CN114820698A - Formation detection method and device for large-scale event moving phalanx - Google Patents

Abstract

The present invention provides a method and device for detecting the formation of a large-scale event moving square. The method for detecting the formation of a large-scale moving square includes: acquiring a target video of a target moving square in a target environment; The target video frame in the video and the diagonal grayscale projection standard array are used to generate the jitter factor corresponding to the target video frame; in the case where the jitter factor is lower than the target jitter threshold, the target motion is generated based on the target video frame. The target image of the square matrix. The method for detecting the formation of a large-scale moving square matrix of the present invention generates a jitter factor based on the target video frame in the target video and a standard array of diagonal grayscale projections, and screens the target video frame based on the jitter factor to preserve the jitter. A target video frame with a lower amplitude is processed, and a target video frame with a lower jitter amplitude is processed to generate a target image of the target moving square matrix, which effectively improves the imaging resolution and helps to improve the detection result.

Description

Formation detection method and device for large-scale event moving phalanx

technical field

The invention relates to the technical field of image processing, and in particular, to a method and device for formation detection of a large-scale moving square matrix.

Background technique

The real-time detection of moving squares based on video images in large-scale events is mainly to capture video images through panoramic cameras set up at high positions in outdoor venues, and to perform real-time detection of large-area moving squares. However, in the related art, since the photographing equipment (especially outdoors) is easily affected by external factors such as strong wind or shaking of the support frame, it will cause shaking, which will cause problems such as shaking and ghosting of the video picture, and cause the positional displacement of the objects in the picture. Picture distortion affects the clarity of the image, thus affecting the detection results.

SUMMARY OF THE INVENTION

The invention provides a formation detection method and device for a large-scale moving square matrix, which are used to solve the defect of poor detection effect of the moving square matrix in the prior art and improve the detection effect.

The present invention provides a method for detecting the formation of a large-scale moving square matrix, comprising:

Obtain the target video of the target moving square matrix in the target environment;

generating a jitter factor corresponding to the target video frame based on the target video frame and the diagonal grayscale projection standard array in the target video;

In the case where the jitter factor is lower than a target jitter threshold, a target image of the target motion square matrix is generated based on the target video frame.

According to a method for detecting the formation of a large-scale moving square matrix provided by the present invention, the jitter factor corresponding to the target video frame is generated based on the target video frame in the target video and the diagonal grayscale projection standard array. ,include:

Based on the target area of the target video frame, a first diagonal grayscale projection array of the target area is generated, and the target area does not include the feature of the motion square;

The dithering factor is generated based on the first diagonal grayscale projection array and the diagonal grayscale projection criterion array.

According to a method for detecting the formation of a large-scale moving square matrix provided by the present invention, the dithering factor is generated based on the first diagonal grayscale projection array and the diagonal grayscale projection standard array, including: :

Apply the formula:

Generate the dither factor, where θ is the dither factor, the Gr_bi(m) is the diagonal grayscale projection standard value of the mth position in the diagonal grayscale projection standard array, and the Gr_ci(m ) is the first diagonal grayscale projection value of the mth position in the first diagonal grayscale projection array, and d is the side length of the target area.

According to a method for detecting the formation of a large-scale activity moving square matrix provided by the present invention, the generating a target image of the target moving square matrix based on the target video frame includes:

performing frame difference operation on the target video frame and the target background model to generate a first image;

performing binarization processing and noise reduction processing on the first image to generate a third image;

A foreground image in the target video frame is extracted based on the third image, and a target image of the target motion square matrix is generated based on the foreground image.

According to a method for detecting the formation of a large-scale activity moving square matrix provided by the present invention, the generating the target image of the target moving square matrix based on the foreground image includes:

converting the foreground image into a target format to generate a fourth image;

Performing shadow region segmentation processing and erosion expansion processing based on the target shadow threshold on the fourth image to generate a target image of the target movement square matrix.

According to a method for detecting the formation of a large-scale activity moving square array provided by the present invention, before the acquisition of the target video of the target moving square array in the target environment, the method further includes:

acquiring multiple frames of environmental background images within the first target time period under the target environment;

generating a target background model based on the multi-frame environment background images;

Based on the target background model, a standard array of diagonal grayscale projections corresponding to the target background model is generated.

