CN112001974A - Calibration device and calibration method for underwater stereoscopic observation system - Google Patents

Abstract

The invention discloses a calibration device for an underwater stereoscopic observation system. equipment; the calibration plate moving device includes a positioning track, the positioning track is fixed on the top of the glass water tank through a hanger, the positioning track includes a set of track short shafts and a set of track long shafts that are oppositely arranged, and a set of track short shafts is opened on the track short shaft. Several first card slots, the two ends of the moving rod are set on the track short axis of the positioning track through the first card grooves, and several groups of second card grooves are opened on the moving rod, and the top of the moving mechanism passes through the second card grooves. The rod is detachably connected, and the calibration plate is fixedly connected with the moving mechanism; the calibration plate changes its position in the glass water tank through the moving mechanism and the moving rod, and the binocular vision device obtains the sampling image of the calibration plate at the current position.

Description

An underwater stereoscopic observation system calibration device and calibration method

technical field

The invention relates to the field of aquaculture, in particular to a calibration device and a calibration method for an underwater stereoscopic observation system.

Background technique

In aquaculture, it is often necessary to calibrate the target object in the glass tank. However, when using the visual measurement method to reconstruct the target body in three dimensions, the visual system and the measured target are often in the same medium environment (such as air). The system is placed outside the glass water tank, and the movement of the target object in the water is observed through the glass. Therefore, in addition to the air medium, there are glass medium and water medium between the measured target and the vision system. Under the action of these three different refractive index media, the original imaging optical path of the target object has changed, and the reconstruction of the spatial coordinates of the target point by means of the intersection of the straight-line optical paths with the image point positions of different angles will cause a large error.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a calibration device and a calibration method for an underwater stereoscopic observation system in view of the deficiencies in the prior art, so as to solve the problems existing in the prior art.

The technical problem solved by the present invention can be realized by the following technical solutions:

An underwater stereoscopic observation system calibration device, comprising a calibration plate moving device arranged on a glass water tank, a calibration plate fixed inside the glass water tank through the calibration plate moving device, and a binocular vision device arranged outside the glass water tank;

The calibration plate moving device includes a positioning track, the positioning track is fixed on the top of the glass water tank through a hook, the positioning track includes a set of track short shafts and a set of track long shafts that are oppositely arranged, and a plurality of first track shafts are opened on the track short shafts. A card slot, the two ends of the moving rod are arranged on the track short axis of the positioning track through the first card groove, and several groups of second card grooves are opened on the moving rod. The top of the moving mechanism can be connected to the moving rod through the second card grooves Dismantling the connection, the calibration plate is fixedly connected with the moving mechanism;

The position of the calibration plate in the glass water tank is changed by the moving mechanism and the moving rod, and the binocular vision device obtains the sampling image of the calibration plate at the current position.

Further, the sampling process of the binocular vision device is as follows:

1) Install binocular vision equipment, calibration board moving device and calibration board, adjust the position of the binocular vision equipment, so that the glass water tank can be completely imaged;

2) Place the moving rod at the starting position of the short axis, place the moving mechanism in the slot at the starting position of the long axis, take the position of the calibration plate at this time as the sampling starting point, and set the target point in the lower left corner of the calibration plate as the world coordinate system Origin, take the first frame of image;

3) Translate the moving mechanism twice in sequence along the second slot of the moving rod, and shoot the 2nd to 3rd frame images respectively;

4) Keep the position of the moving mechanism stationary, translate the moving rod along the short axis of the track to the next first card slot, and shoot the 4th frame image;

5) The moving mechanism is reversely translated twice along the second slot of the moving rod, and the 5th to 6th frame images are taken;

6) Keep the position of the moving mechanism still, translate the moving rod along the short axis of the track to the next first card slot, repeat steps 2)-5) until the last first card slot position, and complete the full-view sampling.

Further, the calibration method of the binocular vision device is:

1) Obtain the position information of the calibration board through the machine vision software, and after obtaining the position of the calibration board, use the operator to segment the circles in the area, and find the number, perimeter, coordinate position of the circle and the consistent target in the description file of the calibration board point;

2) Calculate the pixel coordinates of the target point; set the X axis of the world coordinate system to be parallel to the long axis of the fish tank, the Y axis to be perpendicular to the bottom of the fish tank, the Z axis to be parallel to the short axis of the fish tank, and the origin to be the lower right target point of the starting position of the calibration board , the world coordinate system coordinates of the target point are obtained by combining the moving distance of the calibration plate and the spatial distribution of the target point;

3) Divide the effectively collected target points into a training set and a test set; use the genetic algorithm to optimize the parameters of the training set of the coordinate system, and when the iteration reaches the maximum number of times, output the optimal combination to obtain the optimized model parameters;

4) Use the model parameters determined by the genetic algorithm to construct the SVR training model of the coordinate system, and predict the three parameters in the world coordinate system; wherein, the target point includes left and right image pixel coordinates and world coordinate information, and the unit of the image coordinate value is Pixel, that is, the pixel coordinates of the target point in the image.

