TWI602156B - Moving foreground detection system and method thereof - Google Patents

Description

System and method for detecting moving foreground

The present invention relates to a system for detecting a moving foreground and a method thereof, and more particularly to a system and method for detecting a moving foreground based on a hysteresis threshold and a Gaussian mixture mode.

In recent years, due to the emphasis on home security, video image detection systems have become a hot research field, because image detection systems can be widely used to protect people's personal safety, such as building monitors in the community. Record accidents or pay attention to suspicious persons.

Traditional image detection systems, such as surveillance systems, must constantly watch the surveillance screens in a human-powered manner so as not to miss important information. Therefore, there are quite a few studies focusing on how to use the computer system to detect and output images of moving foregrounds. Save labor costs. In the prior art, the system for detecting the foreground of the movement must first establish a background model to separate the image of the moving foreground and the fixed background. However, when establishing the background model, many challenges, such as noise and movement of light changes, are often encountered. The interference of the foreground shadows or the slow moving motion of the moving motion becomes part of the fixed background.

In view of the above, the inventors of the present invention contemplate and design a system and method for detecting the foreground of the mobile, and improve the lack of the prior art to present a more Fine detection of images.

The embodiment of the present invention is directed to a system for detecting a moving foreground, which can improve the noise problem caused by the light change in the system for detecting the moving foreground in the past and effectively eliminate the shadow of the moving foreground, and can also move through the foreground. The historical position corrects the misjudgment caused by the slow movement of the moving foreground, thereby providing the user with an accurate and complete system for detecting the foreground of the movement.

The invention provides a system for detecting a moving foreground, which comprises a camera module, a processing module and a storage module. The camera module captures an original image from a scene and transmits the original image to a processing module. The processing module extracts a texture information from the original image in a Gaussian Mixture Model. Hysteresis thresholding is used to generate a Dominated Color (DC) binary image and an Adjunct Color (AC) binary image according to the original image, and the hysteresis threshold is applied to the Gaussian mixture model according to the texture information. A primary image (Dominated Texture, DT) binary image and an auxiliary texture (AT) binary image; the primary texture binary image is used to compare the primary color binary image to generate a shadow for removing the moving foreground a first binary image; the first binary image is used to compare the primary color binary image to extract a frame of the moving foreground and generate a second binary image; and then, the second binary image is used to compare the auxiliary color binary image And the auxiliary texture binary image repairing the lost fine cell block in the moving foreground frame of the second binary image to generate a third binary image; finally using the third binary image comparison Zhang two yuan has a plurality of position history image to the moving foreground movement locus of the foreground image of the corrected third two yuan, two yuan generates and outputs a fourth image to the storage module.

Preferably, the camera module, the processing module and the storage module can be integrated into the same package In the middle, or the camera module is installed at the remote end, and the processing module and the storage module are installed on the servo end, and the camera module and the servo end are connected through the network.

The hysteresis threshold may include a first threshold value and a second threshold value and the first threshold value is greater than the second threshold value; the processing unit separates the original image into a first moving foreground image in the Gaussian mixture model by using the first threshold value and A first background image is used to generate a primary color binary image; the second threshold is used to separate the original image into a second moving foreground image and a second background image auxiliary color binary image in the Gaussian mixture model.

The hysteresis threshold may include a first threshold value and a second threshold value and the first threshold value is greater than the second threshold value; the processing unit uses the first threshold value to separate the texture information into a third moving foreground image in the Gaussian mixture model and a primary texture binary image of the third background image; and separating the texture information into a fourth moving foreground image and a fourth background image auxiliary texture binary image in the Gaussian mixture model by using the second threshold value.

Furthermore, the present invention further provides a method for detecting a moving foreground, comprising: capturing an original image from a scene using a camera module and extracting a texture information from the original image using a processing module in a Gaussian mixture model Applying a hysteresis threshold to generate a primary color binary image and an auxiliary color binary image according to the original image, and applying a hysteresis threshold in the Gaussian mixture model to generate a primary texture binary image and an auxiliary texture according to the binary image of the texture information. Binary image; comparing the primary color binary image with the primary color binary image to generate a first binary image that removes the shadow of the moving foreground; and then extracting the moving foreground frame in the first binary image and generating a first a binary image; the second binary image is used to compare the auxiliary color binary image and the auxiliary texture binary image to repair the missing block in the moving foreground frame in the second binary image to generate a third binary image; Using the third binary image to compare a plurality of binary images containing the historical position of the moving foreground to correct the third binary image, generating and outputting a fourth binary image to Storage module.

