CN101276415A - Device and method for realizing multi-resolution image acquisition with multiple fixed-focus cameras - Google Patents

Abstract

A device and method for realizing multi-resolution image acquisition by using multiple fixed-focus cameras, including five fixed-focus cameras with position constraints, wherein four cameras are on the same horizontal plane, and are located on four vertices of the same square, and The focal lengths of the 4 cameras are the same, set as f _s ; the other camera is located on one side of the above 4 cameras, the axis of the camera is perpendicular to the plane of the 4 cameras and passes through the center point O of the square, and the 5th camera The distance between one camera and the plane where the other four cameras are located is d, and the focal length of the camera is set as f _t , and the parameters and positions of the five fixed-focus cameras are initialized, and their positional relationship satisfies the following constraints: f _t = f _s + d. The invention realizes the function of non-destructive zoom shooting of multiple fixed-focus cameras, the details of scene images can be adjusted, it is flexible and controllable, and multi-resolution images can be quickly collected in larger visible scenes.

Description

Device and method for realizing multi-resolution image acquisition with multiple fixed-focus cameras

technical field

The invention belongs to the field of video processing, in particular to a device and method for acquiring scene images applied to computer vision.

Background technique

Multi-resolution image acquisition refers to the process of consistent acquisition of image data of the same scene at different resolution details. The emphasis is on artificially adjusting the image resolution of the shooting scene according to the user's needs for image details. Existing methods and devices use zoom cameras to achieve multi-resolution image acquisition.

A zoom camera refers to a camera whose lens focal length can be adjusted and changed, so that different visual and shooting effects can be obtained by changing the lens focal length, including single-camera optical zoom shooting and single-camera digital zoom shooting.

Single-camera optical zoom relies on the structure of the optical lens to achieve zooming. The camera lens moves to zoom in and out of the scene to be photographed. The larger the optical zoom factor, the farther the scene can be photographed. General optical zoom cameras are expensive, and after zooming, although the scene resolution is improved, the visible scene is reduced exponentially.

The principle of single-camera digital zoom is different from optical zoom. It uses "interpolation" processing method to enlarge some pixels on the original CCD image sensor, and uses interpolation algorithm to enlarge the picture to the whole picture. In fact, it is an electronic magnification of the picture. Although the captured scene is enlarged, the resolution of the scene has not been substantially improved. At the same time, the single-camera digital zoom method also has the disadvantage that the visible scene becomes smaller after zooming.

Some patents at home and abroad also propose devices and methods for multi-resolution image acquisition. Chinese patent CN1176351C proposes a dual acousto-optic lighting transmitter, CN2476040Y proposes a reflector group, CN1164962C proposes a total optical path changing device, and U.S. patent application 20060077475 proposes Using glass mirrors, circuits, lens rods and photosensitive sensor array combination devices, all try to solve multi-resolution image acquisition, but because the device is a single camera type, it does not solve the problem of automatic zooming out of the visible scene. US Patent Application No. 20050111754 proposes a method for multi-resolution image fusion, but does not provide constraints on image acquisition devices and control parameters.

In addition, according to the disadvantage that the visual scene becomes smaller due to the optical zoom of a single camera, the situation of shooting and collecting by multiple optical zoom cameras can be considered. Although it can solve the problem of the size of the visible scene, it has the disadvantage that the image resolution is not easy to control, and the cost of the device will be more expensive.

Generally speaking, the existing multi-resolution image acquisition devices and methods have the disadvantages of small visible scene or difficult acquisition control.

Contents of the invention

One of the purposes of the present invention is: to overcome the deficiencies in the prior art, to provide a multi-resolution image acquisition device that realizes a larger visual scene; the second purpose is to propose a control method for multi-resolution image acquisition based on the above-mentioned device .

The technical solution of the present invention: a device that adopts multiple fixed-focus cameras to realize multi-resolution image acquisition, including 5 fixed-focus cameras with position constraints, of which 4 cameras are on the same horizontal plane, located in the same square 4 vertex, and the focal lengths of the four cameras are the same, set it as f _s ; another camera is located on one side of the above four cameras, and the axis line of this camera is perpendicular to the plane of the four cameras and passes through the center point O of the square , and the distance from the fifth camera to the plane where the other four cameras are located is d, and the focal length of the camera is f _t , the parameters and positions of the five fixed-focus cameras are initialized, and their positional relationship satisfies the following constraints: f _t =f _s +d.

