CN103077523A - Method for shooting and taking evidence through handheld camera - Google Patents

Abstract

The invention relates to image forensics technology. The present invention aims to use the calculated oblique angle of the blanking line to authenticate the camera shooting mode. The technical solution adopted by the present invention is a method for obtaining evidence by hand-held camera shooting, which includes the following steps. Step 1: extracting the straight line information of the original picture; Step 2: Locate the corresponding blanking point according to the extracted straight line information; Step 3: Calculate the camera rotation angle; Step 4: Verify the shooting method according to the rotation angle. The invention is mainly applied to image forensics.

Description

Handheld Camera Shooting Evidence Collection Method

technical field

The invention relates to image forensics technology, belongs to the field of computer image content security, and relates to a method for calculating the rotation angle of a camera. Specifically, it relates to a method for taking evidence with a hand-held camera. the

Background technique

With the rapid development of mobile handheld devices and the Internet, data acquisition, especially image data, has become easier and easier, but the requirements for image information processing are also getting higher and higher. Pictures are the most intuitive expression of recording culture. In addition to the intuitive information contained in the pictures, other hidden clues should also be given equal attention, such as the authentication of the way the pictures were taken. Aiming at the above problems, a general calculation method of camera rotation angle proposed by the present invention provides a feasible solution. the

For the detection of the picture shooting method, the shooting can be basically divided into two categories, fixed camera shooting or hand-held camera shooting. In the present invention, shooting with a fixed camera is limited to the camera being placed on the ground, a tripod or other horizontal structures. If the camera is fixed to shoot, the camera's rotation angle should be consistent with the horizon. While shooting with a handheld camera, it is bound to produce a tilt of the camera. According to the camera rotation angle calculated in the present invention, it can be judged whether the picture is taken by a hand-held camera or by a fixed camera. the

It is assumed that the camera belongs to orthographic projection, therefore, the camera rotation angle is equal to the inclination angle of the Vanishing Line (VL) calculated from the picture, and the Vanishing Line can be obtained from the Vanishing Point. the

Vanishing Point (VP) is the convergence point of a set of straight lines on a 2D image, which are a series of parallel lines in 3D space. The blanking point can be used to indicate the direction of the road, and the extension of the two sides of the street will converge to a point. This blanking point is often used to evaluate the upcoming curve on the road to determine the best travel speed and Line, if the blanking point is getting closer, it indicates that the curve is gradually tightening, and if it is getting farther and farther, it indicates that the curve is gradually straightening. At the same time, because the blanking point implies information such as the spatial line and direction corresponding to the image, the blanking point can also be applied to the restoration of the three-dimensional structure of the scene. the

For digital images, the currently commonly used detection algorithms for blanking points are: the first type uses statistical methods to estimate line parameters according to image edge features, and then uses the obtained parameters to calculate blanking points [1]. Although it has an advantage in theory, the time complexity of the algorithm is high, and the efficiency cannot meet the needs of practical applications. The second type is mainly by transforming some information of the image to a limited space, and using space transformation to transform the hidden points at infinity and finite distances into equivalent ones, but after this processing, the lost Information such as the length of the line segment and the distance of the blanking point. the

Contents of the invention

The purpose of the present invention is to use the calculated oblique angle of the hidden line to carry out the authentication of the camera shooting method. The technical solution adopted by the present invention is a method for obtaining evidence by hand-held camera shooting, including the following steps:

Step 1: Extract the straight line information of the original image

In the world coordinate system, there are many parallel or intersecting straight lines in man-made buildings, and such straight lines also exist in man-made objects such as bridges and roads. The obtained projection planes will intersect at a point, which is the blanking point. The camera rotation angle can be expressed by using the hidden line calculated in the picture, that is, the tilt angle of the horizon. The hidden line is obtained through the hidden point. Several parallel lines that are orthogonal to each other in the world coordinate system intersect at two points A and B respectively after perspective transformation. The straight line AB determined by these two hidden points is blanking line;

Step 2: Locate the corresponding blanking point according to the extracted straight line information