Further, the calibration method of the binocular vision device includes an error evaluation process, and the method is as follows:

A combination of axial error evaluation and single-point error evaluation is adopted. The axial error evaluation adopts the MSE function, and the single-point error evaluation adopts the P _MSE function, as follows:

In the formula, x', y', z' are the actual spatial coordinates of the measurement point, _xi ', y _i ', z _i ' are the prediction data of the measurement point model, and d _i is the actual measurement point on the x, y or z axis. Spatial coordinates, d _i ' are model prediction data in X, Y or Z axis for all measurement points.

Compared with the prior art, the beneficial effects of the present invention are:

Reconstructing the three-dimensional space coordinate information of the target through image acquisition and analysis has the advantages of non-contact, high precision, and wide measurement range, which can well meet the requirements of calibration.

Description of drawings

FIG. 1 is a schematic diagram of the calibration device of the underwater stereoscopic observation system according to the present invention.

FIG. 2 is a schematic diagram of the calibration plate according to the present invention.

FIG. 3 is a schematic diagram of the positions of the binocular vision device and the glass water tank according to the present invention.

Detailed ways

In order to make the technical means, creative features, achievement goals and effects realized by the present invention easy to understand, the present invention will be further described below with reference to the specific embodiments.

Referring to Figure 1, Figure 2 and Figure 3, an underwater stereoscopic observation system calibration device according to the present invention includes a calibration plate moving device arranged on the glass water tank (3), and is fixed on the glass water tank (3) through the calibration plate moving device. 3) The internal calibration plate and the binocular vision device arranged outside the glass water tank (3);

The calibration plate moving device includes a positioning track (1), the positioning track (1) is fixed on the top of the glass water tank (3) by a hook (5), and the positioning track (1) includes a set of relatively arranged track short shafts ( 6) and a set of track long shafts (7), a plurality of first card slots (61) are opened on the track short shafts (6), and both ends of the moving rod (4) are arranged in the positioning position through the first card slots (61). On the track short axis (6) of the track (1), several groups of second clamping grooves (41) are opened on the moving rod (4), and the top of the moving mechanism (2) passes through the second clamping groove (41) and the moving rod. (4) Removable connection, the calibration plate is fixedly connected with the moving mechanism (2);

The position of the calibration plate in the glass water tank (3) is changed by a moving mechanism (2) and a moving rod (4), and a binocular vision device acquires a sampling image of the calibration plate at the current position.

Example 1

The inner diameter of the positioning track (1) is 1300×900mm, the upper and lower thickness is 20mm, the width of one side is 50mm, the distance between the long axis of the track (7) and the inner wall of the glass water tank (3) is 41mm, and the short axis of the track (6) 3) The inner wall spacing is 16mm. A hanger (5) is used to connect with the glass water tank (3), and the upper plane of the positioning track (1) is flush with the top of the glass water tank (3).

The moving rod (4) is 1320mm in length, 20mm in upper and lower thickness, and 20mm in width, and is fixed on the track short axis (6) through the first card slot (61).

A first slot (61) (length 10mm, width 20mm, depth 10mm) is machined every 50mm from the initial position of the track short axis (6), and a total of 17 pairs of slots are machined.

Along the direction of the long axis (7) of the track, a pair of second slots (41) (length 20mm, width 20mm, depth 10mm) are machined on the moving rod (4) at intervals of 400mm from the starting position of the long axis (length 20mm, width 20mm, depth 10mm). The clamping grooves (41) are at an interval of 550 mm and have the same width as the hanging arm of the moving mechanism (2). A total of 3 pairs of second clamping grooves (41) are processed.

The moving mechanism (2) is 900mm high and 550mm wide, and is fixed on the moving rod through the second card slot (41), and the calibration plate is fixed on the bottom end of the moving mechanism (2).