The working principle of the device of the present invention is: five fixed-focus cameras are divided into two groups of full-scene acquisition cameras and detail acquisition cameras. One full-scene acquisition camera with a fixed position ensures that the image information of the entire scene can be obtained, and the scene range will not change due to image resolution adjustment; four detail acquisition cameras collect images of four sub-square scene areas respectively, according to the imaging principle of the camera The geometric position relationship between the camera and the camera, the data of the detail capture camera is mapped to the image captured by the full scene capture camera, and a multi-resolution image capture device for a larger visible scene is realized. Let the image coordinates captured by the detail acquisition camera be (u ₁ , v ₁ ), (u ₂ , v ₂ ), (u ₃ , v ₃ ), (u ₄ , v ₄ ), and the image coordinates mapped to the whole scene become $({\mathrm{<figure-callout\; id="1"\; label="u"\; filenames="A20081010124100112.png"\; state="\{\{state\}\}">u</figure-callout>}}_1,{\mathrm{<figure-callout\; id="1"\; label="v"\; filenames="A20081010124100112.png"\; state="\{\{state\}\}">v</figure-callout>}}_1+\sqrt{2}m),(u_2+\sqrt{2}m,v_2+\sqrt{2}m),(u_3+\sqrt{2}m,v_3),(u_4,v_4),$ Where m is the vertical distance of the detail capture camera from the full scene capture camera.

A control method for realizing multi-resolution scene acquisition by using multiple fixed-focus cameras proposed by the present invention comprises the following steps:

(1) Automatically specify the visible target objects in the scene, or manually specify the visible target objects in the scene;

(2) Perform background elimination according to the scope of the target object to obtain the foreground of the target object;

(3) Determine the imaging display sub-area of the target object according to the foreground of the target object, and arrange the scene detail acquisition camera and the whole scene acquisition camera in the sub-area;

(4) Collect image data through cameras in the sub-area to generate a high-resolution scene overview; (5) specify the lossless resolution magnification of the target object in the high-resolution scene overview;

(6) Fusing the edge of the enlarged target object with the background difference;

(7) Judging whether the fused target object moves, if it moves, go to step (2); if it does not move, go to step (4), and repeatedly collect image information.

The principle of the method of the present invention is: use image processing methods such as background segmentation to determine the target object, change the data collected by the detail acquisition camera arranged in the sub-region according to the pixel displacement formula, complete the lossless resolution amplification of the target object, and apply the interpolation fusion algorithm to make The target object is fused with the scene image to realize multi-resolution scene acquisition control.

The beneficial effect of the present invention compared with prior art is:

(1) The device of the present invention is equipped with a full-scene capture camera and a detail capture camera to capture images of the same shooting scene, and the obtained target object information includes scene overview information and sub-scene detail information. Using the scene overview information can ensure that the visible scene range does not occur Changes, effectively avoiding the problem that the visible scene becomes smaller when the high-resolution image is obtained when a single camera is zoomed, and realize the "zoom" shooting function of multiple fixed-focus cameras in a larger visible scene. And there are two sets of image acquisition information for the same scene as mutual backup, and the imaging quality is more stable.

(2) The detail acquisition camera that the device of the present invention is provided with obtains the sub-scene information of the target object, can further describe the detail information of the scene, and is real acquisition data, overcomes the problem that the zoom method of digital zoom camera electronic amplification can affect the image The disadvantage of precision.

(3) Compared with other image acquisition methods, the method of the present invention does not need to manually adjust the focal length or aperture of the camera, only needs to manually or automatically select the target object in the first step of the inventive method, and set the lossless resolution in the fifth step The magnification can automatically complete the rapid acquisition of multi-resolution scene images. It has the advantages of simple manual operation steps, flexible selection of target objects, adjustable image details, and easy control.

Description of drawings

Fig. 1 is a schematic diagram of a multi-focus camera device of the present invention;

Fig. 2 is the geometric relation figure that 5 camera focal points form in the example of the present invention;

Fig. 3 is a schematic diagram of the relationship between camera imaging resolution and focal length of the present invention;

FIG. 4 is a flow chart of the multi-resolution image acquisition method of the present invention.