In perspective geometric space, the point on a straight line passing through point A in 3D space with direction D=(d ^T , 0) ^T can be written as X(λ)=A+λD, where T represents the transpose operation, d is the direction vector of the line, D is the homogeneous coordinate representation of the vector, and λ is the parameter of the line equation. When the parameter λ changes from 0 to ∞, point X(λ) changes from finite point A to infinite point D, under the action of simple projection camera P=K[I|0], where P is the camera matrix, K is The internal reference matrix, I is the identity matrix, 0 means the zero vector, and the point X(λ) is imaged as

x(λ)=PX(λ), (1)

After expanding (1):

x(λ)=PA+λPD=a+λKd, (2)

Where a is the image of A, so the blanking point v of the line is obtained by taking the limit:

$\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}<span\; class="notranslate">\; \&Right\; Arrow;</span><span\; class="notranslate">\; \&infin;</span>}{\mathrm{<span\; class="notranslate">\; lim</span>}}\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&lambda;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}<span\; class="notranslate">\; \&Right\; Arrow;</span><span\; class="notranslate">\; \&infin;</span>}{\mathrm{<span\; class="notranslate">\; lim</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&lambda;Kd</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$

From this equation (3), it can be seen that parallel straight lines in 3D space, that is, the direction vector D=(d ^T , 0) ^T of the straight line are the same, and the vanishing point obtained by projecting onto the image after perspective transformation is theoretically are the same, so it is sufficient to select two straight lines that are parallel to each other. In this way, only two sets of such two parallel lines are needed to calculate the hidden line, and a total of 4 straight lines are required, and these two sets of straight lines are required to be related. The first choice of the present invention is the group of straight lines orthogonal to each other, so finally only 8 points need to be determined on the picture after perspective transformation to obtain 4 straight lines, thereby calculating two blanking points A and B, and the blanking points thus obtained The hidden point can calculate the slope angle of the hidden line;

If the selected pictures do not have parallel lines, or only have parallel lines but none of these lines intersect, it is no longer applicable;

Step 3: Calculation of camera rotation angle

From the straight line determined in step 2, select 8 points, as shown in Figure 1, {a ₁ , b ₁ ; c ₁ , d ₁ } and {a ₂ , b ₂ ; c ₂ , d ₂ } are two sets of parallel lines, and are correlated between groups. The blanking points A and B can be obtained by the following formula:

A＝(a ₁ ×b ₁ )×(c ₁ ×d ₁ ) (4)

B=(a ₂ ×b ₂ )×(c ₂ ×d ₂ ) (5)

The angle between the hidden line determined by the hidden points A and B and the real horizon is the calculated camera rotation angle;

Step 4: Verify the shooting method according to the rotation angle

Hand-held shooting will cause the camera to tilt, so if the obtained camera rotation angle θ satisfies τ≤θ≤90°, where τ is the threshold, here we set it as a constraint of 1.5°, it will be judged as someone else’s hand-held shooting, otherwise it is fixed shoot. the

In step 1, two sets of parallel lines are mainly used, and these two sets of parallel lines are related and non-parallel, and orthogonality is preferred. the

Technical characteristics and effects of the present invention:

There has not been any report on judging the shooting method of historical pictures by using the angle of rotation of the camera. the

The present invention estimates the rotation angle by using the blanking point, which is simpler and universal than that determined by camera calibration. And the low complexity of the algorithm is also a major advantage of the present invention. the

The method provided by the invention is simple and universal. And as cultural construction is getting more and more attention, and pictures are the most intuitive expression of recording culture, the importance of pictures is self-evident. In addition to the intuitive information contained in the picture, other hidden information should also be given equal attention, such as the information authentication of the shooting method in the present invention. It can be seen that the problem solved by the present invention has a very wide application prospect. the

Description of drawings

Figure 1 is a schematic diagram of hidden lines. In the figure, several parallel lines orthogonal to each other in the world coordinate system intersect at two points A and B respectively after perspective transformation. The straight line AB determined by these two hidden points is blanking line. the

Figure 2 is a fixed shot picture. the

Figure 3 is a picture taken by a handheld camera. the

Detailed ways

The present invention is a detection algorithm of the third type of blanking point, which uses the straight line information in the image to calculate the blanking point [2] [3]. The method has the advantages of being simple and general, and the algorithm complexity is low. the

