CN114166121A

CN114166121A - Log measuring method based on four-side scale

Info

Publication number: CN114166121A
Application number: CN202111454867.0A
Authority: CN
Inventors: 洪茂雄
Original assignee: Fujian University of Technology
Current assignee: Fujian University of Technology
Priority date: 2021-12-01
Filing date: 2021-12-01
Publication date: 2022-03-11
Anticipated expiration: 2041-12-01
Also published as: CN114166121B

Abstract

The invention relates to a log scale detection method based on a four-side scale. The four-side ruler is taken as an auxiliary tool, a hand-held camera is used for collecting an image of the end face of the log, the log image is corrected through a machine vision algorithm, and finally the diameter of the end face of the log is measured in the corrected image. The invention abandons the method of fixing camera and calibration board in traditional method, and adopts the method of correcting image with four-side scale. The invention firstly uses the geodesic measuring pole and the T-shaped connecting block to assemble the four-side ruler, collects the close-shot image of the four-side ruler and the end surface of the log, detects the reference point of the frame of the four-side ruler, obtains the grid by interpolation, obtains the correction diagram by the perspective transformation of the grid, and finally, on the correction diagram, the diameter of the end surface of the log can be calculated by the size of the grid, thus finishing the visual ruler checking work of the log pile.

Description

Log measuring method based on four-side scale

Technical Field

The invention relates to the field of image processing and computer vision, in particular to a log scale detection method based on a four-side scale.

Background

Wood volume measurement is one of the important tasks in the timber harvesting process, and the sale and processing of wood also depends on the measurement of wood volume. The volume calculation is approximated by the volume of the cylinder, and can be accomplished by knowing the log caliper and the caliper length or by looking up the log volume table according to the log volume calculation formula taught in the national standard log volume table (GB 4814-84).

For a long time, for wood processing enterprises and wood selling enterprises, the wood volume measurement work is finished by using an artificial measuring tape, and the length and the diameter of the measuring tape are directly measured by using tools such as a measuring tape or a caliper. Manual dipsticks suffer from a number of disadvantages, among which are obvious: 1) the cost of labor is high, and the scale is inefficient, 2) the scale rule is more complicated, and different types of log have different scale rules, 3) have very high requirement to the working attitude and the working experience of the scale operator for the highly accurate scale, 4) general timber is piled up and is put certain height, and the scale operator is climbing from top to bottom often, and the scale measurement in-process can take place dangerously.

At present, the mode of manual scale checking firstly numbers the logs, and the scale checker measures the logs and inputs the measurement data, so that the scale checking mode is an invisible scale checking result, namely the log image and the scale checking result are associated, and meanwhile, the condition of the log scale checking cannot be traced back.

The image of the object to be measured is collected, the size of the object is measured by the computer vision technology, and the automatic detection of many industrial products, such as metal workpieces, plastic products, electronic products and the like, is almost successfully realized, however, the traditional computer vision measurement technology is difficult to be directly applied to the log stacking ruler, and the reasons are as follows: 1) generally, an image-based measurement technology needs a Calibration (Calibration) process before implementation, a Calibration board is used to collect a plurality of images, and a Calibration algorithm is used to calculate camera parameters. 2) In the conventional measurement technique, a camera is usually fixed and then calibrated, then an image of an object to be measured is acquired, and measurement data is calculated according to camera parameters. Maintaining the fixed position and pose of the camera, the facilities, appliances and techniques required to achieve outdoor environments are much more difficult and costly than indoor environments. 3) The log collection area scope is big, with the concentrated fortune of log to the check ruler station, consuming time and consuming transportation cost, lets the check ruler personnel carry the measuring equipment and disperses to the point check ruler of gathering, saves transportation cost on the one hand, and on the other hand can reduce the influence of check ruler operation to log felling operation, consequently, designs a correction-free flow, can be with handheld camera to and easily carry the measuring equipment, accomplish log heap check ruler work, becomes the research subject of being worth.

In many nameplates, a rectangle is one of the most common shapes, and due to the view angle factor of the camera, the imaged nameplate image is distorted (distorted), as shown in fig. 1(a) - (b), the shape of the rectangular license plate becomes an arbitrary quadrangle in the image, and at the same time, the sizes of letters and numbers in the license plate also change, the sizes of letters and numbers near the camera become larger, and the sizes of letters and numbers near the camera become smaller, in the actual license plate, all letters and numbers are the same in size, and through Perspective transformation (Perspective transformation), the arbitrary quadrangle can be corrected back to a rectangle, as shown in fig. 1(c) - (d), the shape of the license plate is corrected back to the rectangle of the view angle, and the letters and numbers in the license plate are restored to equal sizes, so the rectangle with known size has the characteristic of correcting the distorted image.