The present invention also provides a formation detection device for a large-scale activity moving square, including:

The first acquisition module is used to acquire the target video of the target motion square matrix under the target environment;

a first generation module, configured to generate a jitter factor corresponding to the target video frame based on the target video frame and the diagonal grayscale projection standard array in the target video;

A second generating module, configured to generate a target image of the target motion square matrix based on the target video frame when the jitter factor is lower than the target jitter threshold.

The present invention also provides an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, when the processor executes the program, the large-scale activities as described above are implemented by the processor. The steps of the formation detection method of the moving phalanx.

The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements the formation detection method for a large-scale activity moving square array as described above.

The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements the formation detection method for a large-scale activity moving square array as described above.

The method and device for detecting the formation of a large-scale moving square matrix provided by the present invention generate a dither factor based on the target video frame in the target video and a standard array of diagonal grayscale projections, and screen the target video frame based on the dither factor, In order to retain the target video frame with low jitter amplitude, and process the target video frame with low jitter amplitude, the target image of the target motion square matrix is generated, which overcomes the defect that other grayscale projection methods are easily affected by local motion, and realizes the The effective jitter detection of target video frames in a moving state significantly improves the rapidity and real-time performance of video frame jitter detection, and improves the imaging effect of target images, thereby helping to improve the accuracy and precision of detection results.

Description of drawings

In order to explain the present invention or the technical solutions in the prior art more clearly, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are of the present invention. For some embodiments of the present invention, for those of ordinary skill in the art, other drawings can also be obtained from these drawings without any creative effort.

Fig. 1 is one of the schematic flow charts of the formation detection method of the large-scale activity moving square matrix provided by the present invention;

Fig. 2 is the second schematic flow chart of the formation detection method of the large-scale activity moving square matrix provided by the present invention;

Fig. 3 is one of the principle schematic diagrams of the formation detection method of the large-scale activity moving square matrix provided by the present invention;

Fig. 4 is the second principle schematic diagram of the formation detection method of the large-scale activity moving square matrix provided by the present invention;

5 is the third schematic diagram of the principle diagram of the formation detection method of the large-scale activity moving square array provided by the present invention;

6 is the fourth schematic diagram of the formation detection method of the large-scale activity moving square array provided by the present invention;

7 is the fifth schematic diagram of the formation detection method of the large-scale activity moving square array provided by the present invention;

8 is the sixth schematic diagram of the principle diagram of the formation detection method of the large-scale activity moving square array provided by the present invention;

9 is a schematic structural diagram of a formation detection device for a large-scale event moving square array provided by the present invention;

FIG. 10 is a schematic structural diagram of an electronic device provided by the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention. , not all examples. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

The formation detection method of the large-scale moving square array of the present invention will be described below with reference to FIGS. 1 to 6 .

The execution body of the method for detecting the formation of a large-scale moving square matrix may be a server, or a user's terminal, such as a mobile phone or a computer.

As shown in FIG. 1 , the method for detecting the formation of the large-scale moving square matrix includes: step 110 , step 120 and step 130 .

Step 110, obtaining the target video of the target motion square matrix under the target environment;

In this step, the target environment is the scene where the target movement square is located.

The target moving square is a square that needs to be detected among the multiple moving squares.

The target video is a real-time video stream including the target moving square, and the target video may include multiple frames of images, and the image includes the target moving square.

For example, in a sports meeting, the target environment is the environment of the sports field, and the sports square is the sports square composed of athletes.

It can be understood that the moving square can be in a moving state.

The target video may be collected by an image sensor, where the image sensor may be a panoramic camera or other image sensor, which is not limited in this application.

In the actual implementation process, the image sensor needs to be set up in a high position to ensure that the image of the entire target moving square can be collected.

In some embodiments, after step 110, the method may further include: performing distortion correction on each video frame of the target video.

In this embodiment, the distortion correction includes radial distortion and tangential distortion.

You can use the formula:

其中(x²+y²＝r²)；

where (x ² +y ² =r ² );

Perform radial distortion correction,

Among them, (x', y') represents the distorted pixel coordinates of each video frame in the target video, (x, y) represents the corrected and distorted pixel coordinates, and k1, k2 and k3 are calibrated by the image sensor. The resulting distortion parameters.

You can use the formula:

其中(x²+y²＝r²)；

where (x ² +y ² =r ² );

tangential distortion correction,

Among them, (x', y') represents the distorted pixel coordinates of each video frame in the target video, (x, y) represents the corrected and distorted pixel coordinates, and p1 and p3 are calibrated by the image sensor. Distortion parameter.

It should be noted that the distortion parameter can be obtained by calibrating the image sensor.