The calibration plate moving device is processed by a high-precision CNC machine tool, the processing accuracy is ±0.05mm, and the installation accuracy of the second clamping groove (61) and the second clamping groove (41) is ±0.1mm. When the camera is calibrated, the calibration plate is vertically fixed on the moving rod (4), the calibration plate is moved along the long axis direction and the moving rod is moved along the short axis direction according to the position of the card slot, and sample images are taken at each position, so that the collected The calibration plate image can cover the entire space of the glass water tank (3), so as to realize the sampling of the whole field of view.

The sampling process of the binocular vision device is as follows:

1) Install binocular vision equipment and calibration plate moving device, adjust the position of the binocular camera, so that the water tank can be completely imaged. The equipment layout is shown in Figure 2;

2) Place the moving rod in the initial position of the short axis as shown in Figure 1, and place the moving mechanism in the initial position of the long axis on the calibration rod. Use this position as the sampling starting point, and set the target point in the lower left corner of the calibration plate Take the first frame image as the origin of the world coordinate system;

3) Translate the moving mechanism 2 times in sequence along the long axis according to the position of the card slot, and the moving distance is 400mm, and shoot the 2nd to 3rd frame images respectively;

4) Keep the position of the moving mechanism, translate the moving rod 50mm along the short axis direction, place it in the next slot of the positioning track, and take the fourth frame of image;

5) Translate the moving mechanism in reverse order twice, the moving distance is 400mm, and take the 5th to 6th frame images;

6) Keep the position of the moving mechanism, translate the moving rod 50mm along the short axis direction, and repeat steps (2) to (5) until the last slot stops and complete the full-view sampling.

According to the installation layout of the equipment and the size of the calibration board, the upper limit of the movable distance of the calibration board from left to right is set to 1200mm, and the upper limit of the movable distance of the calibration rod from front to back is 850mm; after the above sampling process is completed, the sample library will have 51 pairs of 102 images , select the first 47 pairs as the training set, the 48th pair as the test set, and the last 3 pairs as the evaluation samples. Since each image has 49 target points (71), there are 2303 target points (71) in the first 47 pairs of images as the training set, the 48th pair of images has a total of 49 target points (71) as the test set, and the last 3 pairs of images A total of 147 target sites (71) were used as evaluation samples.

The calibration method of the binocular vision device includes an error evaluation process, and the method is as follows:

The position information of the calibration board is obtained through the find_caltab() operator that comes with the HALCON software. After the position of the calibration board is obtained, the operator find_marks_and_pose() is used to segment the circles in the area, and the number, perimeter, and coordinate position of the circles should be found. and consistent target spots in the calibration plate description file (71).

The pixel coordinates of the target point (71) can be obtained through the above-mentioned operator operations. Set the X axis of the world coordinate system to be parallel to the long axis of the fish tank, the Y axis to be perpendicular to the bottom of the fish tank, the Z axis to be parallel to the short axis of the fish tank, and the origin to be the lower right target point (71) of the starting position of the calibration plate, and the moving distance and From the spatial distribution of the target point (71), the coordinates of the world coordinate system of the target point (71) can be obtained. A total of 2352 target points (71) were effectively collected, the first 2303 points were selected as the training set, and the last 49 points were selected as the test set.

The genetic algorithm was used to optimize the parameters of the training set of the three coordinate systems. Considering the operation speed of the algorithm and the convergence of the fitness functions of different calibration models, the population size and evolutionary algebra of the three groups of optimization samples were uniformly selected as 60 and 200. The search ranges for parameters C and g are [1000, 10000], [0.0001, 0.000001], respectively. When the maximum number of iterations is reached, the optimal combination is output, and the optimized model parameters are obtained as shown in the following table.

A combination of axial error evaluation and single-point error evaluation is adopted. The axial error evaluation adopts the MSE function, and the single-point error evaluation adopts the P _MSE function, as follows:

In the formula, x', y', z' are the actual spatial coordinates of the measurement point, _xi ', y _i ', z _i ' are the prediction data of the measurement point model, and d _i is the actual measurement point on the x, y or z axis. Spatial coordinates, d _i ' are model prediction data in X, Y or Z axis for all measurement points.

In addition, since the μm measurement accuracy is already a very good measurement accuracy, the predicted value based on the training model is not affected by the actual measurement accuracy, and the predicted value can reach infinite digits. Therefore, in order to ensure the validity of the evaluation accuracy, the accuracy error is set three decimal places.

The basic principles and main features of the present invention and the advantages of the present invention have been shown and described above. Those skilled in the art should understand that the present invention is not limited by the above-mentioned embodiments. The above-mentioned embodiments and descriptions only illustrate the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will also have Various changes and modifications fall within the scope of the claimed invention. The claimed scope of the present invention is defined by the appended claims and their equivalents.

