according to the time lapse, carrying out signal mean value filtering, calculating the difference value of the one-dimensional signal waveform, obtaining the reference value of the background image and realizing the detection of the foreground target video image.
6. The apparatus according to claim 5, wherein the global video processing unit comprises a global down-sampling module, and the global down-sampling module is configured to down-sample the video image globally, and the global down-sampling satisfies the following formula:
M＝W/w,N＝H/h
in the above equation, the scaling factor in the image width direction and the same scaling factor in the height direction are W and H, the image width is W and the image height is H, and the new downsampled image length and height are M, N.
7. The apparatus according to claim 5, wherein the local signal conversion unit comprises a video region dividing module for dividing the global video image into a plurality of sub-regions, and defining an image dividing coefficient as k, so that the global video image is divided into k × k sub-region local images R by the video region dividing modulen(x,y,t)。
8. The apparatus according to claim 5, wherein the local signal conversion unit comprises a region mean filter module for performing region mean filtering on the sub-regions, the region mean filter being expressed by
j × j represents a mean filtering template, Fn(x, y, t) represents a subregion filtered image.
9. The apparatus according to claim 5, wherein the one-dimensional signal processing unit comprises a foreground object detection module, the foreground object detection module is used for obtaining a reference value of a background image to realize foreground object video image detection, and the formula is used for
Obtaining a reference value, Sa, for the background imagen(T) is the signal value at the current time T, and when the signal value is greater than alpha times TnWhen the threshold value is reached, the foreground object is judged to appear in the video monitoring area, and the automatic trigger switch trigger for generating and storing data of the video image is used for triggering the switch triggernSatisfy the formula
And then, automatically acquiring the image of the current scene, and storing the important video image when the foreground target appears.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810201588.5A CN108596169B (en) | 2018-03-12 | 2018-03-12 | Block signal conversion and target detection method and device based on video stream image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810201588.5A CN108596169B (en) | 2018-03-12 | 2018-03-12 | Block signal conversion and target detection method and device based on video stream image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108596169A CN108596169A (en) | 2018-09-28 |
| CN108596169B true CN108596169B (en) | 2021-05-14 |
Family
ID=63626105
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810201588.5A Active CN108596169B (en) | 2018-03-12 | 2018-03-12 | Block signal conversion and target detection method and device based on video stream image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108596169B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109589071B (en) * | 2018-12-05 | 2022-03-29 | 余姚市恒正金属制品有限公司 | Water pressure regulating and controlling system for spray head |
| CN109691922B (en) * | 2018-12-27 | 2021-04-20 | 安徽舜邦精细化工有限公司 | Automatic discharging mechanism for shampoo |
| CN110598239B (en) * | 2019-07-19 | 2023-11-17 | 中国建设基础设施有限公司 | Application method based on point cloud big data in track area |
| CN111063146A (en) * | 2019-12-18 | 2020-04-24 | 浙江大华技术股份有限公司 | Defense method and equipment for monitoring picture and storage device |
| CN113011218B (en) * | 2019-12-19 | 2024-03-01 | 合肥君正科技有限公司 | Method for automatically updating background of personnel getting on or off in-car monitoring detection |
| CN111524158B (en) * | 2020-05-09 | 2023-03-24 | 黄河勘测规划设计研究院有限公司 | Method for detecting foreground target in complex scene of hydraulic engineering |
| CN113744137B (en) * | 2020-05-27 | 2024-05-31 | 合肥君正科技有限公司 | Frame difference smoothing method of spiral matrix |
| CN112801049B (en) * | 2021-03-26 | 2021-07-23 | 北京至真互联网技术有限公司 | Image classification method, device and equipment |
| CN113822240B (en) * | 2021-11-22 | 2022-03-25 | 广东电网有限责任公司中山供电局 | Method and device for extracting abnormal behaviors from power field operation video data |
| CN115856771A (en) * | 2022-11-22 | 2023-03-28 | 哈尔滨工程大学 | A method for joint estimation of near-field source number and positioning parameters |
