An Automatic Interpretation Method for Measuring Images of a High-Speed Small Target Photoelectric Theodolite

Abstract

The present invention discloses an automatic interpretation method for the measurement images of a high-speed small target theodolite, which can improve the automatic interpretation efficiency of high-speed small targets. The present invention uses the second difference of the interpretation positions of the previous several frames as weights for correction, and combines the target image windowing position prediction method based on the actual rotation angle of the main optical axis of the optoelectronic theodolite to formulate the prediction principles in the case of target loss and out-of-field of view, ensuring the accuracy and stability of the predicted position while guaranteeing the prediction efficiency. By adopting the change angle amount of the main optical axis of the theodolite, a prediction method for the windowing position of the target recognition in the measurement image is proposed, reducing the number of target searches, which has great significance for improving the automatic interpretation efficiency of high-speed small targets in the theodolite measurement images.

Description

Technical Field

[0001] The present invention belongs to the field of image measurement, and particularly relates to an automatic interpretation method for measurement images of a high-speed small target optoelectronic theodolite. Background Art

[0002] When an optoelectronic theodolite tracks and shoots a target to be measured, the optoelectronic theodolite must be continuously rotated, resulting in a continuously changing shooting background. Therefore, the measurement image of the optoelectronic theodolite must be an image in which both the target and the background are continuously changing. Since high-speed small targets are small, have no texture, and have large morphological changes, it is difficult to use traditional methods such as difference, matching, and centroid. Only the entire image can be used for tracking, identification, and positioning. However, since the measurement image generally has a large resolution, the efficiency of automatic interpretation is very low. The general solution is to open a tracking window in the image. The simple method is to use the position of the target in the previous few frames of images for fitting extrapolation and position difference to predict the position of the window opening in the next frame of image. However, in a complex measurement environment, the target imaging is weak, and there are often large errors in the interpreted position or direct loss in the previous few frames. Therefore, the method of using fitting extrapolation and position difference for prediction has problems of low efficiency caused by fitting and unstable predicted position caused by position difference, and the practicability is poor. Coupled with the fact that the optoelectronic theodolite itself is not a staring shot, there is a problem of relative movement between the background and the target. Therefore, direct prediction is almost impossible.

[0003] It can be seen that currently, the automatic tracking, identification, and positioning of high-speed small targets in the measurement image of the theodolite generally use the entire image. Since the measurement image generally has a large resolution, whether the centroid method or the matching method is used, there will be a huge amount of data, resulting in a current pure software interpretation efficiency of only about 10 frames. Therefore, the existing technologies generally have the disadvantages of large computational amount and low efficiency. Summary of the Invention

[0004] In view of this, the present invention provides an automatic interpretation method for measurement images of a high-speed small target optoelectronic theodolite, which can improve the automatic interpretation efficiency of high-speed small targets.

[0005] To achieve the above object, the technical solution of the present invention is as follows:

[0006] An automatic interpretation method for measurement images of a high-speed small target theodolite of the present invention includes the following steps:

[0007] Statistically calculate the first-order difference mean of the azimuth and elevation comprehensive angle measurements of the first three frames, use this mean to predict the comprehensive angle of the next frame of image, subtract the azimuth and elevation angles of the main optical axis of the theodolite at the moment of this frame of image from the azimuth and elevation comprehensive angles of this frame of image to obtain the offset angle amount of the target from the center of the image in this frame of image, and use the sensor dimension to obtain the pixel deviation displacement of the target from the center of the image.

[0008] When the target is lost within the frame, take the first-difference mean and the second-difference mean of the comprehensive angles of the interpreted targets. The sum of these two means is the amount by which the target deviates from the image center, and the product of this center amount and the number of frames of the lost image is used as the current frame offset;

[0009] If the sum of the first differences between the comprehensive angle of the last frame of the lost target and the previous n frames is greater than the comprehensive angle value of this frame, then the target is outside the image; otherwise, it is inside the image;

[0010] When the target is outside the frame, use the encoder values of the previous n frames for prediction.

