CN105513072A - PTZ correction method - Google Patents

Abstract

The present invention relates to a PTZ correction method, belonging to the technical field of PTZ calibration. Firstly, the image of a preset position is collected as a template image, and an azimuth, a pitch angle and a camera zoom ratio at that time are recorded; then the template image is subjected to modeling, and the outline of a target object is extracted with the center point of the template image as a center; then a PTZ and a camera are set to carry out image collection with the preset position of the template image in collection, the collected image and the template image are compared, and the offset pixel of the target object is calculated; and finally the PTZ offset angle is calculated according to the obtained offset pixel, and the PTZ offset angle is added to the PTZ preset position, and a corrected absolute angle is obtained. The machine vision technology is used to identify a target point, the PTZ posture can be quickly adjusted, an external sensor is not needed, the alignment precision of the PTZ is improved, and the target deviation error in shooting the image is reduced.

Description

A kind of The Cloud Terrace bearing calibration

Technical field

The present invention relates to a kind of The Cloud Terrace bearing calibration, belong to The Cloud Terrace collimation technique field.

Background technology

The place of mobile monitor is needed in the transformer station of unattended duty, the remote monitoring etc. of machine room, staff not relevant is at ordinary times on duty, and the operation maintenance personnel in the controlling heart is carried out movable rail video camera in operation supervise and control place by Long-distance Control cradle head device and carried out daily patrolling and examining work to equipment.Such as, crusing robot in transformer station needs to patrol and examine in transformer station, patrol and examine in process, need to put specific the collection picture that stops, due to the reason of navigation accuracy, the parking error of robot can be excessive, causing pictures taken to depart from, in order to improve the alignment precision of The Cloud Terrace, needing to correct The Cloud Terrace.Traditional bearing calibration mostly adopts mechanical Correction Strategies, such as application number is the patent document of 201410694969, this patent document discloses a kind of posture self-correcting monopod video camera control device and method thereof, and specifically disclose sensor technology such as employing accelerometer and gyroscope etc. to detect the attitudes vibration of orbital platform and monopod video camera, according to the actual travel distance of institute's test pose change calculations orbital platform and monopod video camera deflection angle and luffing angle, calculating machine error is carried out again by carrying out contrast with preset value, and according to this error, The Cloud Terrace position is revised.Visible, need additional sensor to realize in this file, both added cost, correction accuracy may be caused not high due to the error of additional sensor itself again.

Summary of the invention

The object of this invention is to provide a kind of The Cloud Terrace bearing calibration, to improve the alignment precision of The Cloud Terrace, to reduce target biased error during pictures taken.

The present invention provides a kind of The Cloud Terrace bearing calibration for solving the problems of the technologies described above, and the step of this bearing calibration is as follows:

1) gather the picture of a certain preset position as template picture, and record position angle now, the angle of pitch and camera zoom multiple;

2) modeling is carried out to template picture, centered by the central point of template picture, extract the profile of target object;

3) preset position when gathering with template picture arranges The Cloud Terrace and camera carries out picture collection, the picture of collection and template picture is compared, and calculates the offset pixels of target object;

4) calculate The Cloud Terrace deviation angle according to obtaining offset pixels, and will obtain The Cloud Terrace deviation angle and be added on cradle head preset positions, obtain the absolute angle after correction.

Described step 4) computation process of The Cloud Terrace deviation angle is as follows:

A. the physical deflection of target object is calculated according to the offset pixels of target object,

d _x＝c _x×ρ

d _y＝c _y×ρ

Wherein, d _xand d _yfor the physical deflection of target object, c _xand c _yfor the offset pixels of target object, ρ is the Pixel Dimensions of video camera used;

B. its current focal distance f is calculated by video camera current zoom multiple k,

f＝kf ₁

Wherein f ₁for the minimum focus of video camera;

C. according to the physical deflection of target object and the current focal length of video camera, calculate the deviation angle of The Cloud Terrace, comprise orientation angles deviation delta θ and luffing angle deviation delta gamma,

$\begin{array}{c}\mathrm{\Δ}\mathrm{\θ}=arctan\left(\frac{d_x}{f}\right)\\ \mathrm{\Δ}\mathrm{\γ}=arctan\left(\frac{d_y}{f}\right)\end{array}.$

Described template picture gathers under Large visual angle.

The offset pixels c of described target object _xand c _ycomputation process be: by conspicuousness property calculation, analyze candidate region, adopt local auto-adaptive kernel regression calculate matching area, target's center position offset pixels c can be obtained _xand c _y.

The invention has the beneficial effects as follows: first the present invention gathers the picture of a certain preset position as template picture, and record position angle now, the angle of pitch and camera zoom multiple; Then modeling is carried out to template picture, centered by the central point of template picture, extract the profile of target object; Preset position when gathering with template picture again arranges The Cloud Terrace and camera carries out picture collection, the picture of collection and template picture is compared, and calculates the offset pixels of target object; Last basis obtains offset pixels and calculates The Cloud Terrace deviation angle, and will obtain The Cloud Terrace deviation angle and be added on cradle head preset positions, obtains the absolute angle after correction.The present invention adopts machine vision technique to identify impact point, can rapid adjustment The Cloud Terrace attitude, without the need to additional sensor, improves the alignment precision of The Cloud Terrace, reduces target biased error during pictures taken.The The Cloud Terrace that the present invention is not only applicable to Intelligent Mobile Robot field corrects, and the The Cloud Terrace being applicable to other field too corrects.