Detailed ways

As shown in Figure 1, the present invention is provided with 5 fixed-focus cameras, and the 4 cameras on the right are on the same horizontal plane, located on the 4 vertices of the same square, corresponding to the four points B, C, D, and E in the figure, And the focal lengths of the four cameras are the same, which is set to f _s ; another camera is located on the left of the above four cameras, corresponding to point A in the figure, and its axis line is perpendicular to the plane of the four cameras and passes through the center point of the square, corresponding to Point O in the middle, and the distance to this point is d, let the focal length of the camera be f _t . The parameters and positions of the five fixed-focus cameras are initialized, and their positional relationship satisfies the following constraints: f _t =f _s +d.

The working principle and process of the device of the present invention are as follows: five fixed-focus cameras are divided into two groups of full-scene acquisition cameras and detail acquisition cameras. One full-scene acquisition camera with a fixed position ensures that the image information of the entire scene can be obtained, and the scene range will not change due to image resolution adjustment; four detail acquisition cameras collect images of four sub-square scene areas respectively, according to the camera imaging principle According to the geometric position relationship between the camera and the camera, the data of the detail acquisition camera is mapped to the image captured by the full scene acquisition camera, so as to realize the multi-resolution image acquisition device of a larger visual scene. The full-scene capture camera starts to capture images first, and the remaining four detail capture cameras work autonomously according to the multi-resolution image capture control method, thereby obtaining different resolution pixel representations of the surface of the object to be photographed, and completing multi-resolution adjustment.

As shown in Figure 2, after the initial settings in step 1, the focal points of the five cameras in the scene are coplanar, and the geometric relationship shown in Figure 2 exists. Take the lower left corner of the imaging plane of the four cameras on the right as the origin of the image coordinate system, which corresponds to the points O ₁ , O ₂ , O ₃ , and O ₄ in the figure, and the distance from the center of the viewpoint of each camera to the origin is m. The visible scene space of the 5 fixed-focus cameras is the intersection space of the union of the cone platforms from the focal point to the rear depth of field of the 4 cameras on the right and the visual cone platform of the left camera.

As shown in Figure 3, according to the camera imaging principle: O _c -X _c Y _c Z _c is the camera coordinate system, O-XY is the imaging plane coordinate system, and O′-UV is the image plane coordinate system. The coordinates of point P in the camera coordinate system are (X _c , Y _c , Z _c ), the coordinates of point P′ are (x, y, f); the coordinates of point P′ in the image coordinate system are (u, v) . According to the three-dimensional geometric proportion relationship:

$\frac{\mathrm{<span\; class="notranslate">\; f</span>}}{{\mathrm{<span\; class="notranslate">\; Z</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; x</span>}}{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; the\; y</span>}}{{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 0.1</span>}<span\; class="notranslate">\; )</span>$

Suppose the viewing angle of the camera on the x-axis is θ _x , the resolution is u _x , the viewing angle on the y-axis is θ _y , and the resolution is u _y , then the pixel coordinates of P′ on the image are (u, v) :

$\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}{{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; \&CenterDot;</span>{\mathrm{<span\; class="notranslate">\; tg\&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; x</span>}}{\mathrm{<span\; class="notranslate">\; dx</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}\\ \mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}{{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>{\mathrm{<span\; class="notranslate">\; tg\&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; the\; y</span>}}{\mathrm{<span\; class="notranslate">\; dy</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; v</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}\end{array}\right.<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 02</span>}<span\; class="notranslate">\; )</span>$

Among them, f _x and f _y represent the focal length of the camera in the x and y directions, and dx and dy are the physical dimensions of each pixel on the x-axis and y-axis of the image plane. (u ₀ , v ₀ ) is the physical intersection of the camera optical axis and the image plane.