In the actual execution process, prior to step 110, the image sensor may be calibrated preferentially to obtain the distortion parameter and the internal parameter matrix of the image sensor.

In this embodiment, by performing distortion correction on the target video, it is possible to avoid picture distortion caused during the image acquisition process using the panoramic camera, which helps to improve the imaging effect and the imaging authenticity.

After acquiring the target video, the target video can be stored in the local server or the cloud server, and can be retrieved when needed.

Step 120, based on the target video frame and the diagonal grayscale projection standard array in the target video, generate the corresponding jitter factor of the target video frame;

In this step, the target video frame is the video frame that needs to be detected.

The target video frame can be any image frame in the target video.

The diagonal grayscale projection standard array is the diagonal grayscale projection value of the image without dithering, that is, the initial value of the diagonal grayscale projection value.

The diagonal grayscale projection standard array is a value determined in advance before detection. The present invention will describe the steps of determining the diagonal grayscale projection standard array in detail in subsequent embodiments, which will not be repeated here.

The jitter factor is used to characterize the jitter degree of the target video.

It can be understood that, affected by external environmental factors, different video frames in the same target video may have different degrees of jitter.

In this embodiment, the jitter factor corresponding to the target video frame can be determined based on the target video frame in the target video and the diagonal grayscale projection standard array.

The specific implementation manner of this step will be described below.

In some embodiments, step 120 may include:

Based on the target area of the target video frame, a first diagonal grayscale projection array of the target area is generated, and the target area does not include the feature of the moving square;

A dithering factor is generated based on the first diagonal grayscale projection array and the diagonal grayscale projection criterion array.

In this embodiment, the target area is any area in the target video frame except the moving square matrix area, and the number of target areas may be one or more.

The target area can be customized based on the user.

For example, the target area can be set as a square area on the four corners of the target video frame, and the side length of each square area is d; or the target area can be set as a square area on any two corners of the target video frame; or the target The area is set to a square area on either corner of the target video frame.

By selecting the square areas on the four corners of the target video frame, it is possible to avoid the influence on the shake detection caused by the movement of the large moving square matrix in the middle of the video frame.

The first diagonal grayscale projection array is the actual diagonal grayscale projection value corresponding to the target video frame in the target video.

The first diagonal grayscale projection array may be represented by Gr, wherein the number of the first diagonal grayscale projection array is consistent with the number of target regions.

For example, in the case of determining the square area on the four corners of the target video frame as the target area, calculate the first diagonal grayscale projection array of the four target areas of the target video frame, using Gr_c1, Gr_c2, Gr_c3 and Gr_c4 representation. Among them, Gr_c1 is used to characterize the diagonal grayscale projection array of the upper left corner of the target video, Gr_c2 is used to characterize the diagonal grayscale projection array of the upper right corner of the target video, and Gr_c3 is used to characterize the diagonal grayscale projection array of the lower right corner of the target video. Gr_c4 is used to characterize the diagonal grayscale projection array in the lower left corner of the target video.

Then, the weighted variance factors between the first diagonal grayscale projection array and the diagonal grayscale projection standard array of the four target regions of the current target video frame are calculated respectively.

In some embodiments, the maximum value among the weighted variance factors of the four target regions can be selected as the jitter factor θ to reflect the most real shaking situation.

That is,

θ _max = max(θ ₁ , θ ₂ , θ ₃ , θ ₄ );

Among them, θ _max is the jitter factor, and θ ₁ , θ ₂ , θ ₃ , and θ ₄ are the weighted variance factors corresponding to the four target regions, respectively.

It can be understood that in the case where the diagonal grayscale projection standard arrays are multiple groups, the first diagonal grayscale projection array and each diagonal grayscale projection standard array are calculated respectively, and the obtained multiple dither factors are calculated. The maximum value in is used as the final jitter factor.

For example, in the case of 1000 sets of diagonal grayscale projection standard arrays, the first diagonal grayscale projection array and 1000 sets of diagonal grayscale projection standard arrays are calculated respectively, and the maximum value is selected as the dithering factor.

