阅读说明：本技术 利用多线结构光测量间隙面差的方法 (Utilize the method for multiple line structure light measurement clearance surface difference ) 是由 谢康康 冯伟昌 郭寅 郭磊 于 2019-10-09 设计创作，主要内容包括：本发明公开了一种利用多线结构光测量间隙面差的方法,包括：相机采集多条线结构光被调制所形成的结构光图像；得到光条图像,对光条图像进行轮廓分割,将每个左/右光条轮廓中的所有点记为一个单侧点集,筛选每个单侧点集所对应的特征点；利用间隙单侧的所有特征点进行直线拟合,作为基准线,求取间隙另一侧所有特征点到基准线的距离,取平均后作为间隙值；利用间隙一侧所有单侧点集内的点拟合平面,作为基准面,求取间隙另一侧所有特征点到基准面的距离,取平均后作为面差值,本方法能够一次性获得更多的被测物表面点云信息,同时,本方法提出了一种快速、高效的特征点寻找方法,保证了整个测量过程的高效性、测量结果的准确性。(The invention discloses a kind of methods using multiple line structure light measurement clearance surface difference, comprising: a plurality of line-structured light of camera acquisition, which is modulated, is formed by structure light image；Optical strip image is obtained, contours segmentation is carried out to optical strip image, all the points in each left/right striation profile are denoted as a unilateral point set, screen characteristic point corresponding to each unilateral point set；Straight line fitting is carried out using all characteristic points of gap unilateral side, as reference line, all characteristic points in the gap other side is sought to the distance of reference line, gap width is used as after being averaged；Utilize the point fit Plane in all unilateral point sets in gap side, as datum level, seek all characteristic points in the gap other side to datum level distance, face difference is used as after being averaged, this method can disposably obtain more measured object surfaces point cloud information, meanwhile method proposes a kind of fast and efficiently characteristic point finding method, the high efficiency of entire measurement process, the accuracy of measurement result ensure that.)

1. a kind of method using multiple line structure light measurement clearance surface difference, which comprises the steps of:

1) multiple line structure light projector projects a plurality of line-structured light being parallel to each other to measured object surface, and camera acquires a plurality of knot Structure light, which is modulated, is formed by structure light image；The a plurality of line-structured light at least three；

2) to the structure light image carry out Light stripes center extraction, image procossing, obtain optical strip image, to the optical strip image into Row contours segmentation is labeled as left/right striation profile according to the relative position in gap, by the institute in each left/right striation profile It is denoted as a unilateral point set a little, starting endpoint wherein will be labeled as away from interstitial site farthest point for single unilateral point set；

3) it chooses any unilateral point set and first straight line is labeled as to all the points fitting a straight line；

4) it is concentrated in the unilateral point, the point within the scope of step-length t is screened since starting endpoint, point set I is labeled as, utilizes point set I fitting a straight line is labeled as second straight line, seeks the angle theta between the first and second straight line；It counts the unilateral point and concentrates rejecting point The quantity for collecting left point after point in I, is labeled as m；

If angle theta is more than or equal to preset angle, filtered out newly for new step-length with the above method with the step-length of last computation plus h Point set I in point, again be fitted second straight line, seek new angle theta and quantity m；It so repeats, until θ is less than preset angle； Wherein,

If angle theta is less than preset angle, the point in this point set I near interstitial site is denoted as characteristic point；

5) same to step 3), 4) method handle remaining unilateral point set, until obtaining spy corresponding to each unilateral point set Sign point；

6) straight line fitting is carried out as reference line using all characteristic points of gap unilateral side seek all features in the gap other side Point arrives the distance of reference line, and gap width is used as after being averaged；

It seeks all characteristic points in the gap other side as datum level using the point fit Plane in the point set of gap side unilateral side and arrives The distance of the datum level is used as face difference after being averaged.

2. a kind of method using multiple line structure light measurement clearance surface difference, it is characterised in that: angle theta will be obtained for the first time in step 4) Corresponding step-length and quantity m are respectively labeled as t when less than preset angle_JustAnd m_Just；It is handled as follows:

With t_JustH is subtracted as new step-length, screens the point within the scope of new step-length since the starting point of unilateral point set, fitting a straight line, Ask whether its angle theta between first straight line is more than or equal to preset angle；Wherein,

If so, being labeled as characteristic point away from the nearest point in gap in point used when by this fitting a straight line；

If it is not, the quantity that unilateral point concentrates left point after point used when rejecting this fitting a straight line is then counted, labeled as new m_Just, Correspondence seeks new h value, with t_JustNew h is subtracted as new step-length, using same procedure reconnaissance, fitting a straight line, then seeks angle theta, It so repeats, until angle theta is more than or equal to preset value again；Point nearest away from gap in point used when by this fitting a straight line Labeled as characteristic point.

3. utilizing the method for multiple line structure light measurement clearance surface difference as claimed in claim 1 or 2, it is characterised in that: described initial It always counts in the unilateral point set of the value of step-length tIt arrives

4. utilizing the method for multiple line structure light measurement clearance surface difference as claimed in claim 1 or 2, it is characterised in that: in step 2) Image procossing includes: image binaryzation, image filtering and three-dimensionalreconstruction, and the three-D profile after reconstruct is transformed into optical plane coordinate System, obtains the optical strip image under optical plane coordinate system.

5. utilizing the method for multiple line structure light measurement clearance surface difference as claimed in claim 1 or 2, it is characterised in that: step 2) is right The optical strip image carries out contours segmentation, in the following way:

Using certain end point in optical strip image as starting point, search range d is set, is scanned for along striation direction, it will be in the d of search range The point in addition to starting point be denoted as similar point, the similar point farthest away from starting point along striation direction that will be searched, as new starting point, It continues searching, until no longer occurring similar point in the d of search range；The point not in search range is rejected, left and right striation wheel is obtained It is wide.

6. utilizing the method for multiple line structure light measurement clearance surface difference as claimed in claim 1 or 2, it is characterised in that: step 2) is right The optical strip image carries out contours segmentation, in the following way:

Using certain end point in optical strip image as starting point, along striation direction, search range d is set, starting point will be removed in the d of search range Outer point is denoted as similar point, and the similar point farthest away from starting point along striation direction searched is continued to search as new starting point Rope, until no longer occurring similar point in the d of search range；Counting includes the sum put in each secondary search range d, and it is low to delete sum The point corresponding to the search range of preset value；Plain range is searched for new with d1, is scanned for from the off, by search range The point in addition to starting point in d1 is denoted as similar point, is risen away from the farthest similar point of starting point as new using what is searched along striation direction Point continues searching, until no longer occurring similar point in the d1 of search range；Based on left point position be respectively labeled as it is left/ Right striation profile；Wherein, d < d1 < 1.5d.

7. utilizing the method for multiple line structure light measurement clearance surface difference as described in claim 5 or 6, it is characterised in that:Wherein, T=2~5.

8. utilizing the method for multiple line structure light measurement clearance surface difference as described in any one of claim 5~7, it is characterised in that: Judge that the quantity summation put in all unilateral point sets and the quantity put in each unilateral point set account for optical strip image midpoint respectively Whether the ratio of quantity summation is in preset range simultaneously, if so, contours segmentation result is effective, if it is not, then needing to re-start Contours segmentation.

9. utilizing the method for multiple line structure light measurement clearance surface difference as described in claim 1, it is characterised in that: the fitting a straight line Method are as follows: Ransac algorithm or least square method.

Technical field

The present invention relates to field of visual inspection, and in particular to a method of utilize multiple line structure light measurement clearance surface difference.

Background technique

Gap, face difference are the common forms in industrial processes, processing dimension accurately whether, be related to the beauty of entire workpiece The property seen, leakproofness, are a highly important parameters in quality monitoring, such as during automobile production, covering scale error The larger face mistake difference that will lead to vehicle body is larger, to influence the appearance of vehicle, wind-dryness, sealing etc..

About gap, the measurement of face difference, traditional artificial detection method is mainly measured using clearance ruler, face difference ruler, This method low efficiency, precision are low and are easy to be influenced by the subjective factor of survey crew, and it is high-precision to be unable to satisfy the high beat of high efficiency The measurement demand of degree；With the development of Computer Vision Detection, measured based on line-structured light recovery of vision three-D profile Technology receives more and more attention, this mode, can more efficiently, accurately obtain tested object plane difference feature, when measurement, Line structure optical sensor is made of line-structured light projector and camera, in active incident line structure light to testee, passes through knot The deformation (protrusion, recess, fluctuation etc.) of structure light determines the dimensional parameters of measured object, and in gap, face difference image, striation is usual It will appear protrusion and disconnect (measured object surface has both face difference and clearance features) situation；Based on above-mentioned characteristics of image, structure is found Characteristic point in light image, the left and right profile for being accurately and effectively partitioned into striation are that entire clearance surface difference measurement process is crucial Step, but in existing method, there is no largely effective, quick Feature Points Extraction.

Summary of the invention

In view of the above-mentioned problems, the present invention proposes a kind of method using multiple line structure light measurement clearance surface difference, multiple line structure Light can disposably obtain more measured object surfaces point cloud information, meanwhile, this method be based on multi-line structured light the characteristics of, propose A kind of fast and efficiently characteristic point finding method, ensure that the high efficiency of entire measurement process, the accuracy of measurement result.

A method of using multiple line structure light measurement clearance surface difference, include the following steps:

1) multiple line structure light projector projects a plurality of line-structured light being parallel to each other to measured object surface, and camera acquisition is a plurality of Line-structured light, which is modulated, is formed by structure light image；The a plurality of line-structured light at least three；

2) Light stripes center extraction, image procossing are carried out to the structure light image, optical strip image is obtained, to the striation figure As carrying out contours segmentation, it is labeled as left/right striation profile according to the relative position in gap, it will be in each left/right striation profile All the points be denoted as a unilateral point set, will be wherein away from the farthest point of interstitial site labeled as starting point for single unilateral point set Point；

3) it chooses any unilateral point set and first straight line is labeled as to all the points fitting a straight line；

4) it is concentrated in the unilateral point, the point within the scope of step-length t is screened since starting endpoint, is labeled as point set I, is utilized Point set I fitting a straight line is labeled as second straight line, seeks the angle theta between the first and second straight line；The unilateral point concentration is counted to pick Except the quantity of left point after point in point set I, it is labeled as m；

If angle theta is more than or equal to preset angle, with the step-length of last computation plus h for new step-length, screened with the above method Point in new point set I out is fitted second straight line again, seeks new angle theta and quantity m；It so repeats, until θ is less than in advance If angle；Wherein,

If angle theta is less than preset angle, the point in this point set I near interstitial site is denoted as characteristic point；

5) same to step 3), 4) method handle remaining unilateral point set, until obtaining corresponding to each unilateral point set Characteristic point；

6) straight line fitting is carried out using all characteristic points of gap unilateral side, as reference line, it is all seeks the gap other side Characteristic point is used as gap width to the distance of reference line after being averaged；

All features in the gap other side are sought as datum level using the point fit Plane in the point set of gap side unilateral side Point arrives the distance of the datum level, and face difference is used as after being averaged.