There are four steps in the hand-held camera shooting forensic technology. The following is a detailed description of these four steps:

Step 1: Extract the straight line information of the original image

In the world coordinate system, there are many parallel or intersecting straight lines in man-made buildings, and such straight lines also exist in man-made objects such as bridges and roads. The obtained projection planes will intersect at a point, which is the blanking point. The camera rotation angle can be expressed by using the hidden line calculated in the picture, that is, the tilt angle of the horizon. The hidden line is obtained through the hidden point, so the selection of the straight line is very important for calculating the rotation angle. If all the straight lines in a picture are parallel to each other, then in theory, these straight lines will only intersect at one point in the end, and it is impossible to calculate the hidden line. Therefore, there should be several groups of parallel straight lines in the picture to be tested, but the groups are not parallel. Several parallel lines orthogonal to each other in the world coordinate system intersect at two points A and B respectively after perspective transformation, and the straight line AB determined by these two hidden points is the hidden line. the

As shown in Figure 1. the

The present invention mainly uses two sets of parallel lines, and these two sets of parallel lines are related and non-parallel, and orthogonal is preferred. the

Step 2: Locate the corresponding blanking point according to the extracted straight line information

Due to the limited space of imaging, the vanishing point is generally located outside the image, and may even be at infinity (when the set of parallel lines and the image plane are parallel to each other). In perspective geometric space, the point on a straight line passing through point A in 3D space with direction D=(d ^T , 0) ^T can be written as X(λ)=A+λD, where T represents the transpose operation, d is the direction vector of the line, D is the homogeneous coordinate representation of the vector, and λ is the parameter of the line equation. When the parameter λ changes from 0 to ∞, point X(λ) changes from finite point A to infinite point D. Under the action of a simple projection camera P=K[I|0], where P is the camera matrix, K is the internal reference matrix, I is the identity matrix, 0 represents the zero vector, and the point X(λ) is imaged as

x(λ)=PX(λ), (1)

After expanding (1):

x(λ)=PA+λPD=a+λKd, (2)

Where a is the image of A, so the blanking point v of the line is obtained by taking the limit:

$\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}<span\; class="notranslate">\; \&Right\; Arrow;</span><span\; class="notranslate">\; \&infin;</span>}{\mathrm{<span\; class="notranslate">\; lim</span>}}\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&lambda;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}<span\; class="notranslate">\; \&Right\; Arrow;</span><span\; class="notranslate">\; \&infin;</span>}{\mathrm{<span\; class="notranslate">\; lim</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&lambda;Kd</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$

From this equation (3), it can be seen that parallel straight lines in 3D space, that is, the direction vector D=(d ^T , 0) ^T of the straight line are the same, and the vanishing point obtained by projecting onto the image after perspective transformation is theoretically Are the same. Therefore, it is sufficient to select two straight lines that are parallel to each other. In this way, only two sets of such two parallel lines are needed to calculate the hidden line, and a total of 4 straight lines are required, and these two sets of straight lines are required to be related. The first choice of the present invention is Groups of lines that are orthogonal to each other. Therefore, in the end, it is only necessary to determine 8 points on the picture after perspective transformation to obtain 4 straight lines, so as to calculate two hidden points A and B, and the oblique angle of the hidden line can be calculated from the obtained hidden points .

If the selected picture does not have straight lines parallel to each other, or only has straight lines parallel to each other but these straight lines do not intersect, then the method of the present invention is no longer applicable. the

Step 3: Calculation of camera rotation angle

From the straight line determined in step 2, select 8 points, as shown in Figure 1, {a ₁ , b ₁ ; c ₁ , d ₁ } and {a ₂ , b ₂ ; c ₂ , d ₂ } are two sets of parallel lines , and there is a correlation between each group. The blanking points A and B can be obtained by the following formula:

A＝(a ₁ ×b ₁ )×(c ₁ ×d ₁ ) (4)

B=(a ₂ ×b ₂ )×(c ₂ ×d ₂ ) (5)