In some embodiments, the dithering factor is generated based on the first diagonal grayscale projection array and the diagonal grayscale projection criterion array, including:

Apply the formula:

Generate a dither factor, where θ is the dither factor, Gr_bi(m) is the diagonal grayscale projection standard value of the mth position in the diagonal grayscale projection standard array corresponding to the target area i, and Gr_ci(m) is the target area. The first diagonal grayscale projection value of the mth position in the first diagonal grayscale projection array corresponding to i, d is the side length of the target area, "?m≤d-1:2d-m-1" uses To determine the value of m at the denominator position in the formula, specifically: when m≤d-1, m=d-1 in the denominator "m+1"; otherwise, m=d in the denominator "m+1" 2d-m-1.

After the jitter factor is obtained, step 130 can be executed.

During the research and development process, the inventor found that in the related art, there are methods based on feature point matching and methods based on optical flow to detect jitter, but both rely on the detected feature points, which usually involve a very large amount of calculation, and the feature points The quality of detection will also affect the accuracy of the optical flow algorithm, and the movement of objects in the picture will also cause the optical flow algorithm to produce wrong estimates, resulting in poor final image quality.

In this application, by extracting the target area of the target video frame (such as the four corners of the target video frame) for jitter detection, the influence of local motion (such as the influence of the central motion square of the target video frame) can be effectively avoided, Suitable for dynamic shooting environments.

In addition, selecting the value with the largest jitter degree as the jitter degree of the entire target video frame can effectively avoid the interference of the moving objects in the target video frame to the jitter detection.

Step 130 , under the condition that the jitter factor is lower than the target jitter threshold, based on the target video frame, generate a target image of the target motion square matrix.

In this step, the target jitter threshold is the maximum jitter value at which the image is approximately considered to be in a static state.

When the jitter factor is lower than the target jitter threshold, the target video frame corresponding to the jitter factor can be approximately considered to be still.

The target image is an image of the moving pedestrian and the shadow of the pedestrian in the target moving square matrix after processing the original image.

The target image is used to characterize the real-time state of the target moving matrix.

In this embodiment, when the jitter factor of the target video frame is lower than the target jitter threshold, it is approximately considered that the target video frame is non-jitter, and the next step is performed on the target video frame.

In other embodiments, if the jitter factor of the target video frame is not lower than the target jitter threshold, it is considered that the jitter degree of the target video frame is relatively large, the target video frame corresponding to the jitter factor is skipped, and the next A frame of video frame.

Next, the generation method of the target image in this step will be described in detail.

In some embodiments, step 130 may further include:

Perform a frame difference operation on the target video frame and the target background model to generate a first image;

Perform binarization processing and noise reduction processing on the first image to generate a third image;

The foreground image in the target video frame is extracted based on the third image, and the target image of the target motion square matrix is generated based on the foreground image.

In this embodiment, the foreground image includes moving pedestrians in the target movement square and projections of the moving pedestrians in the target movement square on the ground.

In the actual execution process, for the target video frame without jitter, the coarse-grained detection of the foreground area can be performed first.

For example, as shown in FIG. 4 , a frame difference operation is performed between the target video frame and the target background model to generate the first image.

The target background model is a model established in advance, and the steps for establishing the target background model will be described in subsequent embodiments, which will not be described here.

Then, the result after the frame difference is first subjected to binarization processing, and then noise reduction processing is performed to eliminate the noise of the first image and the influence of the intrusion interference of small moving objects, so as to generate a third image, as shown in FIG. The three images are used to characterize the contour features of the moving square in the target video frame;

Among them, the noise reduction process can include basic morphological operations of erosion and expansion.

Then, a foreground image in the target video frame within the contour corresponding to the third image is extracted based on the third image, and a target image is generated based on the foreground area, as shown in FIG. 8 .

The foreground area at this time will contain pedestrians and their shadows in the square matrix, as shown in Figure 6.

According to the method for detecting the formation of a large-scale moving square array provided by the embodiment of the present invention, the moving target detection is performed by combining the background modeling method and the frame difference method, thereby realizing the improvement in detail.

In some embodiments, generating the target image of the target motion square matrix based on the foreground image may further include:

Convert the foreground image to the target format to generate a fourth image;

The fourth image is subjected to shadow region segmentation and erosion expansion processing based on the target shadow threshold to generate a target image of the target movement square matrix.

In this embodiment, color space conversion and square matrix area fine detection can also be performed on the foreground image after the coarse-grained detection of the foreground area, so as to eliminate the shadow of the foreground area and make the foreground extraction more accurate.

Among them, the target shadow threshold can be user-defined.

For example, the third image is converted from the RGB color space to the HSV color space to generate the fourth image. Among them, the HSV color space includes color (H), saturation (S) and brightness (V).