Disclosure of Invention

The invention aims to develop a visual measurement algorithm without a correction process for the end face of a log pile by taking a four-side ruler as an auxiliary tool, and particularly provides a log checking method based on the four-side ruler.

In order to realize the purpose of the invention, the device is divided into two parts: 1) the four-side ruler design utilizes detachable red and white alternative geodesic measuring marker posts and T-shaped connecting blocks sold in the market at present, a rectangular frame is combined by a left vertical post, a right vertical post, an upper transverse post and a lower transverse post, and red and white alternative sections form scales with fixed length and are used as reference points for image correction. 2) The vision measurement algorithm inputs the image of the log end face in the four-side scale frame, detects the reference points of the four-side scale frame, obtains grids (Grid) by an interpolation method, corrects the quadrilateral grids into squares through perspective transformation one by one, obtains a correction graph after the image is transformed along with a series of perspectives, and on the basis of the assumption that the log end face and the grids are in the same plane, the diameter of the log end face can be calculated by the grids with known sizes on the correction graph.

In order to achieve the purpose, the technical scheme of the invention is as follows: a log scale detection method based on a four-side scale comprises the following steps:

s1, placing the side of the end face of the log pile to enable the end face of the log pile to be framed by the four-side ruler, enabling the rectangular frame of the four-side ruler to be on the same plane with the end face of the log pile, and collecting images containing the four-side ruler and the log pile;

step S2, inputting a test image, and manually dragging an arbitrary quadrangle on a program user interface to ensure that the quadrangle can be matched with a four-side ruler;

step S3, after dragging the scale fitting the four sides, the four sides of any quadrangle are pasted on the marker post, the gray scale image is sampled for the second time along the four sides to form a one-dimensional signal, the marker post is painted in red and white alternatively, so that the one-dimensional signal is similar to a square wave signal, noise is removed through smooth filtering, then a first derivative is obtained for the smoothed signal, a local maximum value of which the first derivative is larger than a threshold value and a local minimum value of which the first derivative is smaller than the threshold value are selected, and the position of a red-white boundary point of the marker post is obtained according to the local maximum value and the local minimum value;

s4, detecting red and white junction points on the marker post to obtain four groups of segmentation points of the upper side, the lower side, the left side and the right side, obtaining four corner drop points by an extrapolation method according to the segmentation points, obtaining grid points in a frame by an interpolation method, and forming a complete grid by adding the four-side segmentation points and the four corner drop points by the interpolation grid points;

step S5, after the grids are generated, converting each grid into a rectangle with the same size, calculating a homography matrix of each grid for each grid, converting each grid into a new coordinate, and generating a correction image;

and step S6, for each log in the corrected image, manually dragging a straight line on a program user interface, measuring the log diameter from one point of the outline to another point of the outline through the center of the wood, converting the image coordinate into a real world coordinate by a scale, and finally calculating the distance between the two points to obtain the log diameter.

In an embodiment of the present invention, in step S3, the red-white boundary positions of the marker post are as follows:

wherein (x)_i，y_i) Is the coordinate of the red and white boundary point of the marker post, (x)₀，y₀) Is the coordinate of the starting end of the edge, θ is the oblique angle of the edge, l_iIs the value on the x-axis of the local maximum and local minimum.

In an embodiment of the present invention, step S5 is implemented as follows:

after the grids are generated, each grid is converted into a rectangle with the same size, for each grid, the original (x, y) coordinates are converted into (u, v) coordinates, homography matrix conversion is carried out by the formula (2), 4 pairs of corresponding points are needed for calculating the homography matrix H, and the (x) coordinates are obtained by the method₁，y₁)→(u₁， v₁)、(x₂，y₂)→(u₂，v₂)、(x₃，y₃)→(u₃，v₃)、(x₄，y₄)→(u₄，v₄) Wherein (u)₁，v₁)～(u₄，v₄) Is the coordinates of four corner points of a rectangleThe homography matrix H acts on the corresponding points to obtain an equation of formula (3), and a least square solution of H can be found by using an SVD algorithm

Wherein the H homography matrix is a 3 x 3 matrix, H₁,h₂,…,h₈The first 8 elements of the matrix and the last 1 element is 1, as arranged by equation (2).