The angle between the hidden line determined by the hidden points A and B and the real horizon is the calculated camera rotation angle. the

Step 4: Verify the shooting method according to the rotation angle

This step is an innovative point of the present invention, using the obtained camera rotation angle to determine the shooting mode of the picture, whether it is fixed shooting (placed on the ground, tripod or other horizontal buildings) or handheld shooting. the

Hand-held shooting will cause the camera to tilt, so if the obtained camera rotation angle θ satisfies the constraint of τ≤θ≤90° (where τ is the threshold value, here we set it to 1.5°), then it is judged to be a hand-held shooting by someone else, otherwise it is Shoot stationary (on the ground, on a tripod or other horizontal structure). the

Beneficial effect

1. Judging the shooting method of historical pictures by using the camera rotation angle has not been reported yet. the

2. The present invention estimates the rotation angle by using the blanking point, which is simpler than that determined by camera calibration.

generic. And the low complexity of the algorithm is also a major advantage of the present invention. the

The method provided by the invention is simple and universal. And as cultural construction is getting more and more attention, and pictures are the most intuitive expression of recording culture, the importance of pictures is self-evident. In addition to the intuitive information contained in the picture, other hidden information should also be given equal attention, such as the information authentication of the shooting method in the present invention. It can be seen that the problem solved by the present invention has a very wide application prospect. the

The present invention is a hand-held camera shooting evidence collection technology based on blanking points, the input is some pictures with straight line information,

It mainly includes parallel lines and intersecting lines, extracts the hidden points of the picture, finds the corresponding hidden lines, and then performs camera

Calculation of rotation angle. In this experiment, 6 fixed pictures and 6 hand-held pictures were selected from the collected pictures.

The picture is used as a test picture to verify the shooting method. the

The following are the experimental results:

The present invention divides the pictures collected into two categories, one is fixed shooting (placed on the ground, tripod or other horizontal buildings), and the other is handheld shooting by others. the

(1) Fixed shooting

Note: The yellow dots in the figure determine a pair of horizontal lines, and the yellow star points determine a pair of horizontal lines. The yellow straight line represents the calculated hidden line (ie the horizon)

The relevant test information statistics table is as follows:

2) Handheld pictures:

Note: In Figure 3, the yellow dots determine a pair of horizontal lines, the yellow star points determine a pair of horizontal lines, and the yellow straight line represents the calculated blanking line (that is, the horizon). the

The relevant test information statistics table is as follows:

After the measurement of the six pictures in Figure 2, the rotation angle θ of the camera conforms to the constraints of 0≤θ≤τ (where τ is the threshold, here we set it to 1.5°), so it is judged that the six pictures are taken by a fixed camera ( on the ground, on a tripod or other level structure). For the six pictures in Figure 3, after calculation, the rotation angle θ of the camera satisfies the constraint of τ≤θ≤90° (where τ is the threshold, here we set it to 1.5°), so it is considered that the six pictures The shooting method is hand-held camera shooting. the

Conclusion: When a picture is obtained, first extract the straight line information that meets the conditions in the picture, that is, there are at least two pairs of parallel lines that intersect, and calculate the blanking point from these straight line information. After obtaining the coordinates of the blanking point, you can calculate the blanking point. The tilt angle of the hidden line, which can be used as the rotation angle of the camera. Because the present invention limits shooting with a fixed camera to when the camera is placed on the ground, a tripod or other horizontal buildings, the theoretical value of the calculated angle is 0° if it is consistent with the horizon. very small. Due to the limitation of subjective and objective conditions, there will be errors. Set the threshold τ as 1.5°. If the obtained camera rotation angle θ satisfies the constraint of τ≤θ≤90°, it is judged to be taken by someone else, otherwise it is taken by a fixed camera (placed on the ground) , tripod or other horizontal structure). the

main reference

[1] MeLean CF, Koyyuri D. "Evanescent point detection by linear clustering", Transactions on Pattern Analysis and Machine Intelligence, 1995, 17(11), 1090-1095

[2] Schaffalitzky F, Zisserman A. "Automatic detection of hidden lines and hidden points by plane grouping", Journal of Image and Visual Computing, 2000, 18(9), 647-658

[3] Rother C. "A New Method for Vanishing Point Detection in Architectural Environments", British Machine Vision Conference (BMVC), Bristol, UK, 2000, 382-391.