By converting to the HSV color space, the square matrix area and the shadow area in the foreground area in the third image can be effectively distinguished based on the luminance value. As shown in Figure 7, the difference between the two in luminance value (V space) is obvious.

Using the distribution characteristics of the third image in the HSV color space, after the third image is converted from the RBG color space to the HSV color space to generate the fourth image, the threshold value is set for the image in the foreground area in the fourth image in the brightness V space The shadow area is segmented, and then the erosion and expansion operation is performed to obtain the target image after the shadow interference is eliminated, as shown in Figure 8.

In this embodiment, combined with methods such as distortion correction and corrosion expansion, the influence of disturbances such as jitter, illumination, and shadow under natural outdoor conditions is overcome, and the real-time detection of moving square formations based on video frames is finally realized.

In some embodiments, the method may further include: updating the target shadow threshold every third target time period.

For example, the value distribution of the foreground area (square matrix and shadow area) in the brightness space V can also be used, and the value of the pixel point in the shadow area is determined by prior knowledge in the range of 50-100 in the V space, then take V In the interval of 10-200 of the spatial value distribution, the least squares method is used to fit the curve, and the minimum value of the trough is taken as the target shadow threshold Ψ for dividing the shadow area, and the target shadow threshold Ψ is also updated every third target time period.

Set the image pixel values corresponding to the brightness space V lower than the target shadow threshold Ψ to 0, and then perform the basic morphological operations of erosion and expansion in turn to eliminate the influence of image noise, and then extract a finer foreground area. , and the target image.

According to the method for detecting the formation of a large-scale moving square array provided by the embodiment of the present invention, the target shadow threshold is updated in real time through the "adaptive" target shadow threshold update method, and the target shadow can be updated in time based on the light intensity at different times. The threshold is adjusted to the optimal threshold, which avoids the situation that the fixed threshold cannot maintain the best effect when the natural light intensity changes, and also eliminates the problem that the shadow of the image in the outdoor natural environment is easily affected by the change of light intensity. It can adapt to the detection of moving square formations under natural outdoor conditions for a long time, which helps to improve the segmentation effect of the shadow area, thereby improving the imaging effect of the target image, thereby improving the detection effect.

In this step, by performing multiple processing on the target video frame, noise reduction can be effectively performed, and shadow interference can be eliminated, thereby obtaining a clear target image, which helps to improve the accuracy of the detection result.

As shown in FIG. 2, in some embodiments, in step 130, after determining that the jitter factor is lower than the target jitter threshold, the method may further include:

Obtain the resolution corresponding to the target video frame;

A target image of the target motion square matrix is generated based on the target video frame when the resolution is not lower than the resolution threshold.

When the resolution is lower than the resolution threshold, the jitter factor of the next target video frame is detected.

In this embodiment, the sharpness threshold is the minimum value of the sharpness of the image in a state approximately considered sharp.

The sharpness threshold can be customized based on the user.

In the actual execution process, the Laplacian gradient function can be used to evaluate the clarity of the target video frame.

Specifically, the formula can be used:

Determine, where M and N represent the width and height of the target video frame, respectively, T is the given edge detection threshold, and z(x,y) is the convolution of the Laplacian operator at the pixel point (x,y).

The Laplacian operator matrix is defined as the following formula:

It can be understood that, for the target video frame that has been converted to a grayscale image, the clearer the image of the target video frame input to the convolution is, the larger the Laplacian gradient function value is.

After obtaining the definition Ω _t corresponding to the target video frame, compare the definition with the definition threshold Ω, in the case of Ω _t ≥Ω, it is determined that the target video frame is clear, then based on the target video frame , and generate the target image of the target motion square matrix.

In other embodiments, in the case of Ω _t <Ω, it is determined that the target video frame is unclear, the target video frame is skipped, and the jitter factor of the next video frame is detected.

In this embodiment, by detecting the clarity of the target video frame, on the basis that the jitter factor is lower than the target jitter threshold, the target video frame that does not meet the definition requirement is skipped, so as to avoid camera shake and overexposure. In other cases, the detection of the square array is affected, so as to improve the clarity of the imaging and improve the accuracy of the detection results.

