CN103700140A - Spatial modeling method used for linkage of single gun camera and multiple dome cameras - Google Patents

Abstract

The invention discloses a spatial modeling method used for linkage of a single gun camera and multiple dome cameras, relating to the fields of video surveillance, computer vision, pattern recognition and the like. The method comprises the following steps: (1) performing 3D spatial modeling on a monitored scene (201); (2) calculating the mapping relation of a coordinate point of a picture of the gun camera to a coordinate point of the true monitored ground (202); (3) obtaining a tracking detection result of the gun camera (203); (4) obtaining spatial position information of a tracked object (204); (5) calculating the offsets of the dome cameras (205); (6) controlling the dome cameras (206); (7) outputting the tracking result by the dome cameras (207). The tracked object is always positioned in center areas of the pictures of the dome cameras, so the defects of tracking picture discontinuity and dome camera picture judder are overcome; by using relevant mathematical principles, the position of the true object can be accurately and fast calculated by one-time scene calibration; by using the spatial modeling method, mathematical formulas of space geometry can be used for accurately calculating the offsets of the dome cameras, and the object tracking can be more continuous and more accurate.

Description

Space modeling method for the interlock of single gunlock-many balls machine

Technical field

The present invention relates to the fields such as video monitoring, computer vision and pattern-recognition, be specifically related to a kind of space modeling method for the interlock of single gunlock-many balls machine; The method is particularly suitable for the interlock intelligent visual surveillance system (gunlock refers to that rifle type video camera, ball machine refer to ball-shaped camera) of single gunlock-many balls machine.

Background technology

Video monitoring, as very effective means in modern management monitoring and safety precaution, has been widely used in the various public places such as bank, station and office building.In the last few years, due to the active demand of public safety strick precaution and the development of computer technology, the video monitoring direction from traditional mode to Intellectualized monitoring changes, its correlation technique becomes the study hotspot of computer vision gradually, also in succession has some intelligent monitor systems to occur both at home and abroad.The interlock of rifle ball is a kind of important monitoring mode of intelligent monitor system.Rifle ball interlock refers to: with gunlock, target is detected and in gunlock select target, ball machine is followed the tracks of this target.

Gridding method is the rifle ball conventional method that links, it needs a plurality of presetting bits and through repeatedly demarcating, approximately estimate the side-play amount of ball machine, and common gridding method can only make target in ball machine picture and can not guarantee at ball machine picture center, cause following the tracks of that picture is discontinuous, the float of ball machine is more serious, some gridding methods after improving also need with computing repeatedly with and demarcate a plurality of unique points and carry out the relative side-play amount of estimating ball machine, make the complicated and operation result of operating process might not be accurate.

Summary of the invention

The invention provides a kind of space modeling method for the interlock of single gunlock-many balls machine, be intended to the problem existing for prior art, how to overcome in rifle ball linkage process ball machine and follow the tracks of that picture is discontinuous, target departs from the shortcomings such as ball machine picture center, and make more simple and convenient each parameter of operating process can disposable demarcation well, ball machine side-play amount calculates and more accurately meets True Data.

The object of the present invention is achieved like this:

Specifically, this method comprises the following steps:

1. monitoring scene is carried out to 3d spatial modeling

Initial point using the center of the trapezoidal ground region of gunlock monitoring as three-dimensional coordinate system, the line direction at gunlock vertical rod Dian Dao center is as x axle, the proximal border of gunlock guarded region and the direction of distal border be as y direction of principal axis, perpendicular to the vertical rod direction on ground, is z direction of principal axis; Establish after space coordinates, the presetting bit of ball machine is set to true origin place, and four points of trapezoidal ground region, the volume coordinates such as position of the vertical rod point of ball machine and ball machine camera then the position of the ground vertical rod point of gunlock, gunlock camera, gunlock monitored all demarcate out;

2. calculate gunlock picture coordinate points to the mapping relations of true monitoring ground coordinate points

By true monitoring scene and gunlock picture opening relationships, be equivalent to, in image projection to rectangular frame of an isosceles trapezoid, utilize perspective transform principle, calculate gunlock picture coordinate points to the mapping relations of true monitoring ground coordinate points;

3. the tracking of gunlock detects

Gunlock adopts target detection track algorithm to detect target, and returns to target's center's coordinate;

4. obtain the spatial positional information of tracking target;

Coordinate by target in gunlock picture is updated in the mapping relations that step calculates in 2., thereby obtains target in the co-ordinate position information of true monitoring ground;

5. calculate the side-play amount of ball machine;

Using a upper positional information of target and current location information as input, and calibrate the volume coordinate information of ball machine, build a space multistory geometric model, utilize geometric relationship, can calculate the side-play amount of ball machine.

6. ball machine is controlled;

The ball machine side-play amount that 5. step is calculated is done as required and is inputted ball machine after relevant treatment and carry interface function, with three-dimensional localization function control marble forming machine, rotates, and makes intended target be positioned at the central area of ball machine picture.

7. ball machine tracking results output.