As shown in Figure 2, let the origins of the image plane coordinate system of the four cameras on the right be O ₁ , O ₂ , O ₃ , O ₄ respectively, and the origin of the image plane coordinate system of the right camera be O ₃ , and make them uniform The coordinate origin of the coordinate system. Assuming that the distance between the four cameras on the right and the optical axis of the upper camera is m, then the four cameras on the right are at any point (u ₁ , v ₁ ), (u ₂ , v ₂ ), (u ₃ , v ₃ ), (u ₄ , v ₄ ) correspond to the coordinates of the unified coordinate system as $({\mathrm{<figure-callout\; id="1"\; label="u"\; filenames="A20081010124100112.png"\; state="\{\{state\}\}">u</figure-callout>}}_1,{\mathrm{<figure-callout\; id="1"\; label="v"\; filenames="A20081010124100112.png"\; state="\{\{state\}\}">v</figure-callout>}}_1+\sqrt{2}m),(u_2+\sqrt{2}m,v_2+\sqrt{2}m),(u_3+\sqrt{2}m,v_3),(u_4,v_4),$ The correspondence between the pixels is used as the basis for the preprocessing of the multi-resolution image acquisition.

As shown in Figure 4, a flow diagram of a multi-resolution image acquisition method is given to realize the adjustment and control of the scene image resolution, including (1) automatically or manually specifying the target object for observation; (2) the background of the target object Eliminate; (3) Divide the imaging display sub-region of the target object; (4) Collect image data from the sub-region camera to generate a high-resolution scene overview; (5) Specify the lossless resolution of the target object to enlarge; (6) Target object edge and Background difference fusion; (7) Determine whether the target object is moving, and repeat the seven steps of collecting image information, which are explained separately as follows:

In step 4.1, the visible target objects in the scene are automatically specified by means of pattern recognition, or the visible target objects in the scene are specified manually by manual operation. When the color difference between the background image and the target object image is obvious, use a software filter with the same color as the background to remove the background, and automatically select the remaining visible target object images. When the background is complex, or the color difference between the target object and the image is similar, key points are manually set in the scene overview image, and the key points are connected to each other to form a convex polygon as the specified visible target object.

In step 4.2, the foreground of the target object is separated from the background according to the range of the target object determined in 4.1. If the background has been eliminated by the method of automatically specifying the target object in step 4.1, this step can be omitted; if the method of manually specifying the target object is used, the image outside the convex polygon is used as the background to be removed, and the remaining part of the image is the foreground of the target object.

Step 4.3, according to the foreground of the target object determined in 4.2, according to the image block shown in FIG. 2, determine the display sub-region of the target object imaging. In the scene overview image, divide the image into four equal-sized and symmetrical square sub-scenes according to Figure 2, and ensure that each sub-scene has a sub-scene detail image acquisition camera. The target object image is also divided into four display sub-regions during the scene division process, and of course, the case that some sub-regions do not contain target object pixels is also included.

In step 4.4, according to the imaging display sub-area of the target object calculated in 4.3, it is determined that the camera responsible for the imaging of this area starts to collect data. Based on the pixel coordinates of the scene overview image, the image coordinates acquired by the sub-scene image acquisition camera are (u ₁ , v ₁ ), (u ₂ , v ₂ ), (u ₃ , v ₃ ), (u ₄ , v ₄ ) , corresponding to the scene overview image coordinate positions are respectively $({\mathrm{<figure-callout\; id="1"\; label="u"\; filenames="A20081010124100112.png"\; state="\{\{state\}\}">u</figure-callout>}}_1,{\mathrm{<figure-callout\; id="1"\; label="v"\; filenames="A20081010124100112.png"\; state="\{\{state\}\}">v</figure-callout>}}_1+\sqrt{2}m),(u_2+\sqrt{2}m,v_2+\sqrt{2}m),(u_3+\sqrt{2}m,v_3),(u_4,v_4),$ And the detail pixels are inserted into the scene overview image according to the coordinate position to obtain a high-resolution scene overview image.

Step 4.5, enlarge or adjust the pixel points of the imaging area of the target object determined in 4.4 according to the input parameters. Input the magnification α of the target object, the minimum is 1, which means no zoom. Taking the center of the target object area as the origin, adjust the distance of the pixel points in the target object area, and each original pixel point translates α times of the original pixel point distance in the four directions of up, down, left, and right. Suppose the distance between the upper and lower adjacent pixels is m; the distance between the left and right adjacent pixels is n, then the enlarged distances become α·m and α·n respectively.