According to the method for detecting the formation of a large-scale moving square matrix provided by the embodiment of the present invention, based on the target video frame in the target video and the diagonal grayscale projection standard array, a jitter factor is generated, and the target video frame is screened based on the jitter factor. , in order to retain the target video frame with low jitter amplitude, and process the target video frame with low jitter amplitude to generate the target image of the target motion square matrix, which overcomes the defect that other grayscale projection methods are easily affected by local motion, and realizes the It can effectively detect the jitter of the target video frame in the moving state, significantly improve the rapidity and real-time performance of the video frame jitter detection, and improve the imaging effect of the target image, thereby helping to improve the accuracy and precision of the detection result.

The steps of generating the target background model will be described below through specific embodiments.

In some embodiments, prior to step 110, the method further includes:

Obtain multiple frames of environmental background images within the first target time period under the target environment;

Generate a target background model based on multiple frames of environmental background images;

Based on the target background model, a standard array of diagonal grayscale projections corresponding to the target background model is generated.

In this embodiment, the first target time period may be user-defined.

The ambient background image is the same background image as the target video.

The background image of the environment can be acquired by an image sensor, such as by a panoramic camera.

The number of ambient background images can be customized based on the user, such as set to 1000 frames.

After the environmental background image is collected, distortion correction may be performed on the environmental background image preferentially, and the specific implementation method is the same as that in the above-mentioned embodiment, and details are not described here.

For example, in the actual execution process, a Gaussian background model can be used to model the continuous first 1000 frames of environment background images to generate a target background model.

After generating the target background model, you can save the target background model in the local server or cloud server, and call it when needed.

The specific implementation manner of generating the diagonal grayscale projection standard array corresponding to the target background model based on the target background model will be described below.

First, convert each frame of the environmental background image into a grayscale image, and take four square areas of the same size with a side length of d from the upper left, lower left, upper right and lower right corners of the background model image as the target area to avoid In the future, the larger moving object in the middle of the screen will affect the shake detection.

It should be noted that, the target area in this embodiment needs to be consistent with the position and quantity of the target area in the target video frame extracted in the above embodiment.

Histogram equalization processing is performed on the four target regions respectively.

Specifically, the formula can be used:

Perform histogram equalization processing, where f is a monotonic nonlinear mapping, G _I represents the gray value of the pixels in the environmental background image I, L represents the gray level, and I _n is the number of pixels in the environmental background image. , H _I represents the gray histogram distribution of the environmental background image I, and u represents the gray value.

L usually takes 256.

After the histogram equalization process is performed, then grayscale projection is performed on the four regions after the histogram equalization, so as to generate a grayscale projection array standard value.

As shown in FIG. 3 , the standard value of the grayscale projection array is used to represent the comprehensive grayscale information in the horizontal and vertical directions of the four corner areas of the current environmental background image.

In some embodiments, the formula can be used:

表示环境背景图像中像素点(x,y)的灰度值，d表示每一个目标区域的边长，I_n是环境背景图像中像素点的个数。Calculate the gray mean value of the environmental background image, where I_mean represents the gray value of the pixel in the environmental background image,

Represents the gray value of the pixel ( _x , y) in the environmental background image, d represents the side length of each target area, and In is the number of pixels in the environmental background image.

After obtaining the gray mean value of the environmental background image, based on the gray mean value, the formula is:

Then the standard value of the diagonal grayscale projection array corresponding to each target area can be calculated;

表示环境背景图像中像素点(i,j)的灰度值。Among them, Gr_b1 represents the standard value of the k-th position of the diagonal grayscale projection array in the upper left corner of the environmental background image, d represents the side length of the target area,

Represents the gray value of the pixel point (i, j) in the background image of the environment.

It can be understood that Gr_b2 represents the standard value of the kth position of the diagonal grayscale projection array in the upper right corner of the environmental background image, and Gr_b3 represents the standard value of the kth position of the diagonal grayscale projection array in the lower right corner of the environmental background image. , Gr_b4 represents the standard value of the kth position of the diagonal grayscale projection array in the lower left corner of the ambient background image.

After the diagonal grayscale projection standard array is obtained, the diagonal grayscale projection standard array can be saved for subsequent calculation in the actual detection process.

In some embodiments, the target background model may be updated every second target time period t1 to avoid changes in information such as background brightness caused by the outdoor environment being easily affected by sunlight.

In this embodiment, by combining methods such as distortion correction and histogram equalization, it is beneficial to overcome the influence of disturbances such as jitter, illumination, and shadow under natural outdoor conditions.

According to the object detection method of the motion policy provided by the embodiment of the present invention, a diagonal grayscale projection standard array is generated based on the multi-frame environmental background images in the first target time period, which is used for the jitter detection of subsequent target video frames, with high flexibility and high flexibility. High detection accuracy.