Preferably, the hysteresis threshold includes a first threshold value and a second threshold value, and the method for generating a dominant color binary image and an auxiliary color binary image by applying a hysteresis threshold to the original image in the Gaussian mixture model comprises using the first threshold value The Gaussian mixture model separates the original image into a first moving foreground image and a primary background binary image of the first background image; and uses the second threshold value to separate the original image into a second moving foreground image in the Gaussian mixture model. And an auxiliary color binary image of the second background image.

Preferably, the hysteresis threshold includes a first threshold value and a second threshold value, and the method for applying the hysteresis threshold to the texture information to generate the main texture binary image and the auxiliary texture binary image in the Gaussian mixture model comprises using the first threshold value Separating the texture information into a third moving foreground image and a third texture image in the Gaussian mixture model; and separating the texture information into a fourth in the Gaussian mixture model by using the second threshold value An auxiliary texture binary image that moves the foreground image and a fourth background image.

As described above, the system and method for detecting a moving foreground of the present invention is based on a Gaussian mixture model in which a hysteresis threshold is applied to image processing based on original image and texture information to generate a binary image of a separated moving foreground and background. The binary image contains a complete moving foreground image, and can eliminate the noise caused by the shadow of the moving foreground and the change of the light, and even detect the moving direction of the moving foreground to solve the problem that the moving foreground is misjudged into the background, achieving a perfect and The ability to accurately detect the foreground of the move.

1‧‧‧System for detecting mobile prospects

11‧‧‧ camera module

111‧‧‧ original image

111-1‧‧‧Main color binary image

111-2‧‧‧Auxiliary color binary image

12‧‧‧Processing module

121‧‧‧hysteresis threshold

122‧‧‧ Gaussian mixture model

112‧‧‧Texture Information

112-1‧‧‧Main texture binary image

112-2‧‧‧Auxiliary Texture Binary Image

124‧‧‧ binary image

13‧‧‧ Storage Module

22‧‧‧ camera

23‧‧‧Computer host

I43‧‧‧ first binary image

I51‧‧‧ second binary image

I63‧‧‧ third binary image

I71‧‧‧Multiple binary images of mobile history

I72‧‧‧ fourth binary image

The above and other features and advantages of the present invention will become more apparent from the detailed description of the exemplary embodiments of the invention in which a block diagram of an example; 2 is a schematic diagram of a preferred embodiment of a system for detecting a moving foreground according to the present invention; FIG. 3 is a flow chart of a method for detecting a moving foreground according to the present invention; FIG. 4 is a flowchart according to the present invention; A schematic diagram of a first embodiment of a method for detecting a moving foreground of the invention; a fifth embodiment of the method for detecting a moving foreground according to the present invention; and a sixth embodiment for detecting a moving foreground according to the present invention A schematic diagram of a third embodiment of the method; and FIG. 7 is a schematic diagram of a fourth embodiment of a method for detecting a moving foreground according to the present invention.

As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items. When the phrase "at least one of" is preceded by a list of elements, the entire list of elements is modified instead of the individual elements in the list.

1 is a block diagram showing an embodiment of a system for detecting a moving foreground according to the present invention. Referring to FIG. 1 , the mobile foreground system 1 includes a camera module 11 , a processing module 12 , and a storage module 13 . After the camera module 11 images the scene to form an original image 111, the processing module 12 first extracts a texture information 112 from the original image 111. The texture information 112 is firstly cut into the original image 111. A block of non-overlapping blocks, each block containing n*n pixels, using equation (1) to convert a pixel bij located in (i, j) of the block into a binary value, where m is the original image The average pixel value, x _ij is one of the (i, j) pixels in the block, and Threshold is the user-defined preset threshold, where 1≦i≦n, 1≦j≦n.

Then, the processing module 12 applies a first threshold value and a second threshold value of the hysteresis threshold 121 in the Gaussian mixture model 122 to generate a dominant color (DC) binary image 111-1 and an auxiliary color according to the original image 111. (Adjunct Color, AC) binary image 111-2, and applying a first threshold value and a second threshold value of the hysteresis threshold 121 in the Gaussian mixture model 122 to generate a main texture binary image 112-1 and a texture according to the texture information 112. The auxiliary texture binary image 112-2 uses the main texture binary image 112-1 to compare the main color binary image 111-1 to generate a first binary image that removes the shadow of the moving foreground.