In step 4.6, the target object is enlarged or adjusted according to 4.4. When the α multiple is greater than 1, the original background must be blocked. The enlarged target object is used as a new target object, and the linear interpolation algorithm commonly used in graphics is used to perform interpolation and fusion operations on the edge of the new target object and the new background to achieve a smooth transition between the enlarged target object and the background. Suppose the RGB information of the pixel of the new target object is N _R , N _G , N _B , and the RGB information of the adjacent background pixel is N _R ′, N _G ′, N _B ′, then the inserted n-1 fusion The RGB information of the pixel is as follows:

$<span\; class="notranslate">\; (</span>\frac{<span\; class="notranslate">\; (</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; R</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; R</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; ,</span>\frac{<span\; class="notranslate">\; (</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; G</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; G</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; ,</span>\frac{<span\; class="notranslate">\; (</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>$

$<span\; class="notranslate">\; (</span>\frac{<span\; class="notranslate">\; (</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; R</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; R</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; ,</span>\frac{<span\; class="notranslate">\; (</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; G</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; G</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; ,</span>\frac{<span\; class="notranslate">\; (</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>$

…

$<span\; class="notranslate">\; (</span>\frac{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; R</span>}}<span\; class="notranslate">\; +</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; R</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; ,</span>\frac{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; G</span>}}<span\; class="notranslate">\; +</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; G</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; ,</span>\frac{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}}<span\; class="notranslate">\; +</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; B</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; .</span>$

Step 4.7, determine whether the specified target object moves by the method of video frame difference, if in the two frame images before and after, the pixel point position of the target object deviates, that is, the result of video frame difference is greater than a certain threshold (as N= 10), it means that the target object is moving, then go to step 4.2; otherwise go to step 4.4. Assuming that the RGB information values of the pixels corresponding to the two frames before and after the collected video are N _R , N _G , N _B and ", then the result of video frame difference is N=(N _R ′ _-NR )+(N _G '-N _G )+(N _B '-N _B ).

Since the image acquisition process is a cyclic process, no schematic representation of the end of the process is given in FIG. 4 .

Finally, it should be noted that the above description is only a preferred embodiment of the present invention, and it should be pointed out that, for those of ordinary skill in the art, without departing from the principle of using multiple fixed-focus cameras to realize multi-resolution image acquisition in the present invention Under the premise, several improvements or equivalent replacements can also be made, and these improvements and equivalent replacements should also be regarded as the protection scope of the present invention.

Claims (9)

1, a kind of device that adopts many focusing video cameras to realize the multi-resolution image collection, it is characterized in that: comprise 5 focusing video cameras with position constraint relation, wherein 4 video cameras are on the same surface level, be positioned on same foursquare 4 summits, and 4 focus of camera are identical, are made as f _sOther 1 position for video camera is in a side of above-mentioned 4 video cameras, the axial line of this video camera is perpendicular to 4 shooting machine sides and through foursquare central point O, and the 5th video camera is d to the distance on all the other planes, 4 video camera places, and establishing this focus of camera is f _t, the initialization setting is done in the parameter and the position of 5 focusing video cameras, its position relation satisfies following constraint: f _t=f _s+ d.

2, a kind of acquisition method that adopts many focusing video cameras to realize the device that multi-resolution image is gathered is characterized in that step is as follows:

(1) the visual target object in the automatic given scenario is perhaps by the visual target object in the artificial given scenario of manually-operated;

(2) carry out background according to the scope of destination object and reject, obtain the destination object prospect;

(3), determine that the destination object imaging shows subregion, and in subregion, arrange scene details acquisition camera and whole audience scape acquisition camera according to the destination object prospect;

(4), generate high resolving power scene overview by the camera acquisition view data in the subregion;

(5), specify the harmless resolution enlargement factor of destination object wherein to high resolving power scene overview;

(6) destination object edge and background difference after amplifying are merged;

(7) judge that whether the fusion goal object moves, if move then change step (2); If move then change step (4), repeated acquisition image information.

3, a kind of acquisition method that adopts many focusing video cameras to realize the device that multi-resolution image is gathered according to claim 2, it is characterized in that: the method for the destination object that the automatic or manual appointment in the described step (1) is observed is: when the image aberration of background image and destination object is obvious, adopt the software filter consistent that background is rejected, selected automatically remaining visual target object images with background color; When background is comparatively complicated, or destination object is when close with the image aberration, by manually in the scene overview image key point being set, as the visual target object of appointment.

