Summary of the invention
The object of the present invention is to provide a kind of 360 degree of on-line automatic calibration systems of viewing system, to solve the problem proposed in above-mentioned background technology.
For achieving the above object, the invention provides following technical scheme:
A kind of 360 degree of on-line automatic calibration systems of viewing system, a basic premise of automatic Calibration is that the position of the camera that a kind of vehicle is installed is consistent with angle, difference source between the camera of this vehicle different vehicle is only alignment error, during automatic Calibration, there is one to look around ECU in each car, look around in ECU and run automatic Calibration algorithm.
As the further scheme of the present invention: the automatic Calibration process of described automatic calibration system comprises the following steps:
(1) first on the ground in automatic Calibration place, lay manual calibration point, vehicle is sailed into the region at manual calibration point place, single-view is carried out to this car and manually demarcates, generate standard scale and be downloaded to and look around ECU in each car;
(2) then on the ground in automatic Calibration place, lay automatic Calibration point, vehicle is sailed into the region at automatic Calibration point place, take pictures, start automatic Calibration calculating process, its concrete operation process comprises:
2.1) use the standard scale of looking around in ECU to act on the image photographed, generate 4 width vertical views;
2.2) use and automatically get an algorithm 4 width vertical views are got a little automatically, automatically get an algorithm and automatically can detect the automatic Calibration point in vertical view and automatic Calibration point is numbered;
2.3) the automatic Calibration point that utilization obtains adjusts looking around existing standard scale in ECU, and generates the table after adjustment, to compensate camera alignment error;
(3) table after adjustment shows that automatic Calibration process terminates after generating.
As the further scheme of the present invention: described concrete steps of automatically getting an algorithm are as follows:
(1) detect angle points all in 4 width vertical views, use Harris Corner Detection Algorithm herein;
(2) neighborhood image Block-matching template and the neighborhood loop wire matching template of 4 width vertical views is generated;
(3) the neighborhood image block of all angle points in 4 width vertical views is analyzed, first Threshold segmentation is carried out to neighborhood image, morphology opening operation is done to the image block after Threshold segmentation;
(4) loop wire statistics is carried out to image block after process;
(5) carry out template matches and loop wire coupling to image block after process, the position occurred in conjunction with calibration point in matching value and 4 width vertical views is screened;
(6) qualified angle point is merged;
(7) ineligible angle point is removed according to the line information of calibration point.
Compared with prior art, the invention has the beneficial effects as follows: the first, except first time is demarcated and needs artificial participation, after the demarcation of all vehicles of this vehicle all manually do not participate in, also i.e. a kind of vehicle for once manual calibration process.The second, except first time demarcate be that single-view is demarcated, after all demarcation be all that vertical view is demarcated.Demarcation is carried out to vertical view and has 2 advantages: the first in a top view, calibration graph, less than distortion, is got a ratio automatically and is easier to by vertical view, generally can not undetected and many inspections; It two is that to carry out Adjustable calculation amount to standard scale very little, and this computation process can be run looking around on ECU, and can not be consuming time longer.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, be clearly and completely described the technical scheme in the embodiment of the present invention, obviously, described embodiment is only the present invention's part embodiment, instead of whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art, not making the every other embodiment obtained under creative work prerequisite, belong to the scope of protection of the invention.
In the embodiment of the present invention, a kind of 360 degree of on-line automatic calibration systems of viewing system, a basic premise of automatic Calibration is that the position of the camera that a kind of vehicle is installed and angle are basically identical, difference source between the camera of this vehicle different vehicle is only alignment error, during automatic Calibration, need an automatic Calibration place, have one to look around ECU in each car, look around in ECU and run automatic Calibration algorithm.
The automatic Calibration process of described automatic calibration system comprises the following steps:
(1) first on the ground in automatic Calibration place, lay manual calibration point 2, vehicle is sailed into the region at manual calibration point 2 place, single-view is carried out to this car and manually demarcates, generate standard scale and be downloaded to and look around ECU in each car;
(2) then on the ground in automatic Calibration place, lay automatic Calibration point 1, vehicle is sailed into the region at automatic Calibration point 1 place, take pictures, start automatic Calibration calculating process, its concrete operation process comprises:
2.1) use the standard scale of looking around in ECU to act on the image photographed, generate 4 width vertical views;
2.2) use and automatically get an algorithm and point is got automatically to 4 width vertical views (automatically get an algorithm and mainly comprise Corner Detection, neighborhood image is split, morphology operations, template matches, the image processing algorithms such as neighborhood image loop wire analysis), automatically get an algorithm and automatically can detect the automatic Calibration point 1 in vertical view and automatic Calibration point 1 is numbered;
2.3) the automatic Calibration point 1 that utilization obtains adjusts looking around existing standard scale in ECU, and generates the table after adjustment, to compensate camera alignment error;
(3) table after adjustment shows that automatic Calibration process terminates after generating.