Accompanying drawing explanation

Fig. 1 is template picture and the contrast schematic diagram gathering picture in the embodiment of the present invention;

Fig. 2 is video camera pin-hole imaging model schematic;

Fig. 3 is the principle schematic of The Cloud Terrace bearing calibration of the present invention.

Embodiment

Below in conjunction with accompanying drawing, the specific embodiment of the present invention is described further.

The present invention adopts machine vision technique to identify impact point, by gathering the picture of a certain preset position as template picture, and records position angle now, the angle of pitch and camera zoom multiple; Modeling is carried out to template picture, centered by the central point of template picture, extracts the profile of target object; Preset position when gathering with template picture arranges The Cloud Terrace and camera carries out picture collection, the picture of collection and template picture is compared, and calculates the offset pixels of target object; Calculate The Cloud Terrace deviation angle according to obtaining offset pixels, and will obtain The Cloud Terrace deviation angle and be added on cradle head preset positions, obtain the absolute angle after correction.Be described in detail for the The Cloud Terrace trimming process of Intelligent Mobile Robot below, concrete implementation step is as follows:

1. first collect the picture of a certain preset position as template, graphic template gathers under Large visual angle, records azimuth angle theta now _twith angle of pitch γ _t, azimuth angle theta _t, angle of pitch γ _tpreserve as presetting bit with camera zoom multiple k.Here preset position refers to specific one point, namely the particular orientation angle of pitch.

2. then modeling is carried out to this template, centered by target's center's point, extract the profile of target object, then preserve template.When robot fully automatic working, presetting bit when gathering with template arranges The Cloud Terrace and camera, carries out picture collection.By the picture of collection and template matching, as shown in Figure 1, identify the offset pixels c of target object _xand c _y.

Offset pixels c _xand c _ycomputation process is as follows: first by conspicuousness property calculation, analyzes candidate region, then adopts the kernel regression of local auto-adaptive to calculate matching area, can obtain target's center position offset pixels c _xand c _y.This process is prior art, no longer describes in detail here to its detailed process.

3. utilize the offset pixels c of object in picture _xand c _ycalculate azimuth pitch error angle Δ θ and the Δ γ of The Cloud Terrace.Concrete method is: demarcate used video camera, calibrates accurate Pixel Dimensions ρ.By offset pixels c _xand c _yphysical deflection d can be calculated _xand d _y:

d _x＝c _x×ρ

(1)

d _y＝c _y×ρ

Calculated by current zoom multiple k and work as front focal length: f=kf ₁, wherein f ₁for the minimum focus of video camera.As shown in Figures 2 and 3, The Cloud Terrace deviation angle can be calculated:

$\begin{array}{c}\mathrm{\Δ}\mathrm{\θ}=arctan\left(\frac{d_x}{f}\right)\\ \mathrm{\Δ}\mathrm{\γ}=arctan\left(\frac{d_y}{f}\right)\end{array}---\left(2\right)$

Angular deviation Δ θ and Δ γ is added in presetting bit, obtains the absolute angle after correcting.

The present invention identifies impact point by adopting machine vision technique, can rapid adjustment The Cloud Terrace attitude, achieve the correction to The Cloud Terrace, the The Cloud Terrace that the present invention is not only applicable to Intelligent Mobile Robot field corrects, and the The Cloud Terrace being applicable to other field too corrects.

Claims (4)

1. a The Cloud Terrace bearing calibration, is characterized in that, the step of this bearing calibration is as follows:

1) gather the picture of a certain preset position as template picture, and record position angle now, the angle of pitch and camera zoom multiple;

2) modeling is carried out to template picture, centered by the central point of template picture, extract the profile of target object;

3) preset position when gathering with template picture arranges The Cloud Terrace and camera carries out picture collection, the picture of collection and template picture is compared, and calculates the offset pixels of target object;

4) calculate The Cloud Terrace deviation angle according to obtaining offset pixels, and will obtain The Cloud Terrace deviation angle and be added on cradle head preset positions, obtain the absolute angle after correction.

2. The Cloud Terrace bearing calibration according to claim 1, is characterized in that, described step 4) computation process of The Cloud Terrace deviation angle is as follows:

A. the physical deflection of target object is calculated according to the offset pixels of target object,

d _x＝c _x×ρ

d _y＝c _y×ρ

Wherein, d _xand d _yfor the physical deflection of target object, c _xand c _yfor the offset pixels of target object, ρ is the Pixel Dimensions of video camera used;

B. its current focal distance f is calculated by video camera current zoom multiple k,

f＝kf ₁

Wherein f ₁for the minimum focus of video camera;

C. according to the physical deflection of target object and the current focal length of video camera, calculate the deviation angle of The Cloud Terrace, comprise orientation angles deviation delta θ and luffing angle deviation delta gamma,

$\begin{array}{c}\mathrm{\Δ}\mathrm{\θ}=arctan\left(\frac{d_x}{f}\right)\\ \mathrm{\Δ}\mathrm{\γ}=arctan\left(\frac{d_y}{f}\right)\end{array}.$

3. The Cloud Terrace bearing calibration according to claim 2, is characterized in that, described template picture gathers under Large visual angle.

4. The Cloud Terrace bearing calibration according to claim 3, is characterized in that, the offset pixels c of described target object _xand c _ycomputation process be: by conspicuousness property calculation, analyze candidate region, adopt local auto-adaptive kernel regression calculate matching area, target's center position offset pixels c can be obtained _xand c _y.