The following describes the formation detection device for a large-scale event moving phalanx provided by the present invention. The formation detection device for a large-scale event moving phalanx described below and the above-described large-scale event movement phalanx formation detection method can be used. refer to each other.

As shown in FIG. 9 , the apparatus for detecting the formation of the large-scale moving square matrix includes: a first acquiring module 910 , a first generating module 920 and a second generating module 930 .

The first acquisition module 910 is used to acquire the target video of the target motion square matrix under the target environment;

The first generation module 920 is used to generate the corresponding jitter factor of the target video frame based on the target video frame and the diagonal grayscale projection standard array in the target video;

The second generating module 930 is configured to generate a target image of the target motion square matrix based on the target video frame when the jitter factor is lower than the target jitter threshold.

According to the object detection apparatus for a motion policy provided by an embodiment of the present invention, a jitter factor is generated based on a target video frame in a target video and a standard array of diagonal grayscale projections, and the target video frame is screened based on the jitter factor to preserve the jitter amplitude lower target video frame, and process the target video frame with lower jitter amplitude to generate target image of target moving square matrix, which effectively improves the clarity of imaging and helps to improve detection results.

In some embodiments, the first generation module 920 is used to:

Based on the target area of the target video frame, a first diagonal grayscale projection array of the target area is generated;

A dithering factor is generated based on the first diagonal grayscale projection array and the diagonal grayscale projection criterion array.

In some embodiments, the first generation module 920 is used to: apply the formula:

Generate a dither factor, where θ is the dither factor, Gr_bi(m) is the standard value of the diagonal grayscale projection at the mth position in the standard array of diagonal grayscale projections, and Gr_ci(m) is the first diagonal grayscale projection array The first diagonal grayscale projection value of the mth position in , and d is the side length of the target area.

In some embodiments, the second generation module 930 is used to:

Perform a frame difference operation on the target video frame and the target background model to generate a first image;

Perform binarization processing and noise reduction processing on the first image to generate a third image;

The foreground image in the target video frame is extracted based on the third image, and the target image of the target motion square matrix is generated based on the foreground image.

In some embodiments, the second generation module 930 is used to:

Convert the foreground image to the target format to generate a fourth image;

The fourth image is subjected to shadow region segmentation and erosion expansion processing based on the target shadow threshold to generate a target image of the target movement square matrix.

In some embodiments, the apparatus further includes:

The second acquisition module is used to acquire multiple frames of environmental background images in the first target time period under the target environment before acquiring the target video of the target motion square matrix in the target environment;

The third generation module is used to generate the target background model based on the multi-frame environment background images;

The fourth generation module is used for generating a standard array of diagonal grayscale projections corresponding to the target background model based on the target background model.

FIG. 10 illustrates a schematic diagram of the physical structure of an electronic device. As shown in FIG. 10 , the electronic device may include: a processor (processor) 1010, a communication interface (Communications Interface) 1020, a memory (memory) 1030, and a communication bus 1040, The processor 1010 , the communication interface 1020 , and the memory 1030 communicate with each other through the communication bus 1040 . The processor 1010 can invoke the logic instructions in the memory 1030 to execute a method for detecting the formation of a large-scale moving square, the method comprising: acquiring a target video of the target moving square in a target environment; A target video frame and a standard array of diagonal grayscale projections are used to generate a jitter factor corresponding to the target video frame; when the jitter factor is lower than a target jitter threshold, the target motion equation is generated based on the target video frame. array of target images.

In addition, the above-mentioned logic instructions in the memory 1030 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention can be embodied in the form of a software product in essence, or the part that contributes to the prior art or the part of the technical solution. The computer software product is stored in a storage medium, including Several instructions are used to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, Read-Only Memory (ROM, Read-Only Memory), Random Access Memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program codes .

In another aspect, the present invention also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, when the program instructions are executed by a computer When executing, the computer can execute the formation detection method of the large-scale activity moving square provided by the above methods, the method includes: acquiring the target video of the target moving square in the target environment; based on the target video in the target video frame and diagonal grayscale projection standard array to generate the jitter factor corresponding to the target video frame; when the jitter factor is lower than the target jitter threshold, based on the target video frame, generate the target motion square matrix. target image.

