CN105120229A - Wide-angle lens omnidirectional monitor and image matching method - Google Patents

Abstract

The invention relates to a columnar panoramic monitoring device, comprising a multi-path image collecting unit, an image processor and an image displayer. The output end of the multi-path image collecting unit is connected with the input end of the image processor. The output end of the image processor is connected with the input end of the image displayer. The multi-path image collecting unit comprises 2-16 paths of image collecting devices arranged by layer. The 2-16 paths of image collecting devices are uniformly distributed on the circumferences of the upper layer and the lower layer. The image processor comprises an image reading and preprocessing module, an image distortion correction conversion module, an index image distortion correction module, an image projecting cylinder module, a template matching algorithm based module, a graying and gap rejecting module, and an ease-in-out fusion module. The video acquired by the monitor is an omnidirectional image without dead angle in a monitored area, and one monitor can replace a plurality of monitors shooting in different angles, so that the monitoring cost is saved, and the omnidirectional clear video image of the same area can be acquired conveniently.

Description

The method of wide-angle lens conduct monitoring at all levels device and images match

Technical field

The present invention relates to a kind of method of column full view monitoring device and images match, being specially adapted to a camera cannot the comprehensive occasion without dead angle shooting monitoring image.

Background technology

Traditional camera often adopts a photo-sensitive cell, thereafter a video processing unit is connect, the image that watch-dog obtains is the image frame in the region of this pick-up lens aligning, this scheme monitoring limited coverage area, there is large-scale blind area and dead angle, the picture beyond pick-up lens alignment area cannot be photographed.

Summary of the invention

The object of the invention is to solve Problems existing in above-mentioned technology, and provide one all-roundly to monitor, there is no the column full view monitoring device at blind area and dead angle and the method for images match.

For achieving the above object, a kind of column full view monitoring device of the present invention's design, comprise multiplex image acquisition unit, image processor, an image display, the output of described multiplex image acquisition unit connects the input of image processor, and the output of described image processor connects the input of image display, described multiplex image acquisition unit is the image capture device of 2-16 road layered arrangement, and described 2-16 road image capture device is evenly distributed on lower leaf circumferentially, described image processor comprise image reading and pretreatment module, image distortion correction conversion module, thumbnail distortion correction module, image projection cylinder module, based on template matching algorithm module, gray processing and reject blank module, be fade-in and gradually go out Fusion Module, the output of described image reading and pretreatment module connects the input of image distortion correction conversion module, the input of the output join index image distortion correction module of described image distortion correction conversion module, the output connection layout of described thumbnail distortion correction module is as the input of projection column face mould block, the output of described image projection cylinder module connects the input based on template matching algorithm module, the described output based on template matching algorithm module connects gray processing and rejects the input of blank module, the output connection of described gray processing and rejecting blank module is fade-in the input gradually going out Fusion Module.

In technique scheme, described image distortion correction conversion module comprises distortion correction direct transform module, result deposits matrix module and distortion correction inversion module, containing distortion correction direct transform algorithm in described distortion correction direct transform module, the size of correcting image is gone out, described distortion correction direct transform formula by distortion correction direct transform formulae discovery:

$\left\{\begin{array}{c}x=u\[1+k\left(u^2+v^2\right)\]\\ y=v\[1+k\left(u^2+v^2\right)\]\end{array}\right.$

It is two matrixes created after distortion correction direct transform that the above results deposits matrix module, for depositing above-mentioned distortion correction direct transform result;

Containing inversion algorithm in described distortion correction inversion module, obtain coordinate points (x, y) → (u, the v) after image rectification by inverse transformation formulae discovery, inverse transformation formula is:

$\left\{\begin{array}{c}u=x\frac{\sqrt[3]{\sqrt{\frac{1}{4}-\frac{1}{27k(x^2+y^2)}}-\frac{1}{2}}+\sqrt[3]{-\sqrt{\frac{1}{4}-\frac{1}{27k(x^2+y^2)}}-\frac{1}{2}}}{\sqrt[3]{k(x^2+y^2)}}\\ v=y\frac{\sqrt[3]{\sqrt{\frac{1}{4}-\frac{1}{27k(x^2+y^2)}}-\frac{1}{2}}+\sqrt[3]{-\sqrt{\frac{1}{4}-\frac{1}{27k(x^2+y^2)}}-\frac{1}{2}}}{\sqrt[3]{k(x^2+y^2)}}\end{array}\right.$

In above-mentioned distortion correction direct transform formula and inverse transformation formula, x, y are abscissa and the ordinate of the image mid point that image reading and pretreatment module obtain, u, v are the abscissa and ordinate put after image rectification, k is the average distortion coefficient after pick-up lens way is determined, inputs to thumbnail distortion correction module for thumbnail distortion correction after the corresponding relation storage of coordinate.