| CN116071220B (en) * | 2023-03-06 | 2023-06-20 | 浙江华感科技有限公司 | Image window data processing method, device, equipment and medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102288166A (en) * | 2011-05-13 | 2011-12-21 | 河海大学 | Video-based multi-model combined surface ship detection method |
| CN104103039A (en) * | 2013-04-08 | 2014-10-15 | 上海联影医疗科技有限公司 | Image noise estimation method |
| CN107767363A (en) * | 2017-09-05 | 2018-03-06 | 天津大学 | It is a kind of based on natural scene without refer to high-dynamics image quality evaluation algorithm |
-
2018
- 2018-03-12 CN CN201810201588.5A patent/CN108596169B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102288166A (en) * | 2011-05-13 | 2011-12-21 | 河海大学 | Video-based multi-model combined surface ship detection method |
| CN104103039A (en) * | 2013-04-08 | 2014-10-15 | 上海联影医疗科技有限公司 | Image noise estimation method |
| CN107767363A (en) * | 2017-09-05 | 2018-03-06 | 天津大学 | It is a kind of based on natural scene without refer to high-dynamics image quality evaluation algorithm |
Non-Patent Citations (2)
| Title |
|---|
| 《Sparsity and Structure in Hyperspectral Imaging:Sensing,reconstruction,and Target Detection》;Rebecca M.Willett等;《IEEE Signal Processing Magazine》;20131203;第31卷(第1期);全文 * |
| 《基于亮度直方图匹配的运动目标检测算法》;薛丽霞 等;《计算机工程与应用》;20120321;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108596169A (en) | 2018-09-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108596169B (en) | Block signal conversion and target detection method and device based on video stream image | |
| CN112800860B (en) | High-speed object scattering detection method and system with coordination of event camera and visual camera | |
| CN102663869B (en) | Indoor fire detection method based on video monitoring platform | |
| CN103077539B (en) | Motion target tracking method under a kind of complex background and obstruction conditions | |
| CN100474337C (en) | Noise-possessing movement fuzzy image restoration method based on radial basis nerve network | |
| US10032283B2 (en) | Modification of at least one parameter used by a video processing algorithm for monitoring of a scene | |
| EP2549759B1 (en) | Method and system for facilitating color balance synchronization between a plurality of video cameras as well as method and system for obtaining object tracking between two or more video cameras | |
| CN102567979A (en) | Vehicle-mounted infrared night vision system and multi-source images fusing method thereof | |
| CN109359593B (en) | Rain and snow environment picture fuzzy monitoring and early warning method based on image local grid | |
| CN107194946B (en) | FPGA-based infrared salient object detection method | |
| CN110415268A (en) | A Foreground Image Algorithm Based on the Combination of Background Difference Method and Interframe Difference Method | |
| CN114332644B (en) | A large field of view traffic density acquisition method based on video satellite data | |
| CN111582076A (en) | Picture freezing detection method based on pixel motion intelligent perception | |
| CN111028263B (en) | Moving object segmentation method and system based on optical flow color clustering | |
| CN115410324A (en) | Car as a house night security system and method based on artificial intelligence | |
| CN118096815A (en) | Road abnormal event detection system based on machine vision | |
| CN112770090A (en) | Monitoring method based on transaction detection and target tracking | |
| CN106056078A (en) | Crowd density estimation method based on multi-feature regression ensemble learning | |
| CN119763073A (en) | Rail foreign matter detection method and system | |
| CN103093481B (en) | A kind of based on moving target detecting method under the static background of watershed segmentation | |
| KR101520293B1 (en) | Scheduling method for detention of object and of obtaining forensic image by visual attention, and system thereof | |
| Chaiyawatana et al. | Robust object detection on video surveillance | |
| CN113111699B (en) | Human body motion detection method based on infrared camera | |
| CN103259964A (en) | Image signal filtering method of video dynamic target capture | |
| Pratama et al. | Vehicle counting and classification for traffic data acquisition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20241120 Address after: Room 1518B, Unit 2, 12th Floor, Huizhi Building, No. 9 Xueqing Road, Haidian District, Beijing, 100085 Patentee after: Beijing contention Technology Co.,Ltd. Country or region after: China Address before: Beijing Jianzhu University, No.1 zhanzhanguan Road, Xicheng District, Beijing 100044 Patentee before: Beijing University of Civil Engineering and Architecture Country or region before: China Patentee before: Beijing Jiaotong University |