In yet another aspect, the present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored, the computer program being implemented by a processor to execute the formation of the large-scale activity moving square arrays provided above A detection method, the method comprising: acquiring a target video of a target motion square matrix under a target environment; generating a jitter factor corresponding to the target video frame based on the target video frame and the diagonal grayscale projection standard array in the target video; In the case where the jitter factor is lower than a target jitter threshold, a target image of the target motion square matrix is generated based on the target video frame.

The device embodiments described above are only illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in One place, or it can be distributed over multiple network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution in this embodiment. Those of ordinary skill in the art can understand and implement it without creative effort.

From the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and certainly can also be implemented by hardware. Based on this understanding, the above-mentioned technical solutions can be embodied in the form of software products in essence or the parts that make contributions to the prior art, and the computer software products can be stored in computer-readable storage media, such as ROM/RAM, magnetic A disc, an optical disc, etc., includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform the methods described in various embodiments or some parts of the embodiments.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, but not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that it can still be The technical solutions described in the foregoing embodiments are modified, or some technical features thereof are equivalently replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims (10)

1. the formation detection method of the movement square matrix of a large-scale activity, is characterized in that, comprises: Obtain the target video of the target moving square matrix in the target environment; generating a jitter factor corresponding to the target video frame based on the target video frame and the diagonal grayscale projection standard array in the target video; In the case where the jitter factor is lower than a target jitter threshold, a target image of the target motion square matrix is generated based on the target video frame. 2. the formation detection method of the motion square matrix of large-scale activities according to claim 1, is characterized in that, described based on the target video frame in described target video and diagonal grayscale projection standard array, generate described target The jitter factor corresponding to the video frame, including: Based on the target area of the target video frame, a first diagonal grayscale projection array of the target area is generated, and the target area does not include the feature of the motion square; The dithering factor is generated based on the first diagonal grayscale projection array and the diagonal grayscale projection criterion array. 3. The formation detection method of the moving square matrix of large-scale activities according to claim 2, is characterized in that, described based on described first diagonal grayscale projection array and described diagonal grayscale projection standard array, generate The jitter factor includes: Apply the formula:

Generate the dither factor, where θ is the dither factor, the Gr_bi(m) is the diagonal grayscale projection standard value of the mth position in the diagonal grayscale projection standard array, and the Gr_ci(m ) is the first diagonal grayscale projection value of the mth position in the first diagonal grayscale projection array, and d is the side length of the target area. 4. the formation detection method of the moving square matrix of large-scale activities according to claim 1, is characterized in that, described based on described target video frame, the target image of described target moving square matrix is generated, comprising: performing frame difference operation on the target video frame and the target background model to generate a first image; performing binarization processing and noise reduction processing on the first image to generate a third image; A foreground image in the target video frame is extracted based on the third image, and a target image of the target motion square matrix is generated based on the foreground image. 5. The formation detection method of the moving square matrix of large-scale activities according to claim 4, wherein the generating the target image of the target moving square matrix based on the foreground image comprises: converting the foreground image into a target format to generate a fourth image; Performing shadow region segmentation processing and erosion expansion processing based on the target shadow threshold on the fourth image to generate a target image of the target movement square matrix. 6. The formation detection method of the moving square matrix of large-scale activities according to any one of claims 1-5, wherein, before the acquisition of the target video of the target moving square matrix under the target environment, the method Also includes: acquiring multiple frames of environmental background images within the first target time period under the target environment; generating a target background model based on the multi-frame environment background images; Based on the target background model, a standard array of diagonal grayscale projections corresponding to the target background model is generated. 7. A formation detection device for a large-scale activity moving square, characterized in that it comprises: The first acquisition module is used to acquire the target video of the target motion square matrix under the target environment; a first generating module, configured to generate a jitter factor corresponding to the target video frame based on the target video frame and the diagonal grayscale projection standard array in the target video; A second generating module, configured to generate a target image of the target motion square matrix based on the target video frame when the jitter factor is lower than the target jitter threshold. 8. An electronic device comprising a memory, a processor and a computer program stored on the memory and running on the processor, wherein the processor implements the program as claimed in claim 1 when executing the program The formation detection method of the moving square in any one of to 6. 9. A non-transitory computer-readable storage medium on which a computer program is stored, characterized in that, when the computer program is executed by a processor, the movement method of the large-scale activity according to any one of claims 1 to 6 is realized. Array formation detection method. 10. A computer program product, comprising a computer program, characterized in that, when the computer program is executed by a processor, the method for detecting the formation of a large-scale moving square matrix according to any one of claims 1 to 6 is implemented.

Images