In technique scheme, described thumbnail distortion correction module obtains result and deposits two index matrixs in matrix module, makes the method for index of reference accelerate distortion correction speed.

In technique scheme, described image projection cylinder module for by the image projection after distortion correction to the same reference coordinate face of cylinder, projective transformation formula:

$\left\{\begin{array}{c}u=r\sin \left(\frac{\mathrm{\ω}}{2}\right)+r\sin \[\arctan \left(\frac{x-W/2}{r}\right)\]\\ v=\frac{H}{2}-\frac{r(y-H/2)}{d}\end{array}\right.$

Wherein: x, y are abscissa and the ordinate of pixel on the image after distortion correction, u, v are abscissa and the ordinate of pixel after projective transformation, and r is focal length, and W is the pixel wide of every road image acquisition units, H is the pixels tall of every road image acquisition units

$d=\sqrt{r^2+{(W/2-x)}^2},$ ω＝2arctan(W/2f)。

The present invention provides a kind of column full view monitoring device to carry out the method for images match simultaneously, by following steps, realizes the coupling of image:

(1), the horizontal displacement TempDsplyX of two width images and vertical displacement TempDsplyY is simulated by two-layer circulation;

(2), template size parameter TemplateX and TemplateY;

(3), then choose one piece of fixing size area at image lap to carry out grey value difference statistics and compare stored in the Difference variable that TempDifference and initial value are 0, preserve the less value of numerical value in Difference, and current TemplateX and TemplateY value is stored in DsplyX, DsplyY respectively simultaneously; Because be fixed area, so cycling condition TempDsplyX initial value can not be less than template width;

(4), circulation terminate after, the value stored in variables D splyX, DsplyY is exactly two required figure position relationship values, thus realizes the registration of image.

In technique scheme, described gray processing and rejecting blank module are used for image empty useless region to remove, and gray level image reduces amount of calculation, accelerates the speed of service.

In technique scheme, described in be fade-in and gradually go out Fusion Module by following steps, realize image 1 and image 2 fused image in two width images level and smooth, visually can not differentiate difference:

(1), at image about 1 add null matrix, carry out the vertical translation of analog image with this;

(2), at image 1 add null matrix below, carry out the horizontal translation of analog image with this;

(3), realize being fade-in of image by circulation and gradually go out splicing;

(3.1), for the content after lap, the content of direct duplicating image 2;

(3.2), in overlapping region, image 2 has data, the unsure certificate of image 1, equals image 2 data;

(3.3), for the region of lap, adopt and be fade-in the method gradually gone out:

(3.31), suppose that image lap is W unit;

(3.32), first unit gray value: image 1* (1/W)+image 2* [1-(1/W)];

(3.33), the after this distribution of coefficient changes a unit respectively.

The way of the pick-up lens in the present invention is indefinite, can be 2-16 road, multichannel pick-up lens, from different angle acquisition image informations, reads certain frame from video, carry out distortion correction, and by under being based upon same referential between image and image, using the face of cylinder as with reference to system, realize conic projection, method based on template matches realizes the coupling of image, image empty useless region is removed, and gray level image reduces amount of calculation, accelerate the speed of service.After the coupling of picture position, can due to reasons such as exposures between image, make the luminance difference of image larger, direct fused images, obvious difference can be brought, merge so employing is fade-in the mode gradually gone out, make image smoothing, visually can not differentiate difference, obtain the comprehensive clear picture of complete reaction material object.

The image that watch-dog of the present invention obtains is that in guarded region, omnidirectional images is without dead angle, and a watch-dog can replace multiple watch-dog taken from different perspectives, saves monitoring cost, conveniently can obtain the comprehensive clear video image in same region simultaneously.

Accompanying drawing explanation

Fig. 1 is principle of the invention block diagram;

Fig. 2 is image processor of the present invention (2) theory diagram;

In figure: 1-pick-up lens; 2-image processor; 3-image display;

2.1-image reading and pretreatment module; 2.2-image distortion correction conversion module; 2.3-thumbnail distortion correction module; 2.4-image projection cylinder module; 2.5-is based on template matching algorithm module; 2.6-gray processing and rejecting blank module; 2.7-is fade-in and gradually goes out Fusion Module.