The present invention has following advantages and good effect:

1, tracking target is all the time in ball machine picture central area, and overcome the shortcoming of following the tracks of the discontinuous and ball machine float of picture;

2, use relevant mathematical principle, by disposable scene calibration, just can calculate quickly and accurately the position of real goal;

3, the method that adopts spatial modeling, can calculate the side-play amount of ball machine accurately with the mathematical formulae of space geometry, make target following more accurate more continuously.

Accompanying drawing explanation

Fig. 1 is the structural representation of native system;

Fig. 2 is the workflow diagram of this method;

Fig. 3 is the space geometry graph of a relation of ball machine and target front and back position;

Fig. 4 calculates gunlock picture coordinate points to the process flow diagram of true monitoring ground coordinate points mapping relations;

Fig. 5 is the process flow diagram that calculates ball machine side-play amount;

Fig. 6 is the block diagram of intelligent network camera.

Embodiment

Below in conjunction with drawings and Examples to the detailed description of the invention:

One, the spatial modeling system (abbreviation system) linking for single gunlock-many balls machine

1, overall

As Fig. 1, native system comprises gunlock 101, ball machine 102 and switch 103;

The single machine 101 of robbing is connected with switch 103 respectively with a plurality of ball machines 102.

2, functional part

1) gunlock 101

As Fig. 6, robbing machine 101 is existing intelligent network camera A, comprises LENS optical lens 1, ccd image sensor 2, signals collecting converter 3, HV signal driver 4, FPGA programmable logic processor 5, H264 record a video scrambler 6, SDRAM storer 7 and ARM center processor 8;

Be provided with DSP image processor 9;

LENS optical lens 1 is connected with ccd image sensor 2 front and back, and ccd image sensor 2, signals collecting converter 3, FPGA programmable logic processor 5, HV signal driver 4 and ccd image sensor 2 connect and compose successively loop and realize image sequence acquisition function;

FPGA programmable logic processor 5, H264 video recording scrambler 6 are connected successively with ARM center processor 8, realize image/video code storage function;

FPGA programmable logic processor 5, SDRAM storer 7, DSP image processor 9 and ARM center processor 8 are connected successively, realize image sequence intellectual analysis function.

DSP image processor 9 is High Performance DSP image processors of dominant frequency 800Mhz, and this device is selected the dsp processor of the series such as TMS320DM64X, TMS320DM81X.

DSP image processor 9 is realized the lights state detection method based on video analysis that the present invention proposes, and moving vehicle is caught to record and the testing result of making a dash across the red light is exported to ARM center processor 8.

Its function is to realize video data to gather and intellectual analysis, i.e. the detection and tracking of target all here complete.

2) ball machine 102

Ball machine 102 is a kind of universal products, and its function is that the picture of target is carried out to real-time follow-up.

3) switch 103

Switch 103 is a kind of universal products, and its function is that the data of gunlock 101 and ball machine 102 are transmitted.

Two, method

1, total step

As Fig. 2, this method comprises the following steps:

1. monitoring scene is carried out to 3d spatial modeling-201;

2. calculate gunlock picture coordinate points to mapping relations-202 of true monitoring ground coordinate points;

3. tracking testing result-203 of gunlock;

4. obtain spatial positional information-204 of tracking target;

5. calculate side-play amount-205 of ball machine;

6. ball machine controls-206;

7. the output of ball machine tracking results-207.

2, step by step

1) 2. step calculates gunlock picture coordinate points to the mapping relations of true monitoring ground coordinate points

As Fig. 4, concrete steps are as follows:

A, by demarcation, obtain coordinate figure-401 on four summits of gunlock monitoring ground;

B, 4 apex coordinates of the coordinate of 4 frontier points of gunlock picture and monitoring ground respectively in the mathematical formulae of substitution projective transformations, are obtained to 4 matrix equality-402;

C, according to 4 matrix equalities, can calculate 8 parameters of the inside, obtain expression formula-403 of projective transformation.

2) 5. step calculates side-play amount-205 of ball machine

As Fig. 5, concrete steps are as follows:

A, using the front and back position of ball machine and target as input-501

As Fig. 3, by a upper positional information and the current location information of the spatial positional information of ball machine, target, put A, B, O, D respectively as known input;

B, set up solid geometric pattern-502

As Fig. 3, in the direction on BO limit, add an auxiliary point C, make BC=BD, ball machine is the skew of vertical direction from O to C, only has the skew of horizontal direction from C to D, α, β angle are the deviation angle of ball machine;

C, calculating deviation angle

Or calculating three sides of a triangle-503, vertical deviation angle place

As Fig. 3, calculate three limits of Δ OAC:

OA:

in Δ OAB, use Pythagorean theorem,

OC:OC=BC-OB=BD-OB, if BD is greater than OB,

Or OC=OB-BC=OB-BD, if OB is greater than BD,

AC： $\mathrm{AC}-\sqrt{{\mathrm{<figure-callout\; id="2"\; label="AB"\; filenames="HDA0000453366370000041.png"\; state="\{\{state\}\}">AB</figure-callout>}}^2+{\mathrm{<figure-callout\; id="2"\; label="BC"\; filenames="HDA0000453366370000041.png"\; state="\{\{state\}\}">BC</figure-callout>}}^2}=\sqrt{{\mathrm{<figure-callout\; id="2"\; label="AB"\; filenames="HDA0000453366370000041.png"\; state="\{\{state\}\}">AB</figure-callout>}}^2+{\mathrm{<figure-callout\; id="2"\; label="BD"\; filenames="HDA0000453366370000041.png"\; state="\{\{state\}\}">BD</figure-callout>}}^2},$ In Δ ABC, use Pythagorean theorem,