The method for generating the auxiliary texture binary image and the main texture binary image is as follows: each pixel of the texture information is pre-set with a binary bitmap of K different weight values w, {BM1, BM2,... , BMK}, where w is a real number between 0 and 1 and each binary bitmap size is p*p. When a binary value of one pixel in the texture information is generated, the processing unit generates a similar distance value Dist(BMnew,BMm), l≦m≦K, using the formula (2), and the hysteresis threshold includes the first threshold value and the second threshold value. And the first threshold is greater than the second threshold. If the similar distance value is less than the first threshold, the pixel is determined to be the background, and the opposite is determined to be the moving foreground, where BM _new is formed by the texture information centered on the pixel. p*p binary bitmap. The binary image determined by the first threshold and the second threshold is the primary texture binary image and the auxiliary texture binary image.

And then using the first binary image to compare the primary color binary image 111-1 to extract the moving foreground frame in the first binary image and generate a second binary image; using the second binary The images respectively compare the auxiliary color binary image 111-2 and the auxiliary texture binary image 112-2 to repair the missing fine cell block in the moving foreground frame in the second binary image and generate a third binary image; finally, The third binary image is compared with a movement history to correct the moving trajectory of the moving foreground in the third binary image, and a binary image 124 is generated and outputted to the storage module 13.

The camera module 11, the processing module 12, and the storage module 13 can be integrated into the same device, or the camera module 11 is installed at the remote end, and the processing module 12 and the storage module 13 are installed at the server. The camera module 11 and the servo end are connected through a network.

Figure 2 is a schematic diagram showing a preferred embodiment of the system for detecting a moving foreground of the present invention. Referring to FIG. 2, after the camera 22 captures the original image 111, the host computer 23 processes the original image 111 into a moving foreground and a background separated binary image 124 according to the Gaussian mixture model and the hysteresis threshold.

Figure 3 is a flow chart showing the steps of the method for detecting a moving foreground of the present invention. Referring to FIG. 3, in step S30, a camera module performs image capture on a scene and transmits the original image to a processing module. In step S31, the processing module captures the texture information of the original image. In step S32, the processing module applies a Gaussian mixture mode and a hysteresis threshold to generate a primary color binary image and an auxiliary color binary image according to the original image. Here, how to generate the primary color binary image and the auxiliary color binary image has been described in the foregoing, and therefore will not be described herein.

In step S33, the processing module applies a Gaussian mixture model and a hysteresis threshold to generate a main texture binary image and an auxiliary texture binary image according to the texture information. In step S34, the primary color binary image is compared with the primary color binary image to generate a first binary image that removes the moving foreground shadow. In step S35, the primary binary image is compared with the first binary image to extract a frame containing the moving foreground in the first binary image, and a second binary image is generated. In step S36, the second binary image is used to compare the auxiliary color binary image and the auxiliary texture binary image to repair the missing fine block in the frame of the moving foreground in the second binary image and generate a third binary image. image. In step S37, the third binary image is used to compare the movement history to correct the moving foreground moving track of the third binary image, and generate and output a fourth binary image to a storage module.

The primary threshold of the hysteresis threshold is generated by the Gaussian mixture model, and the primary color binary image is generated. The second threshold of the hysteresis threshold is substituted by the Gaussian mixture model to generate the auxiliary color binary image.

Figure 4 is a view showing a first embodiment of the method of detecting a moving foreground of the present invention. Referring to FIG. 4, the primary texture binary image 112-1 is used to compare the primary color binary image 111-1 to generate a first binary image i43 that eliminates moving foreground shadows and background noise, wherein the primary texture binary Each of the pixels of the image 112-1 converts the eight neighboring points around the corresponding pixel in the primary color binary image 111-1, if it does not include a pre-moving attraction, converts the main texture binary image 112-1 The pixel is a background point, and conversely, the pixel is converted into a pre-moving attraction; the first binary image of the moving foreground shadow and the background noise is generated by comparing all the pixels in the main color binary image 111-1. I43.