In an embodiment of the present invention, step S6 is implemented as follows:

obtaining a corrected image after each grid conversion, manually dragging a straight line on a program user interface for each log in the corrected image, measuring the diameter of the log from one point of the outline to another point of the outline by passing through the center of the log, and taking the image coordinates (x) of two end points A, B of each straight line_A，y_A) And (x)_B，y_B) Converting the image coordinates into real world coordinates by a scale, scale (x scale, s) in the x direction_x) Scale bar in y-direction (y scale, s) for actual length of the target/width of the grid image_y) Finally, the distance between two points is calculated A, B by equation (4) to calculate the log diameter for the actual length of the marker post/the height of the grid image:

where dist (A, B) is the distance between A, B two points, i.e. the diameter of the log.

Compared with the prior art, the invention has the following beneficial effects: compared with the traditional method, the visual ruler method for the log pile is characterized in that a four-side ruler is used, a correction process is not needed, and a camera is not needed to be fixed.

Drawings

FIG. 1 is a correction of a license plate image;

FIG. 2 is a geodetic pole and a four-sided scale;

FIG. 3 is a close-up image of a log model and a four-sided ruler;

FIG. 4 is a drawing of an arbitrary quadrilateral;

FIG. 5 is a one-dimensional signal processing;

FIG. 6 is a grid generation;

FIG. 7 is a grid coordinate transformation;

fig. 8 is a corrected image.

Detailed Description

The technical scheme of the invention is specifically explained below with reference to the accompanying drawings.

The technical solution of the present invention will be clearly and completely described below with reference to the embodiments.

The invention aims to develop a vision measurement algorithm without a correction process for the end surface of a log pile by taking a four-side ruler as an auxiliary tool.

[ ASSEMBLY FOUR-EDGE SCALE ]

Two sets of geodetic marker poles are purchased, 1 group has 5 releasable connection quarter butt, 1 quarter butt is 100 centimetres long, 1 quarter butt has 5 red and white festivals, every festival is 20 centimetres long, as figure 2(a), use 4T word connecting blocks to assemble into the four sides scale, as figure 2(b), assemble montant about earlier, pass T word connecting block, the horizontal pole about the equipment again, T word connecting block is in festival alternative department adjustment, make the inside casing height and the width of four sides scale can be the integral multiple of festival length, as figure 2(b), inside casing height 140 centimetres is exactly 7 knots long, inside casing width 100 centimetres is exactly 5 knots long.

[ Vision measuring method ]

Placing the side of the end face of the log pile to enable the end face of the log pile to be framed by the four-side ruler, enabling the rectangular frame of the four-side ruler and the end face of the log pile to be approximately positioned on the same plane, and collecting images containing the four-side ruler and the log pile.

To verify the method, in the laboratory, a log model was made, logs of varying diameters were sawn, a plywood was installed, placed on the floor, and the captured image was framed with four-sided rulers, which were dispersed in the four-sided ruler frame at a level close to the log end face, as shown in fig. 3.

Step two, inputting a test image, dragging an arbitrary quadrangle on a program user interface in a manual mode to enable the quadrangle to be matched with a quadrangle scale, as shown in fig. 3, a quadrangle is added on an original image in fig. 3, dragging control small squares are arranged on four corners, the dragging control small squares are clicked to a T-shaped connecting block respectively, after four times of dragging are completed, the quadrangle can be matched with the quadrangle scale, and fig. 4 shows that the dragging of the arbitrary quadrangle is completed.

Step three, after the any quadrangle is dragged to fit with the four-side ruler, four sides of the any quadrangle are pasted on the marker post and respectively follow the four sides, the gray-scale image is subsampled (Subsampling) to form a one-dimensional signal, as shown in fig. 5(a), where the x-axis is the position of the edge, the y-axis is the subsampled gray-scale value, because the marker post is painted in red and white, the one-dimensional signal can be similar to a square wave signal, the noise is removed through smooth filtering, then, the first derivative of the smoothed signal is obtained, as shown in fig. 5(b), a Local maximum (Local maximum) whose first derivative is greater than the threshold is selected, and a Local minimum (Local minimum) less than a threshold, the 2 thresholds being set at 6.0 and-6.0 respectively, these local maxima and local minima are located exactly at the red-white cross-over point of the benchmarks, and finally, the x-axis values (l) of the local maxima and local minima._i) The coordinate (x) of red and white boundary point of the marker post is converted by formula 1_i,y_i) Wherein (x)₀,y₀) Is the coordinate of the starting end of the edge and θ is the oblique angle of the edge.