Claims (4)

1. an underwater stereoscopic observation system calibration device, it is characterized in that, comprise the calibration plate moving device that is arranged on the glass water tank (3), be fixed on the calibration plate inside the glass water tank (3) by the calibration plate moving device and set Binocular vision equipment outside the glass water tank (3); The calibration plate moving device comprises a positioning track (1), the positioning track (1) is fixed on the top of the glass water tank (3) through a hooking piece (5), and the positioning track (1) includes a set of relatively arranged track short shafts ( 6) and a set of track long shafts (7), a plurality of first card slots (61) are opened on the track short shafts (6), and both ends of the moving rod (4) are arranged in the positioning position through the first card slots (61). On the track short axis (6) of the track (1), several groups of second clamping grooves (41) are opened on the moving rod (4), and the top of the moving mechanism (2) passes through the second clamping groove (41) and the moving rod. (4) Removable connection, the calibration plate is fixedly connected with the moving mechanism (2); The position of the calibration plate in the glass water tank (3) is changed by the moving mechanism (2) and the moving rod (4), and the binocular vision device obtains the sampling image of the calibration plate at the current position. 2. underwater stereoscopic observation system calibration device according to claim 1, is characterized in that, the sampling process of binocular vision equipment is as follows: 1) Install binocular vision equipment, calibration board moving device and calibration board, adjust the position of the binocular vision equipment, so that the glass water tank can be completely imaged; 2) Place the moving rod (4) at the starting position of the short axis, place the moving mechanism (2) in the slot at the starting position of the long axis, take the position of the calibration plate at this time as the sampling starting point, and set the lower left corner of the calibration plate The target point is the origin of the world coordinate system, and the first frame image is taken; 3) The moving mechanism (2) is sequentially translated twice along the second card slot (41) of the moving rod (4), and the second to third frame images are captured respectively; 4) Keeping the position of the moving mechanism (2) still, move the rod (4) in translation along the short axis of the track (6) to the next first card slot (61), and shoot the 4th frame image; 5) The moving mechanism (2) is reversely translated twice along the second slot (41) of the moving rod (4), and the 5th to 6th frame images are taken; 6) Keep the position of the moving mechanism (2) still, translate the moving rod (4) to the next first slot (61) along the short axis of the track (6), repeat steps 2)-5) until the last first The position of the card slot (61) completes the sampling of the whole field of view. 3. underwater stereoscopic observation system calibration device according to claim 2, is characterized in that, the calibration method of described binocular vision equipment is: 1) Obtain the position information of the calibration board through the machine vision software, and after obtaining the position of the calibration board, use the operator to segment the circles in the area, and find the number, perimeter, coordinate position of the circle and the consistent target in the description file of the calibration board point(71); 2) Calculate the pixel coordinates of the target point (71); set the X axis of the world coordinate system to be parallel to the long axis of the fish tank, the Y axis to be perpendicular to the bottom of the fish tank, the Z axis to be parallel to the short axis of the fish tank, and the origin to be the right side of the starting position of the calibration plate. lowering the target point (71), and combining the moving distance of the calibration plate and the spatial distribution of the target point (71) to obtain the world coordinate system coordinates of the target point (71); 3) Divide the effectively collected target points (71) into a training set and a test set; use a genetic algorithm to optimize the parameters of the training set of the coordinate system, when the maximum number of iterations is reached, output the optimal combination to obtain an optimized model parameter; 4) Utilize the model parameters determined by the genetic algorithm to construct the SVR training model of the coordinate system, and predict three parameters in the world coordinate system; wherein, the target point (71) contains left and right image pixel coordinates and world coordinate information, and the image coordinate value The unit of is pixels, that is, the pixel coordinates of the target point (71) in the image. 4. underwater stereoscopic observation system calibration device according to claim 3, is characterized in that, the calibration method of described binocular vision equipment comprises error evaluation process, and its method is as follows: A combination of axial error evaluation and single-point error evaluation is adopted. The axial error evaluation adopts the MSE function, and the single-point error evaluation adopts the P _MSE function, as follows:

In the formula, x', y', z' are the actual spatial coordinates of the measurement point, _xi ', y _i ', z _i ' are the prediction data of the measurement point model, and d _i is the actual measurement point on the x, y or z axis. Spatial coordinates, d _i ' are model prediction data in X, Y or Z axis for all measurement points.

Images