Figure 5 is a diagram showing a second embodiment of the method of detecting a moving foreground of the present invention. Referring to FIG. 5, the first binary image i43 is used to compare the primary color binary image 111-1 to extract the second binary image i51 of the moving foreground frame in the first binary image i43. The first binary image i43 is converted into the first binary image i43 if the eight adjacent points around the corresponding pixel in the primary color binary image 111-1 do not include a pre-moving attraction. The pixel is a background point. Otherwise, the pixel is converted into a pre-moving point of view; after comparing all the pixels in the main color binary image 111-1, the height and width of the moving foreground are used to extract the foreground with the moving foreground. The frame forms a second binary image i51.

Figure 6 is a view showing a third embodiment of the method of detecting a moving foreground of the present invention. Referring to FIG. 6, the auxiliary binary image 111-2 and the auxiliary texture binary image 112-2 are compared in the second binary image i51 to repair the frame of the moving foreground in the second binary image i51. The missing block within the block produces a third binary image i63. Wherein, in the second binary image i51, if the eight neighboring points around the corresponding pixel points in the auxiliary color binary image 112-2 and the auxiliary texture binary image 112-2 do not include a pre-moving attraction, Then converting the pixel in the second binary image i51 to a background point; otherwise, converting the pixel to a pre-moving attraction; comparing the main color auxiliary color binary image 111-2 and the auxiliary texture binary image 112- After all the pixels in 2, the microblocks missing in the moving foreground of the second binary image i51 can be repaired, and the result is output as the third binary image i63.

Fig. 7 is a view showing a fourth embodiment of the method of detecting a moving foreground of the present invention. Referring to FIG. 7, the movement history is a plurality of binary images including the historical position of the moving foreground, and the third binary image i63 is used to compare the moving history i71 to correct the moving trajectory of the moving foreground in the third binary image. And generating and outputting a fourth binary image i72. If the third binary image i63 has a shooting time t, and there are two binary images containing the historical position of the moving foreground, and the shooting time of the two binary images is t-1 and t-2, respectively, searching for the third binary image One of the background points in i63 is not less than a front sight around the corresponding pixel in the binary image at time t-1 and simultaneously searches for the corresponding pixel in the binary image of the background point at time t-2. Is there no less than b front spots around, where a>=b, if yes, set this background point as a front spot, and after changing all the background points of the third binary image i63, output the result as the fourth two. Meta image i72.

The present invention has been particularly shown and described with reference to the exemplary embodiments thereof, and it is understood by those of ordinary skill in the art Various changes in form and detail can be made in the context of the category.

1‧‧‧System for detecting mobile prospects

11‧‧‧ camera module

111‧‧‧ original image

111-1‧‧‧Main color binary image

111-2‧‧‧Auxiliary color binary image

12‧‧‧Processing module

121‧‧‧hysteresis threshold

122‧‧‧ Gaussian mixture model

112‧‧‧Texture Information

112-1‧‧‧Main texture binary image

112-2‧‧‧Auxiliary Texture Binary Image

124‧‧‧ binary image

13‧‧‧ Storage Module

Claims (2)