Step four, the detected red and white junction points on the marker post can be regarded as Segment points (Segment points) on four sides, and four sides obtain four groups of Segment points: the upper side, the lower side, the left side and the right side are as shown in fig. 6(a), then four corner points are supplemented, and the method is obtained by utilizing the existing equal division point extrapolation (extrapolation), for example, two straight lines of the 1 st and 2 nd segmentation points on the upper side, two straight lines of the 1 st and 2 nd segmentation points on the left side, extended intersection of the two straight lines obtain a left upper corner point, for example, two straight lines of the 1 st and 2 nd segmentation points on the upper side, and two straight lines of the 1 st and 2 nd equal division points on the right side, extended intersection of the two straight lines obtain a right upper corner point, and a left lower corner point and a right lower corner point are also obtained by the method.

In order to perform Grid rectification (Grid rectification) to eliminate image distortion, an interpolation method is adopted to obtain Grid points in a frame, the segmentation points on the upper side and the lower side are in one-to-one correspondence, and the segmentation points on the left side and the right side are in one-to-one correspondence, so that a vertical connecting line of the upper segmentation point and the lower segmentation point and a horizontal connecting line of the left segmentation point and the right segmentation point are drawn, a Grid point is obtained at the intersection point of the two direction connecting lines, and the interpolation Grid points, the four side segmentation points and the four corner points form a complete Grid, as shown in fig. 6(b), and finally, a 5 x 7 Grid is generated.

Step five, after the grids are generated, converting each grid into a rectangle with the same size, for each grid, converting the original (x, y) coordinates into (u, v) coordinates, converting by a homography matrix of a formula 2, and calculating the homography matrix H by 4 corresponding points, (x, y) and₁,y₁)→ (u₁,v₁)、(x₂,y₂)→(u₂,v₂)、(x₃,y₃)→(u₃,v₃)、(x₄,y₄)→(u₄,v₄) Wherein (u)₁,v₁)～(u₄,v₄) For the coordinates of the four corner points of the rectangle, as shown in fig. 7, a homography matrix H is applied to the corresponding points to obtain equation 3, and a least squares solution of H can be found using SVD algorithm.

Step six, obtaining a corrected image after each grid is converted, as shown in fig. 8(a), manually dragging a straight line on a user interface for each log in the corrected image, measuring the diameter of the log from one point of the outline to another point of the outline by passing through the center of the log, as shown in fig. 8(B), and taking image coordinates (x) of two end points A and B of each straight line_A,y_A) And (x)_B,y_B) Converting the image coordinates into real world coordinates by a scale, scale (x scale, s) in the x direction_x) Scale bar in y-direction (y scale, s) for actual length of the target/width of the grid image_y) The actual length of the marker post/the height of the grid image is in mm/pixel, and finally, the distance between the point A and the point B is calculated by the formula 4 to calculate the original wood diameter (mm).

12 log models are photographed in a laboratory for 3 times, each log is subjected to 3 times of visual measurement, the results are listed in table 1, and errors between the visual measurement and manual measurement are compared, wherein errors between 4 large-size logs are-8.32-14.51 mm, errors between the rest logs are-4.95-0.62 mm, standard deviations between 3 times of measurement are 4.18 except for 1 log, the rest logs are below 2.03, the maximum error is 5.29% in percentage absolute error, and the average error is 2.35%.

TABLE 1 results of the experiment (mm)

The described embodiments are only some embodiments of the invention, not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

The above are preferred embodiments of the present invention, and all changes made according to the technical scheme of the present invention that produce functional effects do not exceed the scope of the technical scheme of the present invention belong to the protection scope of the present invention.

Claims

1. A log scale detection method based on a four-side scale is characterized by comprising the following steps:

2. The method for measuring logs based on a four-sided ruler as claimed in claim 1, wherein in step S3, the red-white boundary points of the marker post are as follows:

3. The log measuring method based on the four-side ruler of claim 1, wherein the step S5 is implemented as follows:

after the grids are generated, each grid is converted into a rectangle with the same size, for each grid, the original (x, y) coordinates are converted into (u, v) coordinates, homography matrix conversion is carried out by the formula (2), 4 pairs of corresponding points are needed for calculating the homography matrix H, and the (x) coordinates are obtained by the method₁，y₁)→(u₁，v₁)、(x₂，y₂)→(u₂，v₂)、(x₃，y₃)→(u₃，v₃)、(x₄，y₄)→(u₄，v₄) Wherein (u)₁，v₁)～(u₄，v₄) For the coordinates of four corner points of a rectangle, the homography matrix H acts on the corresponding points to obtain an equation of formula (3), and a least square solution of H can be found by using an SVD algorithm

4. The log measuring method based on the four-side ruler of claim 1, wherein the step S6 is implemented as follows: