CN110375645B

CN110375645B - Character size detection method and device

Info

Publication number: CN110375645B
Application number: CN201910600777.4A
Authority: CN
Inventors: 张明勇
Original assignee: Chengdu Juying Intelligent Technology Co ltd
Current assignee: Chengdu Juying Intelligent Technology Co ltd
Priority date: 2019-07-04
Filing date: 2019-07-04
Publication date: 2021-09-28
Anticipated expiration: 2039-07-04
Also published as: CN110375645A

Abstract

The invention discloses a method and a device for detecting character size, wherein the method comprises the following steps: acquiring 3D contour data of the surface of the engraved character; calculating data of a geometric surface where the character is located from the contour data; extracting point cloud data of the lettering characters with depth or height information from the contour data; extracting single characters, and counting the geometric size of each character; and outputting the result. The device comprises a 3D contour measuring unit, a guide rail fixing unit base, a motor driving unit, a displacement sensor and a processing unit; the invention can reduce the manual detection cost of the VIN code, the engine number and the like of the motor vehicle in the field of motor vehicle manufacturing and inspection by carrying out full-automatic detection on the depth or height of the code-engraved character, the size of the character and the distance between adjacent characters, improves the detection precision and the detection efficiency and has important significance for promoting the detection automation of the industry.

Description

Character size detection method and device

Technical Field

The invention relates to the technical field of character size detection, in particular to a character size detection method and device.

Background

The traditional VIN code and engine number of the motor vehicle are detected mainly by manual visual observation and are measured by a dial indicator and a ruler. Typically, many vehicle enterprises use dial gauges with probes to detect the depth of VIN codes and engine numbers. Because the tip of the probe has a certain thickness, the depth of the character ravine details cannot be detected by the probe, and the detection result is seriously influenced by the manual pressing force. In a word, the traditional detection method has the problems of high labor cost, low efficiency, poor quantization precision, subjectivity, easiness in omission and the like. Although a few inventions make a certain technical improvement on the problem, the method is still manual in nature, and the detection of one nick code requires several minutes of time, is time-consuming and labor-consuming, and does not fundamentally change the current situation. With the increase of automobile holding amount and sales volume, the workload of automobile manufacturers and annual inspection mechanisms for detecting VIN codes and engine numbers also increases rapidly, and the traditional manual detection mode cannot meet the working requirements.

Disclosure of Invention

In order to solve the problems in the prior art, the invention provides a character size detection method and a device, and the method and the device can reduce the manual detection cost of the VIN code, the engine number and the like of the motor vehicle in the field of motor vehicle manufacturing and inspection by carrying out full-automatic detection on the depth or height of the code-engraved character, the size of the character and the distance between adjacent characters, improve the detection precision and the detection efficiency, and have important significance for promoting the detection automation of the industry.

In order to achieve the purpose, the invention adopts the following technical scheme:

a method and apparatus for character size detection, comprising the steps of:

s1, acquiring 3D contour data of the surface of the code carving character;

s2, calculating the data of the geometric surface where the character is located from the outline data;

s3, extracting point cloud data of the lettering characters with depth or height information from the outline data;

s4, extracting single characters, and counting the geometric size of each character;

and S5, outputting the result.

Preferably, the step S2 further includes the steps of:

s21, establishing a fitting polynomial curve model for each contour curve;

s22, solving a fitting polynomial curve parameter model of the contour curve through a weighted polynomial curve fitting algorithm or a RANSAC algorithm;

and S23, outputting the fitted polynomial profile curve parameters.

Preferably, in step S22, the step of solving the fitted polynomial curve parameter model of the profile curve by using the weighted polynomial curve fitting algorithm includes the steps of:

s2221, smoothing the profile curve, and removing certain noise and abnormal points;

s2222, fitting a polynomial curve;

s2223, counting the number of the residual points falling on the fitting contour curve on the original contour curve;

s2224, judging whether the number of the remaining points on the fitting contour curve reaches the maximum, if so, entering S2226, and if not, entering S2225;

s2225, removing points outside the fitted hub curve, and returning to the step S2222;

and S2226, outputting the fitted polynomial profile curve parameters.

Preferably, the step S3 further includes the steps of:

s31, calculating the distance from each point on the scanning contour curve to the fitting polynomial curve of the contour curve;

s32, screening out character point cloud data according to the concave-convex characteristics of the characters and the preset depth/height range of the characters;

and S33, outputting the character point cloud data.

Preferably, the following steps are further included between the steps S32 and S33:

s321, judging whether the character surface is a curved surface, if so, going to step S322, and if not, going to step S33;

and S322, converting the character point cloud coordinates on the curved surface into character point cloud data on the plane.

Preferably, the step S4 includes the steps of:

s41, performing binarization processing on the point cloud data to generate a binarized character image;

and S42, extracting the characters from the binary character image by adopting a projection method or a character recognition algorithm, and counting the geometric size of each character.

Preferably, the step S41 is followed by the following steps:

s411, conducting connected domain calculation on the binary character image, and removing isolated points and small-area connected domains.

Preferably, the projection method in step S42 includes the following steps:

s421, projecting the binary image of the character along the x axis and the y axis respectively, namely accumulating the pixel values of the binary image to obtain 2 projection images;

and S422, calculating the boundary of each character along the x axis and the y axis on the 2 projection images by using a threshold segmentation method.

Preferably, the step of counting the geometric size of each character in the step S42 includes the steps of:

s423, acquiring a circumscribed rectangle of the character according to the x-axis and y-axis boundaries of the character obtained in the step S422;

s424, calculating the size of each character and the distance between adjacent characters according to the circumscribed rectangle of each character;

and S425, counting the depth or height data of the character according to the character point cloud data obtained in the step S3 in the circumscribed rectangle of the character.

A character size detection device comprises a 3D contour measurement unit, a guide rail fixing unit base, a motor driving unit, a displacement sensor and a processing unit; the guide rail is installed on the guide rail fixing unit base, the 3D profile measuring unit is installed on the guide rail and can move along the guide rail under the drive of a motor, the displacement sensor measures the displacement of the 3D profile measuring unit moving along the guide rail, the 3D profiler is triggered to perform profile sampling once the guide rail is set to move for an interval distance, and the processing unit receives a profile data result sent by the 3D profile measuring unit and processes the profile data.

Has the advantages that:

1. the invention can reduce the manual detection cost of the VIN code, the engine number and the like of the motor vehicle in the field of motor vehicle manufacturing and inspection by carrying out full-automatic detection on the depth or height of the code-engraved character, the size of the character and the distance between adjacent characters, improves the detection precision and the detection efficiency and has important significance for promoting the detection automation of the industry;

2. the contour data obtained by 3D scanning comprises the information of the geometric surface where the character is located and the geometric information of the character. To extract the geometric data of the character, it is necessary to extract the data of the geometric surface where the character is located from the contour data. To this end, the measures taken by the method of the invention comprise: a fitting polynomial curve model is established for each contour curve to represent the contour surface shape corresponding to the curve, and geometric data of the character can be accurately acquired in such a way;

3. when the weighted polynomial is used for processing, noise and abnormal points are removed, and the result is more accurate;

4. screening out character point cloud data according to the concave-convex characteristics of the characters and the preset depth/height range of the characters, and in this way, the character point cloud data can be accurately screened out;

5. generally, the VIN code and engine number are stamped on a flat surface; in order to increase the application range of the method, the characters are supposed to be engraved on the surface of the curved surface, and a processing mode aiming at the character surface being the curved surface is provided, so that the practicability of the method is further improved;

6. through binarization processing and a projection method or a character recognition algorithm, point cloud data can be more accurately recognized and counted;

7. isolated points and small-area connected domains are removed through connected domain calculation, so that the identification accuracy can be improved;

8. the geometric dimension obtained by the projection method is more accurate;

9. by adopting the character geometric dimension statistical method, the size dimension of the character, the space between adjacent characters, and the depth or height of the character can be accurately obtained.

10. The device can be used for automatically detecting the VIN code, the engine number and the like of the motor vehicle, and the detection precision and the detection efficiency are improved.

Drawings

FIG. 1 is a schematic flow diagram of an embodiment of the process of the present invention;

FIG. 2 is a schematic diagram of a method for extracting parameters of a profile curve model according to the present invention;

FIG. 3 is a schematic diagram of a weighted polynomial curve fitting algorithm of the present invention;

FIG. 4 is a schematic diagram of the present invention for extracting point cloud data of engraved characters;

FIG. 5 is a schematic diagram of the present invention illustrating the expansion of a character point on a curved surface into a coordinate of a character point on a plane;

FIG. 6 is a flow chart of a character extraction and recognition algorithm according to the present invention;

FIG. 7 is a schematic diagram of the character area division by the projection method according to the present invention;

FIG. 8 is a schematic diagram illustrating the calculation of character size and spacing;

fig. 9 is a schematic diagram of the structure of the device of the present invention.

Reference numerals:

10. a guide rail mounting base; 20. a 3D profile measurement unit; 30. a displacement sensor; 40. a guide rail; 50. an electric motor.

Detailed Description

The invention will be further described with reference to the accompanying drawings in which:

example (b):

as shown in fig. 1, a method for detecting a character size includes the following steps:

s1, acquiring 3D contour data of the surface of the code carving character;

s4, extracting single characters, counting the geometric dimension of each character, and checking the space between adjacent characters;

and S5, generating a detection report and outputting the result.

A method of obtaining 3D contour data of a surface of an engraved character, comprising: and triggering measurement once at a certain micro distance by adopting a line laser profiler to obtain a currently scanned 3D profile curve, and continuously scanning the character surface to obtain 3D profile data of the character surface. Satisfactory measurements are typically obtained by having the laser profilometer sample the character profile once every 0.1mm or so.

The invention can reduce the manual detection cost of the VIN code, the engine number and the like of the motor vehicle in the field of motor vehicle manufacturing and inspection by carrying out full-automatic detection on the depth or height of the code-engraved character, the size of the character and the distance between adjacent characters, improves the detection precision and the detection efficiency and has important significance for promoting the detection automation of the industry.

As shown in fig. 9, an apparatus for character size detection includes a 3D profile measuring unit 20, a guide rail 40, a guide rail fixing unit base 10, a motor 50, a motor driving unit, a displacement sensor 30, a processing unit, and a result reporting unit; the guide rail 40 is installed on the guide rail fixing unit base 10, the 3D contour measuring unit 20 is installed on the guide rail 40 and can move along the guide rail 40 under the driving of the motor 50, the displacement sensor 30 measures the displacement of the 3D contour measuring unit 20 moving along the guide rail 40, and triggers the 3D contour measuring unit 20 to perform contour sampling once every time the guide rail 40 moves for a certain distance, and the processing unit receives the contour data result sent by the 3D contour measuring unit 20 and processes the contour data. The result reporting unit provides the detection result; the processing unit performs the processing steps comprising the method of the invention. Fig. 9 shows a schematic of the inventive apparatus without the processing unit, the motor drive unit and the result reporting unit.

The 3D profile measuring unit 20 is a laser profiler, the parameters of which include: the range of the z axis is 2-10 mm, the precision of the z axis is not lower than 0.01mm, and the range of the horizontal axis is 20-50 mm.

The guide rail fixing unit base 10 is a magnetic base including a magnetic circuit switch, and can be fixed to an object having a magnetic conductive property, such as an automobile frame. When the device is used, the device can be placed above a character to be detected, and the magnetic circuit switch is turned on, so that the device can reliably adsorb the character.

To improve the stability of the scanning operation, the guide rail 40 is a precision lead screw guide rail.

And a displacement sensor 30 installed between the 3D profile measuring unit 20 and the guide rail 40, for measuring a displacement of the 3D profile measuring unit 20 moving along the guide rail 40. The displacement sensor 30 may be a grating, a magnetic grating or a laser displacement sensor, and as a typical parameter, its displacement detection sensitivity is not lower than 5 um.

If a servo motor is selected, the rotation angle measurement value of the grating encoder can be used to replace the displacement sensor 30 equivalently because the servo motor is provided with the grating encoder.

And the 3D contour measuring unit 20 is triggered to perform a movement trigger interval of contour sampling once every time the guide rail 40 moves for an interval distance, wherein the trigger interval is 0.01-0.25 mm.

A processing unit comprising: executing the steps of algorithmic processing comprising the method of the invention; the processing unit comprises an industrial personal computer, an embedded processor and the like.

The contour data obtained by 3D scanning comprises the information of the geometric surface where the character is located and the geometric information of the character. To extract the geometric data of the character, it is necessary to extract the data of the geometric surface where the character is located from the contour data. To this end, the measures taken by the method of the invention comprise: and establishing a fitting polynomial curve model for each profile curve to represent the profile surface shape corresponding to the curve.

In one embodiment, as shown in fig. 2, the step S2 further includes the following steps:

s21, establishing a fitting polynomial curve model for each contour curve;

s22, solving a fitting polynomial curve parameter model of the contour curve through a weighted polynomial curve fitting algorithm or a RANSAC algorithm, wherein the RANSAC algorithm is relatively mature, so that the method is not described in detail;

and S23, outputting the fitted polynomial profile curve parameters.

In one embodiment, in order to improve the calculation accuracy of the geometric surface contour where the character is located as much as possible, the method for solving the weighted polynomial curve fitting algorithm of the fitted contour curve of the present invention employs an iterative algorithm, as shown in fig. 3, and in step S22, the step of solving the fitted polynomial curve parameter model of the contour curve using the weighted polynomial curve fitting algorithm includes the following steps:

s2222, fitting a polynomial curve to the profile curve by a least square method;

s2223, calculating the distance between all points on the contour curve and the fitted polynomial curve obtained in the step S2222, removing the points with the distance larger than a set threshold value, namely adjusting the weight of the points to 0, and counting the number of the remaining points falling on the fitted contour curve on the original contour curve;

s2224, judging whether the number of the remaining points falling on the fitting contour curve reaches the maximum, namely, until the number of the remaining points falling on the fitting curve output by the step is not increased any more, if so, entering the step S2226, and if not, entering the step S2225;

it is to be understood that curve smoothing, least squares, and point-to-curve distance calculation are all conventional and general algorithms for those skilled in the art, and the present invention is not explained in detail herein.

and S2226, outputting the fitted polynomial profile curve parameters.

The highest degree of the variable of the polynomial curve can be preset. Generally, the VIN code and engine number are stamped on a flat surface; in order to increase the applicability of the method of the invention, it is assumed that characters may also be engraved on curved surfaces. For a planar surface, the profile curve is a straight line, and therefore, the highest degree of the variation of the polynomial curve corresponding to the plane can be set to 1. For curved surfaces, it is contemplated that the highest degree of variation of the set polynomial curve is 2 or 3.

In one embodiment, as shown in fig. 4, the step S3 further includes the following steps:

s31, for each contour curve, calculating the distance from each point on the original contour curve to the fitting polynomial curve of the contour curve;

s32, selecting points on one side of the curve and with the distance within a set range as point cloud data of the characters according to the concave-convex characteristics of the characters;

and S33, splicing the character point cloud data extracted from all the contour curves into the obtained point cloud data of the whole character surface, and outputting the character point cloud data.

The steps between S32 and S33 further include the following steps:

The above steps are further explained by taking lettering as an example. The lettering characters are usually recessed on the surface of the characters, so that only points below the surface contour line need to be considered when extracting the point cloud data of the characters. And in combination with the reasonable depth of the character under the normal condition, a point set which accords with the geometric characteristics of the character can be quickly screened out and used as point cloud data of the character.

On the other hand, if the geometric surface on which the character is located is an arc-shaped curved surface, in order to improve the calculation accuracy, the arc-shaped curved surface needs to be unfolded into a plane, and the specific principle is as shown in fig. 5. In fig. 5, point p on the original curve is flattened and then located at point q on the straight line. The corresponding relationship is that the length of the curve op is equal to the length of the straight line oq.

In one embodiment, as shown in fig. 6, the step S4 includes the following steps:

s41, performing binarization processing on the point cloud data on the whole character surface, for example, marking the position of a character point as 1, and generating a binarization character image if the rest are 0;

and S42, extracting the characters from the binary character image by adopting a projection method, and counting the geometric dimension of each character.

The step S41 is followed by the following steps:

s411, conducting connected domain calculation on the binary character image, and removing isolated points and small-area connected domains with areas not enough to form minimum point strokes.

In one embodiment, as shown in fig. 7, the projection method in step S42 includes the following steps:

In one embodiment, in step S4, the method for extracting a single character further includes, as a functional option, a character recognition algorithm, that is, extracting and recognizing a character;

the character recognition algorithm comprises the following steps: a template matching algorithm or a neural network algorithm; both of these character recognition algorithms are commonly used in the art.

In one embodiment, the step of counting the geometric size of each character in the step S42 includes the following steps:

and S425, counting the depth or height data of the character according to the character point cloud data obtained in the step S3 in the circumscribed rectangle of the character. The implementation of this step is schematically illustrated in fig. 8. By adopting the character geometric dimension statistical method, the size dimension of the character, the space between adjacent characters, and the depth or height of the character can be accurately obtained.

The foregoing illustrates and describes the principles, general features, and advantages of the present invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principle of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the present invention, which fall within the scope of the invention as claimed.

Claims

1. A method of character size detection, comprising the steps of:

s1, acquiring 3D contour data of the surface of the code carving character;

s2, calculating data of a geometric surface where the character is located from the contour data, wherein the character surface is a curved surface;

s21, establishing a fitting polynomial curve model for each contour curve;

s2222, fitting a polynomial curve;

s2226, outputting fitted polynomial profile curve parameters;

s23, outputting fitted polynomial contour curve parameters;

s322, converting the character point cloud coordinates on the curved surface into character point cloud data on a plane;

s33, outputting character point cloud data;

s411, calculating a connected domain of the binary character image, and removing isolated points and a small-area connected domain;

s42, extracting characters from the binary character image by a projection method or a character recognition algorithm, and counting the geometric size of each character;

s421, projecting the binary image of the character along an x axis and a y axis respectively, namely accumulating pixel values of the binary image to obtain 2 projection images;

s422, calculating the boundary of each character along the x axis and the y axis on the 2 projection images by using a threshold segmentation method;

s424, calculating the size of each character and the distance between adjacent characters according to the circumscribed rectangle of each character; in the circumscribed rectangle of the character, according to the character point cloud data obtained in the step S3, counting the depth or height data of the character;

and S5, outputting the result.