Claims (2)

1. A method for obtaining evidence by hand-held camera shooting, is characterized in that, comprises the steps: Step 1: Extract the straight line information of the original image In the world coordinate system, there are many parallel or intersecting straight lines in man-made buildings, and such straight lines also exist in man-made objects such as bridges and roads. The obtained projection planes will intersect at a point, which is the blanking point. The camera rotation angle can be expressed by using the hidden line calculated in the picture, that is, the tilt angle of the horizon. The hidden line is obtained through the hidden point. Several parallel lines orthogonal to each other in the world coordinate system intersect at two points A and B respectively after perspective transformation. The straight line AB determined by these two hidden points is blanking line; Step 2: Locate the corresponding blanking point according to the extracted straight line information In perspective geometric space, the point on a straight line passing through point A in 3D space with direction D=(d ^T , 0) ^T can be written as X(λ)=A+λD, where T represents the transpose operation, d is the direction vector of the line, D is the homogeneous coordinate representation of the vector, and λ is the parameter of the line equation. When the parameter λ changes from 0 to ∞, point X(λ) changes from finite point A to infinite point D, under the action of simple projection camera P=K[I|0], where P is the camera matrix, K is The internal reference matrix, I is the identity matrix, 0 means the zero vector, and the point X(λ) is imaged as x(λ)=PX(λ), (1) After expanding (1): x(λ)=PA+λPD=a+λKd, (2) where a is the image of A, so the blanking point v of the line is obtained by taking the limit: $\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}<span\; class="notranslate">\; \&Right\; Arrow;</span><span\; class="notranslate">\; \&infin;</span>}{\mathrm{<span\; class="notranslate">\; lim</span>}}\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&lambda;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; \&lambda;</span>}<span\; class="notranslate">\; \&Right\; Arrow;</span><span\; class="notranslate">\; \&infin;</span>}{\mathrm{<span\; class="notranslate">\; lim</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&lambda;Kd</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$ From this equation (3), it can be seen that parallel straight lines in 3D space, that is, the direction vector D=(d ^T , 0) ^T of the straight line are the same, and the vanishing point obtained by projecting onto the image after perspective transformation is theoretically are the same, so it is sufficient to select two straight lines that are parallel to each other. In this way, only two sets of such two parallel lines are needed to calculate the hidden line, and a total of 4 straight lines are required, and these two sets of straight lines are required to be related. Therefore, in the end, it is only necessary to determine 8 points on the picture after perspective transformation to obtain 4 straight lines, so as to calculate two hidden points A and B, and the oblique angle of the hidden line can be calculated from the obtained hidden points ; If the selected picture does not have parallel lines, or only has parallel lines but none of these lines intersect, it is no longer applicable; Step 3: Calculation of camera rotation angle From the straight line determined in step 2, select 8 points, as shown in Figure 1, {a ₁ , b ₁ ; c ₁ , d ₁ } and {a ₂ , b ₂ ; c ₂ , d ₂ } are two sets of parallel lines, and are correlated between groups. The blanking points A and B can be obtained by the following formula: A＝(a ₁ ×b ₁ )×(c ₁ ×d ₁ ) (4) B=(a ₂ ×b ₂ )×(c ₂ ×d ₂ ) (5) The angle between the hidden line determined by the hidden points A and B and the real horizon is the calculated camera rotation angle; Step 4: Verify the shooting method according to the rotation angle Hand-held shooting will cause the camera to tilt, so if the obtained camera rotation angle θ satisfies τ≤θ≤90°, where τ is the threshold, here we set it as a constraint of 1.5°, it will be judged as someone else’s hand-held shooting, otherwise it is fixed shoot. 2. A method for taking evidence with a hand-held camera according to claim 1, wherein in step 1, two sets of parallel lines are mainly used, and these two sets of parallel lines are related and non-parallel, preferably orthogonal.

Images