Embodiment

Below in conjunction with the drawings and specific embodiments, the present invention is described in further detail:

As the column full view monitoring device of Fig. 1, comprise multiplex image acquisition unit (1), an image processor (2), an image display (3), the output of multiplex image acquisition unit (1) connects the input of image processor (2), and the output of image processor (2) connects the input of image display (3), the image capture device that multiplex image acquisition unit (1) is 2-16 road layered arrangement, 2-16 road image capture device is evenly distributed on lower leaf circumferentially, image processor (2) comprises image reading and pretreatment module (2.1), image distortion correction conversion module (2.2), thumbnail distortion correction module (2.3), image projection cylinder module (2.4), based on template matching algorithm module (2.5), gray processing and rejecting blank module (2.6), be fade-in and gradually go out Fusion Module (2.7), as shown in Figure 2, the output of image reading and pretreatment module (2.1) connects the input of image distortion correction conversion module (2.2), the input of output join index image distortion correction module (2.3) of image distortion correction conversion module (2.2), the output connection layout of thumbnail distortion correction module (2.3) is as the input of projection column face mould block (2.4), the output of image projection cylinder module (2.4) connects the input based on template matching algorithm module (2.5), output based on template matching algorithm module (2.5) connects gray processing and rejects the input of blank module (2.6), the output connection of gray processing and rejecting blank module (2.6) is fade-in the input gradually going out Fusion Module (2.7).

Image distortion correction conversion module (2.2) comprises distortion correction direct transform module, result deposits matrix module and distortion correction inversion module, containing distortion correction direct transform algorithm in described distortion correction direct transform module, the size of correcting image is gone out, described distortion correction direct transform formula by distortion correction direct transform formulae discovery:

$\left\{\begin{array}{c}x=u\[1+k\left(u^2+v^2\right)\]\\ y=v\[1+k\left(u^2+v^2\right)\]\end{array}\right.$

It is two matrixes created after distortion correction direct transform that result deposits matrix module, for depositing above-mentioned distortion correction direct transform result;

Containing inversion algorithm in distortion correction inversion module, obtain coordinate points (x, y) → (u, the v) after image rectification by inverse transformation formulae discovery, inverse transformation formula is:

$\left\{\begin{array}{c}u=x\frac{\sqrt[3]{\sqrt{\frac{1}{4}-\frac{1}{27k(x^2+y^2)}}-\frac{1}{2}}+\sqrt[3]{-\sqrt{\frac{1}{4}-\frac{1}{27k(x^2+y^2)}}-\frac{1}{2}}}{\sqrt[3]{k(x^2+y^2)}}\\ v=y\frac{\sqrt[3]{\sqrt{\frac{1}{4}-\frac{1}{27k(x^2+y^2)}}-\frac{1}{2}}+\sqrt[3]{-\sqrt{\frac{1}{4}-\frac{1}{27k(x^2+y^2)}}-\frac{1}{2}}}{\sqrt[3]{k(x^2+y^2)}}\end{array}\right.$

In distortion correction direct transform formula and inverse transformation formula, x, y are abscissa and the ordinate of the image mid point that image reading and pretreatment module (2.1) obtain, u, v are the abscissa and ordinate put after image rectification, k is the average distortion coefficient after pick-up lens (1) way is determined, inputs to thumbnail distortion correction module (2.3) for thumbnail distortion correction after the corresponding relation storage of coordinate.

Thumbnail distortion correction module (2.3) obtains result and deposits two index matrixs in matrix module, makes the method for index of reference accelerate distortion correction speed.

Image projection cylinder module (2.4) for by the image projection after distortion correction to the same reference coordinate face of cylinder, projective transformation formula:

$\left\{\begin{array}{c}u=r\sin \left(\frac{\mathrm{\ω}}{2}\right)+r\sin \[\arctan \left(\frac{x-W/2}{r}\right)\]\\ v=\frac{H}{2}-\frac{r(y-H/2)}{d}\end{array}\right.$

Wherein: x, y are abscissa and the ordinate of pixel on the image after distortion correction, u, v are abscissa and the ordinate of pixel after projective transformation, and r is focal length, and W is the pixel wide of every road image acquisition units, H is the pixels tall of every road image acquisition units