In above-mentioned automatic Calibration process, automatically getting an algorithm is a crucial step, automatically gets an algorithm and is related to automatic Calibration process success, and described concrete steps of automatically getting an algorithm are as follows:
(1) detect angle points all in 4 width vertical views, use Harris Corner Detection Algorithm herein;
(2) neighborhood image Block-matching template and the neighborhood loop wire matching template of 4 width vertical views is generated;
(3) the neighborhood image block of all angle points in 4 width vertical views is analyzed, first Threshold segmentation is carried out to neighborhood image, morphology opening operation is done to the image block after Threshold segmentation;
(4) loop wire statistics is carried out to image block after process;
(5) carry out template matches and loop wire coupling to image block after process, the position occurred in conjunction with calibration point in matching value and 4 width vertical views is screened;
(6) qualified angle point is merged;
(7) ineligible angle point is removed according to the line information of calibration point.
After obtaining a width vertical view, first need to carry out Corner Detection to this vertical view.The reason the doing Corner Detection angle point that to be calibration point must be in image.Carry out Corner Detection and reject the angle point not belonging to calibration point calibration point to be found out.Here Corner Detection adopts Harris Corner Detection Algorithm.The ultimate principle of Harris Corner Detection Algorithm is: the structure tensor matrix obtaining each pixel, if this matrix is A, then the C value of this pixel can utilize following formula to calculate:
C=det(A)-0.04·tr(A)
When the C value of this pixel is greater than 10
4time, then think that this point is angle point.
After Corner Detection is carried out to vertical view, angle points all in vertical view can be obtained.At this moment need in these angle points, find out calibration point and reject non-calibration point.Observation calibration point and neighborhood image thereof can find, the feature of its neighborhood image is that certain pair of horns direction presents black, and another presents white to angular direction, or on the contrary.So proceed as follows the neighborhood image of each angle point: 1. cross entropy Threshold segmentation; 2. morphological erosion; 3. morphological dilations; 4. template matches; 5. determine neighborhood loop wire matching way according to template matching results, these operations go out with black dotted lines frame in flow charts.Introduce the operation of this 5 step below in detail.Cross entropy Threshold segmentation is a kind of Threshold segmentation mode, by the cross entropy of image after computed segmentation and original image, when after segmentation, image is minimum with the cross entropy of original image, corresponding threshold value is segmentation threshold, utilizes segmentation threshold to carry out segmentation to neighborhood image and obtains splitting rear image.After segmentation, image is bianry image, and because original image (vertical view) exists the problems such as uneven illumination is even, the image after segmentation presents a lot of burr, for eliminating these burrs, can utilize morphological operation.Here first morphological erosion is carried out to it, and then carry out morphological dilations elimination burr.The ultimate principle of morphology operations utilizes a structural element traversing graph picture, carries out min/max Value Operations determine result of calculation to the element value in this structural element.Here the size of structural element is 3 × 3.After carrying out morphological operation, after segmentation, image presents comparatively regular result.Now carry out template matches, template is the bianry image generated in advance, in order to increase the robustness of algorithm, working together and first generating 6 two-value templates, adopts these 6 two-value templates to mate with image after segmentation respectively.After obtaining matching result, determine neighborhood loop wire matching way according to this matching result.Adjacent namely around a circle of neighborhood image central point.Each neighborhood loop wire of image after segmentation is all carried out neighborhood loop wire coupling, and record matching result.During neighborhood loop wire coupling, be also mate with the template generated in advance.Here illustratively template matches and neighborhood loop wire coupling similarity and difference.The essence of template matches and neighborhood loop wire coupling is all that element to be matched and element generated are in advance carried out comparison one by one, and obtain comparison value, difference is that template matches is the coupling between two dimensional image, and neighborhood loop wire coupling is the coupling between one-dimension array.
After aforesaid operations is carried out to the neighborhood image of each angle point, below just can screen angle point according to line information and matching value.When template matches value and neighborhood loop wire matching value meet the feature of this calibration point, this angle point can be retained.Most of non-calibration point angle point can be rejected in this way.Carry out angle point merging below.Because the C value of the point around calibration point is all greater than 10
4, so can form " cluster " angle point at calibration point place, the object that angle point merges is that " cluster " angle point is merged into an angle point.The criterion that angle point merges utilizes the ranks of template matches value to this angle point of this angle point to be weighted the ranks value finally being merged angle point.Finally need further to screen angle point.Before and after observing, the calibration point of vertical view, can find that calibration point is on 2 row, observes the calibration point of left and right vertical view, can find that calibration point is on 2 row.Calibration point away from this 2 row or 2 row can be rejected according to this feature.After above-mentioned steps, can obtain final output image, this output image only comprises the calibration point of our needs.
To those skilled in the art, obviously the invention is not restricted to the details of above-mentioned one exemplary embodiment, and when not deviating from spirit of the present invention or essential characteristic, the present invention can be realized in other specific forms.Therefore, no matter from which point, all should embodiment be regarded as exemplary, and be nonrestrictive, scope of the present invention is limited by claims instead of above-mentioned explanation, and all changes be therefore intended in the implication of the equivalency by dropping on claim and scope are included in the present invention.Any Reference numeral in claim should be considered as the claim involved by limiting.
In addition, be to be understood that, although this instructions is described according to embodiment, but not each embodiment only comprises an independently technical scheme, this narrating mode of instructions is only for clarity sake, those skilled in the art should by instructions integrally, and the technical scheme in each embodiment also through appropriately combined, can form other embodiments that it will be appreciated by those skilled in the art that.