A method for detecting a moving foreground, comprising: a. capturing a raw image from a scene using a camera module; and b. using a processing module to extract a texture information from the original image, A Gaussian Mixture Model uses a Hysteresis thresholding to generate a Dominated Color (DC) binary image and an Adjunct Color (AC) binary image based on the original image. Applying the hysteresis threshold to the Gaussian mixture model to generate a Dominated Texture (DT) binary image and an Adjunct Texture (AT) binary image according to the binary image of the texture information; The meta-image is compared to the primary color binary image to generate a first binary image that removes the shadow of the moving foreground; extracting the moving foreground frame in the first binary image and generating a second binary image; The second binary image respectively compares the auxiliary color binary image and the auxiliary texture binary image to repair the lost block in the moving foreground frame in the second binary image, and generates a third binary Using the third binary image to compare a motion history (Motion history) to correct the third binary image, generating and outputting a fourth binary image to a storage module; wherein the primary texture binary image Each pixel in the main color binary image surrounds a plurality of neighboring points, if the neighboring point does not include a pre-moving point of view, the pixel in the main texture binary image is converted into a background point, and vice versa. Converting the pixel to a pre-movement attraction; and each pixel in the first binary image, if a plurality of neighboring points around the corresponding pixel in the binary image of the main color does not include a pre-moving attraction, Converting the pixel in the first binary image to a background point; otherwise, converting the pixel to a pre-movement attraction; wherein, the third binary image capturing time is t, and the historical position of the moving foreground is included When the binary image has a plurality of images and the shooting time of the plurality of binary images is t-1 and t-2, respectively, searching for a background point of the third binary image corresponds to the binary image at the t-1th time Is there no less than a front spot around the pixel, and at the same time searching for whether the background point is not less than b front spots around the corresponding pixel in the binary image of time t-2, where a≧b, if And setting the background point as a front point, and outputting the result to the fourth binary image after changing all the background points of the third binary image; wherein the hysteresis threshold includes a first threshold and a first a threshold value, the method for generating the primary color binary image and the secondary color binary image for the original image by using the hysteresis threshold in the Gaussian mixture model comprises: (b1) using the first threshold value in the Gaussian mixture model Separating the original image into a first Moving the foreground image and the primary color binary image of the first background image; and (b2) separating the original image into a second moving foreground image and a second background in the Gaussian mixture model by using the second threshold value The auxiliary color binary image of the image; wherein the method for generating the main texture binary image and the auxiliary texture binary image by applying the hysteresis threshold to the texture information in the Gaussian mixture model comprises: (b3) separating the texture information into a third moving foreground image and a third background image of the main texture binary image in the Gaussian mixture model by using the first threshold value; and (b4) utilizing the second The threshold value is separated into the auxiliary texture binary image of the fourth moving foreground image and the fourth background image in the Gaussian mixture model; wherein the movement history utilizes a plurality of historical positions including the moving foreground The binary image is used to correct the moving trajectory of the moving foreground of the third binary image. A system for detecting a moving foreground, comprising: a camera module for capturing an original image from a scene to form an original image; a storage module; and a processing module for capturing the original image a texture information, the processing module applies a hysteresis threshold to a Gaussian mixture model to generate a primary color binary image and an auxiliary color binary image according to the original image, and the processing module applies the hysteresis to the Gaussian mixture model The threshold generates a primary texture binary image and a secondary texture binary image according to the texture information; the processing module uses the primary texture binary image to generate a shadow for removing the foreground color of the primary color binary image. a first binary image; the processing module extracts one of the first binary images to move the foreground frame and generates a second binary image; and the processing module compares the auxiliary color according to the second binary image The meta-image and the auxiliary texture binary image are used to repair the missing block in the moving foreground frame in the second binary image, and the processing module generates a third binary image; the processing module utilizes a Dynamic correction history The third binary image is used to generate a fourth binary image, the processing module outputs the fourth binary image to the storage module for storage; and a transmission module transmits the original image to the processing module Wherein, each pixel of the main texture binary image is converted into the main texture binary image if a plurality of neighboring points around the corresponding pixel point in the main color binary image do not include a pre-moving attraction point; The pixel is a background point, and conversely, the pixel is converted into a pre-movement attraction; and each pixel in the first binary image is in a plurality of neighboring points around the corresponding pixel in the main color binary image. When the pre-moving attraction is not included, the pixel in the first binary image is converted to a background point, and vice versa, the pixel is converted into a pre-moving attraction; wherein, the third binary image capturing time is t And if the binary image containing the historical position of the moving foreground has a plurality of images and the shooting time of the plurality of binary images is t-1 and t-2, respectively, searching for one of the third binary images is at the t -1 time binary image Whether there is no less than a front sight around the corresponding pixel, and at the same time, searching for whether the background point is not less than b front sights in the binary image of the binary image at time t-2, where a≧b If yes, the background point is set as a front point, and after the background point of the third binary image is changed, the result is output as the fourth binary image; wherein the hysteresis threshold includes a first threshold And a second threshold value, wherein the processing module separates the original image into a first moving foreground image and a first background image of the primary color binary image in the Gaussian mixture model by using the first threshold value And using the second threshold to separate the original image into a second moving foreground image in the Gaussian mixture model And the auxiliary color binary image of the second background image; and the processing module separates the texture information into a third moving foreground image and a third background in the Gaussian mixture model by using the first threshold value The main texture binary image of the image, and the texture information is separated into the auxiliary texture binary image of the fourth moving foreground image and the fourth background image by using the second threshold value in the Gaussian mixture model; wherein The movement history is that the processing module uses a plurality of binary images including a historical position of the moving foreground to correct a moving trajectory of the moving foreground of the third binary image.