At Δ OAC, use the triangle cosine law, calculate α angle-504;

Or three sides of a triangle-505, calculated level deviation angle place

As Fig. 3, calculate three limits of Δ CAD:

AC: calculate

AD:AD=AC, Δ ABC and Δ ABD congruent triangles,

CD：

Calculate ∠ OBD, known three limits use the triangle cosine law in triangle Δ BOD,

Calculate CD, known both sides and angle use the triangle cosine law in triangle Δ BCD,

At Δ CAD, use the triangle cosine law, calculate β angle-506.

Claims (3)

1. for a space modeling method for single gunlock-many balls machine interlock, it is characterized in that comprising the following steps:

1. monitoring scene is carried out to 3d spatial modeling (201)

Initial point using the center of the trapezoidal ground region of gunlock monitoring as three-dimensional coordinate system, the line direction at gunlock vertical rod Dian Dao center is as x axle, the proximal border of gunlock guarded region and the direction of distal border be as y direction of principal axis, perpendicular to the vertical rod direction on ground, is z direction of principal axis; Establish after space coordinates, the presetting bit of ball machine is set to true origin place, and four points of trapezoidal ground region, the volume coordinates such as position of the vertical rod point of ball machine and ball machine camera then the position of the ground vertical rod point of gunlock, gunlock camera, gunlock monitored all demarcate out;

2. calculate gunlock picture coordinate points to the mapping relations (202) of true monitoring ground coordinate points

Utilize perspective transform principle, calculate gunlock picture coordinate points to the mapping relations of true monitoring ground coordinate points;

3. the tracking of gunlock detects (203)

Gunlock adopts target detection track algorithm to detect target, and returns to target's center's coordinate;

4. obtain the spatial positional information (204) of tracking target

Coordinate by target in gunlock picture is updated in the mapping relations that step calculates in 2., thereby obtains target in the co-ordinate position information of true monitoring ground;

5. calculate the side-play amount (205) of ball machine

Using a upper positional information of target and current location information as input, and calibrate the volume coordinate information of ball machine, build a space multistory geometric model, utilize geometric relationship, can calculate the side-play amount of ball machine;

6. ball machine is controlled (206)

The ball machine side-play amount that 5. step is calculated is done as required and is inputted ball machine after relevant treatment and carry interface function, with three-dimensional localization function control marble forming machine, rotates, and makes intended target be positioned at the central area of ball machine picture;

7. ball machine tracking results output (207).

2. by a kind of space modeling method for the interlock of single gunlock-many balls machine claimed in claim 1, it is characterized in that described step 2.:

A, by demarcation, obtain the coordinate figure (401) on four summits of gunlock monitoring ground;

B, 4 apex coordinates of the coordinate of 4 frontier points of gunlock picture and monitoring ground respectively in the mathematical formulae of substitution projective transformations, are obtained to 4 matrix equalities (402);

C, according to 4 matrix equalities, can calculate 8 parameters of the inside, obtain the expression formula (403) of projective transformation.

3. by a kind of space modeling method for the interlock of single gunlock-many balls machine claimed in claim 1, it is characterized in that described step 5.:

A, using the front and back position of ball machine and target as input (501)

By a upper positional information and the current location information of the spatial positional information of ball machine, target, put A, B, O, D respectively as known input;

B, set up solid geometric pattern (502)

In the direction on BO limit, add an auxiliary point C, make BC=BD, ball machine is the skew of vertical direction from O to C, only has the skew of horizontal direction from C to D, and α, β angle are the deviation angle of ball machine;

C, calculating deviation angle

Or calculating vertical deviation angle place three sides of a triangle (503)

Calculate three limits of Δ OAC:

OA:

in Δ OAB, use Pythagorean theorem,

OC:OC=BC-OB=BD-OB, if BD is greater than OB,

Or OC=OB-BC=OB-BD, if OB is greater than BD,

AC： $\mathrm{AC}=\sqrt{{\mathrm{AB}}^2+{\mathrm{BC}}^2}=\sqrt{{\mathrm{AB}}^2+{\mathrm{BD}}^2},$ In Δ ABC, use Pythagorean theorem,

At Δ OAC, use the triangle cosine law, calculate α angle (504);

Or calculated level deviation angle place three sides of a triangle (505)

Calculate three limits of Δ CAD:

AC: calculate

AD:AD=AC, Δ ABC and Δ ABD congruent triangles,

CD：

Calculate ∠ OBD, known three limits use the triangle cosine law in triangle Δ BOD,

Calculate CD, known both sides and angle use the triangle cosine law in triangle Δ BCD,

At Δ CAD, use the triangle cosine law, calculate β angle (506).

Images