[0011] Among them, the prediction formula for the target in the image is:

[0012]

[0013]

[0014] Among them, x n 、y n are the predicted positions of the target in the image, a n-2 、b n-2 、a n-1 、b n-1 are the encoder azimuth and pitch angle measurements of the previous two frames and the previous frame, x n-2 、y n-2 、x n-1 、y n-1 are the miss distances of the previous two frames and the previous frame, d x 、d y are the sensor dimensions.

[0015] Among them, when the target is lost within the frame, the specific operations are as follows:

[0016]

[0017]

[0018]

[0019]

[0020]

[0021]

[0022]

[0023]

[0024] Among them, is the first-order difference of the azimuth angle of the first n frames of images, is the second-order difference of the pitch angle of the first n frames of images, x n , y n is the predicted target position in the current frame of image.

[0025] Among them, the calculation formula for judging whether the target is within the frame is as follows:

[0026]

[0027]

[0028]

[0029]

[0030] If a1 > a2, the target is outside the image; if a1 < a2, the target is inside the image.

[0031] Among them, when the target is outside the frame, the encoder value of the first n frames of images is used for prediction. The specific formula is as follows:

[0032]

[0033]

[0034]

[0035]

[0036]

[0037]

[0038] Among them, is the first-order difference of the azimuth angle of the first n frames of images, a and b are the last frames of images without lost targets, x n , y n is the predicted target position in the current frame of image.

[0039] Beneficial effects

[0040] 1. The present invention uses the second-order difference of the judgment position of the previous few frames as the weight for correction, and combines the target image windowing position prediction method of the actual rotation angle of the main optical axis of the photoelectric theodolite to formulate the prediction principle in the case of target loss and out-of-field of view, ensuring the accuracy and stability of the predicted position while guaranteeing the prediction efficiency.

[0041] 2. The method proposed by the present invention can perform windowing on images, reducing the computational amount, achieving the technical effect of improving the efficiency of automatic interpretation, realizing efficient automatic interpretation of high-speed small targets, with small prediction errors, high execution efficiency, simple algorithms, and easy implementation.

[0042] 3. The present invention uses the method of image windowing to reduce the workload. The windowing position is very crucial. Since the theodolite measurement image is an image where both the target and the background change simultaneously, it is very difficult to achieve using the traditional method of predicting the target movement trajectory. Therefore, the present invention uses the change angle of the main optical axis of the theodolite to predict the windowing position, reducing the number of target searches, which has great significance for improving the efficiency of automatic interpretation of high-speed small targets in theodolite measurement images. Brief Description of the Drawings

[0043] Figure 1 It is a schematic diagram of the measurement image of the present invention. Detailed Embodiment

[0044] The following takes embodiments in conjunction with the drawings and describes the present invention in detail.

[0045] The automatic interpretation method for theodolite measurement images of high-speed small targets of the present invention includes the following parts:

[0046] The angle measurement of the photoelectric theodolite reflects the rotation angle of the main mirror optical axis, that is, the center of the measurement image. The reason for the failure of the fixed target position prediction method is essentially due to a large offset in the image center. Therefore, before prediction, the position of the target in the image must be corrected using the angle measurement of the photoelectric theodolite. Secondly, since the current photography frequency of the photoelectric theodolite is relatively high, the target spatial position generally changes little between two frames of images. Therefore, it can be assumed that the target spatial position between adjacent two frames of images remains unchanged, that is, the change amount of the target spatial position in the previous two frames of images is used as the change amount of the target spatial position between this frame and the previous frame of image. Then the prediction formula of the target in the image is:

[0047]

[0048]

[0049] where, x n 、y n are the predicted positions of the target in the image, a n-2 、b n-2 、a n-1 、b n-1 are the azimuth and elevation angle measurements of the encoder for the previous two frames and the previous frame, x n-2 、y n-2 、x n-1 、y n-1 are the miss distances for the previous two frames and the previous frame, dx , d y is the sensor dimension. Since the angular change between two adjacent frames needs to be obtained, prediction must start from the third frame. In practical applications, it has been proven that for a normal theodolite, the general prediction accuracy is within a few pixels. The schematic diagram of the measurement image of the present invention is as shown in Figure 1 . The specific method for predicting the window position is as follows: Calculate the mean value of the first-order differences of the azimuth and elevation combined angle measurements of the first three frames, use this mean value to predict the combined angle of the next frame of image, subtract the azimuth and elevation angles of the main optical axis of the theodolite at the moment of this frame of image from the combined azimuth and elevation angles of this frame of image to obtain the offset angle of the target from the center of the image in this frame of image, and use the sensor dimension to obtain the pixel offset displacement of the target from the center of the image.