4, a kind of acquisition method that adopts many focusing video cameras to realize the device that multi-resolution image is gathered according to claim 2, it is characterized in that: the background elimination method of the destination object in the described step (2) is: for artificial appointment target object method, key point is connected to a convex polygon, rejects the image background of convex polygon outside.

5, a kind of acquisition method that adopts many focusing video cameras to realize the device that multi-resolution image is gathered according to claim 2, it is characterized in that: the imaging of the division destination object of described step (3) shows subregion, and the method for arranging scene details acquisition camera and panoramic camera in subregion is: the destination object image is divided into 4 equal and opposite in directions in the process that scene is divided, symmetry, and be shaped as foursquare demonstration subregion, and arrange that 5 have the focusing video cameras that position constraint concerns, wherein 4 video cameras are on the same surface level, be positioned on same foursquare 4 summits, and 4 focus of camera are identical, are made as f _sOther 1 position for video camera is in a side of above-mentioned 4 video cameras, the axial line of this video camera is perpendicular to 4 shooting machine sides and through foursquare central point O, and the 5th video camera is d to the distance on all the other planes, 4 video camera places, and establishing this focus of camera is f _t, the initialization setting is done in the parameter and the position of 5 focusing video cameras, its position relation satisfies following constraint: f _t=f _s+ d.

6, a kind of acquisition method that adopts many focusing video cameras to realize the device that multi-resolution image is gathered according to claim 2, it is characterized in that: the subregion camera acquisition view data of described step (4), the method that generates high resolving power scene overview is: the corresponding relation of determining scene overview image coordinate and sub-scene detail pictures coordinate, based on scene overview image pixel coordinate, it is (u that sub-scene image acquisition camera is gathered the gained image coordinate ₁, v ₁), (u ₂, v ₂), (u ₃, v ₃), (u ₄, v ₄), correspond to scene overview image coordinate position and be respectively $(u_1,v_1+\sqrt{2}m),(u_2+\sqrt{2}m,v_2+\sqrt{2}m),(u_3+\sqrt{2}m,v_3),(u_4,v_4),$ And the details pixel is inserted in the scene overview image by coordinate position, obtain high resolving power scene overview.

7, a kind of acquisition method that adopts many focusing video cameras to realize the device that multi-resolution image is gathered according to claim 2, it is characterized in that: the method that the harmless resolution of the intended target object of described step (5) is amplified is: establishing the resolution enlargement factor is α, the distance of two neighbor pixels is m up and down, about the distance of two neighbor pixels be n, the distance after then amplifying becomes α m and α n respectively.

8, a kind of acquisition method that adopts many focusing video cameras to realize the device that multi-resolution image is gathered according to claim 2, it is characterized in that: the method that amplification target target edges and background difference are merged in the described step (6) is: the interpolation of doing fresh target target edges and new background merges computing, realizes seamlessly transitting of amplification target object and background.If the RGB information of fresh target object pixel point is N _R, N _G, N _B, the RGB information of the background pixel point that is adjacent is N _R', N _G', N _B', then the n-1 of Cha Ruing RGB information that merges pixel is followed successively by:

$(\frac{((n-1)N_R+N_R^{\&prime;})}{n},\frac{((n-1)N_G+N_G^{\&prime;})}{n},\frac{((n-1)N_B+N_B^{\&prime;})}{n}),$

$(\frac{((n-2)N_R+{2N}_R^{\&prime;})}{n},\frac{((n-2)N_G+{2N}_G^{\&prime;})}{n},\frac{((n-2)N_B+2N_B^{\&prime;})}{n}),$

…

$(\frac{(N_R+(n-1)N_R^{\&prime;})}{n},\frac{(N_G+(n-1)N_G^{\&prime;})}{n},\frac{(N_B+(n-1)N_B^{\&prime;})}{n}).$

9, a kind of acquisition method that adopts many focusing video cameras to realize the device that multi-resolution image is gathered according to claim 2, it is characterized in that: judge in the described step (7) that the method whether destination object moves is: be poor N as a result as the foundation of judging whether destination object moves with frame of video, the RGB value of information of establishing the corresponding picture element of video front and back two two field pictures of collection is N _R, N _G, N _BAnd N _R', N _G', N _B', then frame of video is the poor (N of N=as a result _R'-N _R)+(N _G'-N _G)+(N _B'-N _B).

Images