Further, step-length and quantity m corresponding when obtaining angle theta in step 4) for the first time less than preset angle are marked respectively For t_JustAnd m_Just；It is handled as follows:

With t_JustH is subtracted as new step-length, the point within the scope of new step-length is screened since the starting point of unilateral point set, fitting is straight Line, asks whether its angle theta between first straight line is more than or equal to preset angle；Wherein,

If so, being labeled as characteristic point away from the nearest point in gap in point used when by this fitting a straight line；

If it is not, the quantity that unilateral point concentrates left point after point used when rejecting this fitting a straight line is then counted, labeled as new M_Just, correspondence seeks new h value, with t_JustNew h is subtracted as new step-length, using same procedure reconnaissance, fitting a straight line, then seeks pressing from both sides Angle θ is so repeated, until angle theta is more than or equal to preset value again；It is nearest away from gap in point used when by this fitting a straight line Point be labeled as characteristic point.

Further, it always counts in the unilateral point set of the value of the initial step length tIt arrives

Further, image procossing includes: image binaryzation, image filtering and three-dimensionalreconstruction in step 2), after reconstruct Three-D profile is transformed into optical plane coordinate system, obtains the optical strip image under optical plane coordinate system.

Further, step 2) carries out contours segmentation to the optical strip image, in the following way:

Using certain end point in optical strip image as starting point, search range d is set, is scanned for along striation direction, model will be searched for The point in addition to starting point enclosed in d is denoted as similar point, the similar point farthest away from starting point along striation direction that will be searched, as new Starting point continues searching, until no longer occurring similar point in the d of search range；The point not in search range is rejected, is obtained left and right Striation profile.

The above process can effectively inhibit the interference of image noise, and striation profile is broken in order to prevent, be searched again for；

Further, step 2) carries out contours segmentation to the optical strip image, in the following way:

Using certain end point in optical strip image as starting point, along striation direction, search range d is set, by removing in the d of search range Point outside starting point is denoted as similar point, and the similar point farthest away from starting point along striation direction that will be searched continues as new starting point Search, until no longer occurring similar point in the d of search range；Counting includes the sum put in each secondary search range d, deletes sum Point corresponding to search range lower than preset value；Plain range is searched for new with d1, scans for from the off, model will be searched for The point in addition to starting point enclosed in d1 is denoted as similar point, to search along striation direction away from starting point farthest similar point as newly Starting point continues searching, until no longer occurring similar point in the d1 of search range；It is respectively labeled as based on left point position Left/right striation profile；Wherein, d < d1 < 1.5d.

Further,Wherein, T=2~5.

Further, the quantity difference for judging the quantity summation put in all unilateral point sets and being put in each unilateral point set Account for the quantity summation at optical strip image midpoint ratio whether and meanwhile in preset range, if so, contours segmentation result is effective, if It is no, then it needs to re-start contours segmentation.

Further, the method for the fitting a straight line are as follows: Ransac algorithm or least square method.

This method can be primary compared to single line structural light measurement using multiple line structure light collection measured object surface image Property obtain more measured object surfaces point cloud information, be suitable for more measurement scenes, meanwhile, this method is based on multi-line structured light Measuring characteristic, propose a kind of fast and efficiently characteristic point finding method, the left and right profile in structure light image carried out Effectively segmentation, and then ensure that the high efficiency of entire measurement process, the accuracy of measurement result.

Detailed description of the invention

Fig. 1 is image acquisition process schematic diagram in embodiment 1；

Fig. 2 is 1 middle left and right striation outline drawing of embodiment；

Fig. 3 is the schematic diagram that multiple unilateral point sets are partitioned into embodiment 1；

Fig. 4 is in embodiment 1 with local two contours segmentation process schematic of striation example ways of carrying out；

Fig. 5 is to seek characteristic point process schematic in embodiment 1；

Fig. 6 is calculated gap process schematic in embodiment 1；

Fig. 7 is other form optical strip image schematic diagrames that the present invention can apply；

Fig. 8 is the schematic diagram for other form striation unilateral side point sets that the present invention can apply.

Specific embodiment

Technical solution of the present invention is described in detail below in conjunction with drawings and examples.