[0050] Since the infrared radiation characteristics of small high-speed targets are weak, the gray value presented is relatively small, and the number of pixels included in the target is also small. Therefore, it is often blocked by clouds, waves, etc., resulting in the loss of the target within several frames. When the target is lost inside the frame, take the mean value of the first-order differences and the mean value of the second-order differences of the combined angles of the already interpreted targets. The sum of these two mean values is the amount of deviation of the target from the center of the image, and the product of this center amount and the number of lost image frames is used as the current frame offset. The specific operations are as follows:

[0051]

[0052]

[0053]

[0054]

[0055]

[0056]

[0057]

[0058]

[0059] Among them, is the first-order difference of the azimuth angle of the first n frames of images, is the second-order difference of the elevation angle of the first n frames of images, x n , y n is the predicted target position in the current frame of image.

[0060] If the sum of the combined angle of the last frame of the lost target and the first-order differences of the first n frames is greater than the combined angle value of this frame, then the target is outside the image; otherwise, it is inside the image. The calculation formula is as follows:

[0061]

[0062]

[0063]

[0064]

[0065] If a1 > a2, the target is outside the image. If a1 < a2, the target is inside the image.

[0066] When the target is outside the frame, that is, when the target is completely lost, at this time, the photoelectric theodolite can no longer track stably. Therefore, the encoder values of the previous n frames of images are used for prediction. The specific formula is as follows:

[0067]

[0068]

[0069]

[0070]

[0071]

[0072]

[0073] Where is the first difference of the azimuth angle of the previous n frames of images, a and b are the last frame of images of the non-lost target, and x n , y n is the predicted target position in the current frame of image.

[0074] In summary, the above are only the preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An automatic interpretation method for the measurement images of a high-speed small target theodolite, characterized in that, It includes the following steps: Statistically calculate the mean of the first differences in the combined azimuth and elevation angle measurements for the first three frames. Use this mean to predict the combined angle of the next frame. Subtract the azimuth and elevation angles of the main optical axis of the theodolite at the moment of this frame from the combined azimuth and elevation angle of this frame image to obtain the offset angle by which the target in this frame image deviates from the center of the image. Use the sensor dimension to obtain the pixel deviation displacement of the target from the center of the image; When the target is lost within the frame, take the mean of the first differences and the mean of the second differences of the combined angles of the already interpreted target. The sum of these two means is the amount by which the target deviates from the center of the image. Multiply this center amount by the number of frames in which the image is lost to obtain the current frame offset; If the sum of the first differences between the combined angle of the last frame in which the target is lost and the previous n frames is greater than the combined angle value of this frame, then the target is outside the image; otherwise, it is inside the image; When the target is outside the frame, use the encoder values of the previous n frames for prediction.

2. The method according to claim 1, wherein The prediction formula for the target in the image is: where x n and y n are the predicted positions of the target in the image, a n-2 and b n-2 and a n-1 and b n-1 are the azimuth and pitch angles of the encoders of the previous two frames and the previous frame, x n-2 and y n-2 and x n-1 and y n-1 are the miss distances of the previous two frames and the previous frame, and d x and d y are the sensor dimensions.

3. The method according to claim 2, wherein When the target is lost within the frame, the specific operations are as follows: Among them, is the first difference of the azimuth angle of the first n frames of images, is the second difference of the pitch angle of the first n frames of images, x n , y n is the predicted target position in the current frame of image.

4. The method according to claim 3, characterized in that, The calculation formula for determining whether the target is within the frame is as follows: If a1 > a2, then the target is outside the image; if a1 < a2, then the target is inside the image.

5. The method according to claim 3 or 4, characterized in that When the target is outside the frame, use the encoder values of the previous n frames for prediction. The specific formula is as follows: Among them, is the first-order difference of the azimuth angle of the first n frames of images, a and b are the last frames of images of the non-lost targets, and x n , y n is the predicted target position in the current frame of image.

Images