$d=\sqrt{r^2+{(W/2-x)}^2},$ ω＝2arctan(W/2f)。

Column full view monitoring device carries out the method for images match, realizes the coupling of image based on template matching algorithm module (2.5):

(1), the horizontal displacement TempDsplyX of two width images and vertical displacement TempDsplyY is simulated by two-layer circulation;

(2), template size parameter TemplateX and TemplateY;

(3), then choose one piece of fixing size area at image lap to carry out grey value difference statistics and compare stored in the Difference variable that TempDifference and initial value are 0, preserve the less value of numerical value in Difference, and current TemplateX and TemplateY value is stored in DsplyX, DsplyY respectively simultaneously; Because be fixed area, so cycling condition TempDsplyX initial value can not be less than template width;

Gray processing and rejecting blank module (2.6) are for removing image empty useless region, and gray level image reduces amount of calculation, accelerates the speed of service.

(4), circulation terminate after, the value stored in variables D splyX, DsplyY is exactly two required figure position relationship values, thus realizes the registration of image.

Be fade-in and gradually go out Fusion Module (2.7) by following steps, realize image 1 and image 2 fused image in two width images and smoothly, visually can not differentiate difference:

(1), at image about 1 add null matrix, carry out the vertical translation of analog image with this;

(2), at image 1 add null matrix below, carry out the horizontal translation of analog image with this;

(3), realize being fade-in of image by circulation and gradually go out splicing;

(3.1), for the content after lap, the content of direct duplicating image 2;

(3.2), in overlapping region, image 2 has data, the unsure certificate of image 1, equals image 2 data;

(3.3), for the region of lap, adopt and be fade-in the method gradually gone out:

(3.31), suppose that image lap is W unit;

(3.32), first unit gray value: image 1* [1/W]+image 2* [1-1/W];

(3.33), the after this distribution of coefficient changes a unit respectively.

After above-mentioned process, the image that watch-dog obtains be in guarded region omnidirectional images without dead angle, a watch-dog can replace multiple watch-dog taken from different perspectives, saves monitoring cost, conveniently can obtain the comprehensive clear video image in same region simultaneously.

Above embodiment only for illustration of technical scheme of the present invention, is not intended to limit.Other carry out the amendment of structure technology or principle scheme with reference to previous embodiment to the present invention, or modify for partial content and replace or assembling, do not make the essence of appropriate technical solution depart from the spirit and scope of the technical scheme that the present invention relates to.

The content be not described in detail in this specification belongs to the known prior art of professional and technical personnel in the field.

Claims (7)

1. a column full view monitoring device, it is characterized in that: comprise multiplex image acquisition unit (1), an image processor (2), an image display (3), the output of described multiplex image acquisition unit (1) connects the input of image processor (2), and the output of described image processor (2) connects the input of image display (3);

Described multiplex image acquisition unit (1) has 2-16 road layered arrangement image capture device, and described 2-16 road image capture device is evenly distributed on lower leaf circumferentially;

Described image processor (2) comprise image reading and pretreatment module (2.1), image distortion correction conversion module (2.2), thumbnail distortion correction module (2.3), image projection cylinder module (2.4), based on template matching algorithm module (2.5), gray processing and reject blank module (2.6), be fade-in and gradually go out Fusion Module (2.7), the output of described image reading and pretreatment module (2.1) connects the input of image distortion correction conversion module (2.2), the input of output join index image distortion correction module (2.3) of described image distortion correction conversion module (2.2), the output connection layout of described thumbnail distortion correction module (2.3) is as the input of projection column face mould block (2.4), the output of described image projection cylinder module (2.4) connects the input based on template matching algorithm module (2.5), the described output based on template matching algorithm module (2.5) connects gray processing and rejects the input of blank module (2.6), the output connection of described gray processing and rejecting blank module (2.6) is fade-in the input gradually going out Fusion Module (2.7).

2. column full view monitoring device according to claim 1, it is characterized in that: described image distortion correction conversion module (2.2) comprises distortion correction direct transform module, result deposits matrix module and distortion correction inversion module, containing distortion correction direct transform algorithm in described distortion correction direct transform module, the size of correcting image is gone out, described distortion correction direct transform formula by distortion correction direct transform formulae discovery:

$\left\{\begin{array}{c}x=u\[1+k(u^2+v^2)\]\\ y=v\[1+k(u^2+v^2)\]\end{array}\right.$

It is two matrixes created after distortion correction direct transform that the above results deposits matrix module, for depositing above-mentioned distortion correction direct transform result;

Containing inversion algorithm in described distortion correction inversion module, obtain coordinate points (x, y) → (u, the v) after image rectification by inverse transformation formulae discovery, inverse transformation formula is:

$\left\{\begin{array}{c}u=x\frac{\sqrt[3]{\sqrt{\frac{1}{4}-\frac{1}{27k(x^2+y^2)}}-\frac{1}{2}}+\sqrt[3]{-\sqrt{\frac{1}{4}-\frac{1}{27k(x^2+y^2)}}-\frac{1}{2}}}{\sqrt[3]{k(x^2+y^2)}}\\ v=y\frac{\sqrt[3]{\sqrt{\frac{1}{4}-\frac{1}{27k(x^2+y^2)}}-\frac{1}{2}}+\sqrt[3]{-\sqrt{\frac{1}{4}-\frac{1}{27k(x^2+y^2)}}-\frac{1}{2}}}{\sqrt[3]{k(x^2+y^2)}}\end{array}\right.$

In above-mentioned distortion correction direct transform formula and inverse transformation formula, x, y are abscissa and the ordinate of the image mid point that image reading and pretreatment module (2.1) obtain, u, v are the abscissa and ordinate put after image rectification, k is the average distortion coefficient after pick-up lens (1) way is determined, inputs to thumbnail distortion correction module (2.3) for thumbnail distortion correction after the corresponding relation storage of coordinate.

3. column full view monitoring device according to claim 2, is characterized in that: described thumbnail distortion correction module (2.3) obtains result and deposits two index matrixs in matrix module, makes the method for index of reference accelerate distortion correction speed.

4. column full view monitoring device according to claim 3, is characterized in that: described image projection cylinder module (2.4) for by the image projection after distortion correction to the same reference frame face of cylinder, projective transformation formula:

$\left\{\begin{array}{c}u=r\sin \left(\frac{\mathrm{\ω}}{2}\right)+r\sin \[\arctan \left(\frac{x-W/2}{r}\right)\]\\ v=\frac{H}{2}-\frac{r(y-H/2)}{d}\end{array}\right.$

Wherein: x, y are abscissa and the ordinate of pixel on the image after distortion correction, u, v are abscissa and the ordinate of pixel after projective transformation, and r is focal length, and W is the pixel wide of every road image acquisition units, H is the pixels tall of every road image acquisition units

$d=\sqrt{r^2+{(W/2-x)}^2},$ $\mathrm{\ω}=2\arctan (W/2f).$

5. column full view monitoring device according to claim 1, is characterized in that: described gray processing and rejecting blank module (2.6) are for removing image empty useless region, and gray level image reduces amount of calculation, accelerates the speed of service.

6. column full view monitoring device according to claim 1 carries out a method for images match, it is characterized in that: described based on template matching algorithm module (2.5) by following steps, realize the coupling of image:

(1), the horizontal displacement TempDsplyX of two width images and vertical displacement TempDsplyY is simulated by two-layer circulation;

(2), template size parameter TemplateX and TemplateY;

(3), then choose one piece of fixing size area at image lap to carry out grey value difference statistics and compare stored in the Difference variable that TempDifference and initial value are 0, preserve the less value of numerical value in Difference, and current TemplateX and TemplateY value is stored in DsplyX, DsplyY respectively simultaneously; Because be fixed area, so cycling condition TempDsplyX initial value can not be less than template width;

(4), circulation terminate after, the value stored in variables D splyX, DsplyY is exactly two required figure position relationship values, thus realizes the registration of image.

7. column full view monitoring device according to claim 6 carries out the method for images match, it is characterized in that: described in be fade-in gradually to go out Fusion Module (2.7) and pass through following steps, realize image 1 and image 2 fused image in two width images and smoothly, visually can not differentiate difference:

(1), at image about 1 add null matrix, carry out the vertical translation of analog image with this;

(2), at image 1 add null matrix below, carry out the horizontal translation of analog image with this;

(3), realize being fade-in of image by circulation and gradually go out splicing;

(3.1), for the content after lap, the content of direct duplicating image 2;

(3.2), in overlapping region, image 2 has data, the unsure certificate of image 1, equals image 2 data;

(3.3), for the region of lap, adopt and be fade-in the method gradually gone out:

(3.31), suppose that image lap is W unit;

(3.32), first unit gray value: image 1* (1/W)+image 2* [1-(1/W)];

(3.33), the after this distribution of coefficient changes a unit respectively.