Lightweight method for reducing calculated amount in background schlieren imaging of high-speed aircraft

Abstract

The invention discloses a lightweight method for reducing calculation amount in background schlieren imaging of a high-speed aircraft, and belongs to the field of visualization of an air disturbance flow field of the high-speed aircraft. Comprising the following steps: inputting a background image without an aircraft and a background schlieren image containing the aircraft; compressing the two-dimensional image into a one-dimensional signal through background differential projection, and determining a target frame; predicting the vertex coordinates and orientation angles of the aircraft in the flight direction by using the trained lightweight neural network in the normalized target frame; the predicted vertex coordinates and orientation angles of the aircraft are linearly mapped to an original image space, and original vertex coordinates and orientation angles of the aircraft are obtained; and determining a to-be-calculated region through linear segmentation based on the original vertex coordinates and the orientation angle of the aircraft. According to the method, the physical disturbance sensitive domain is dynamically locked through the core thought of lightweight flight target positioning, the vertex-orientation geometric model and picture linear segmentation, and the calculation load and the calculation power demand are greatly reduced on the premise that the shock wave form integrity is guaranteed.

Description

Technical Field

[0001] The present invention belongs to the field of visualization of air disturbance flow fields of high-speed aircraft, and more specifically, relates to a lightweight method for reducing the amount of calculation in background schlieren imaging of high-speed aircraft. Background Art

[0002] Schlieren imaging technology for high-speed aircraft visualizes the turbulent flow field by capturing the changes in air refractive index caused by the aircraft's shock wave, making it a core method for analyzing the aerodynamic characteristics of supersonic flight. With the growing demand for space-based detection and high-resolution imaging, the average daily data volume of the Schlieren system can reach terabytes. It is necessary to perform intensive optical flow calculations and other measures on the full-frame image sequence to extract particle displacements and obtain background Schlieren images. In the case of high data volumes, traditional full-frame processing modes consume a large amount of computing resources. Some experiments capturing supersonic shock waves in the context of solar flares have shown that the processing of single-experiment data requires long-term computing on large-scale computing clusters, which seriously restricts real-time analysis and engineering applications.

[0003] Existing computational optimization solutions primarily alleviate computing pressure by reducing image resolution or frame rate, but this can result in the loss of weak schlieren features. Because background schlieren techniques rely on the precise capture of particle displacements, low image resolution makes it difficult to obtain particle displacements algorithmically, making it impossible to obtain schlieren images. Another common approach is to utilize hardware such as GPUs for parallel accelerated computing. While this approach significantly improves processing speed, its high power consumption fundamentally conflicts with the strict power constraints of space-based platforms, such as satellites, and the computational effort remains essentially unchanged. To reduce inefficient computation, recent research has attempted to introduce motion region segmentation techniques. These methods rely on static background modeling or fixed threshold segmentation, resulting in high missegmentation rates in complex scenarios such as dynamic cloud interference and moving objects on the ground. Existing wind tunnel tests have shown that traditional segmentation algorithms have high response delays to the wake edges of high-speed aircraft, making them unable to meet the millisecond-level dynamic tracking requirements of high-speed aircraft. Furthermore, the segmentation algorithm itself adds additional computing power overhead. Summary of the Invention

[0004] In view of the defects of the prior art, the purpose of the present invention is to provide a lightweight method for reducing the amount of calculation in high-speed aircraft background schlieren imaging, aiming to solve the problem of excessive consumption of full-scale particle displacement calculation resources in high-speed aircraft background schlieren imaging.

[0005] To achieve the above object, the present invention provides a lightweight method for reducing the amount of calculation in background schlieren imaging of high-speed aircraft, comprising the following steps: Input the background image without aircraft and the background schlieren image with aircraft; compressing the background image without the aircraft and the background schlieren image with the aircraft into a one-dimensional signal by background differential projection, determining a target frame according to the one-dimensional signal, and normalizing the target frame; In the target frame normalized in the above steps, a coordinate system is defined, called the normalized coordinate system. The origin of this coordinate system is located at the upper left corner of the normalized target frame, the coordinate axis direction is consistent with the original image coordinate axis direction, and the coordinate range is equal to the size of the target frame; The trained lightweight neural network is used to predict the aircraft vertex coordinates and heading angle in the normalized target frame; Linearly map the predicted vertex coordinates and orientation angles of the aircraft to the original image space to obtain the original vertex coordinates and orientation angles of the aircraft; Based on the original vertex coordinates and orientation angle of the aircraft, a dividing line representing the maximum front of the shock wave is constructed. By calculating the coordinate value of the center point of the target frame relative to the position symbol Sc of the dividing line, all pixels with the same symbol Sc in the background schlieren image containing the aircraft are screened out, forming the area to be calculated.

[0006] The present invention also provides an electronic device, comprising: a computer-readable storage medium and a processor; The computer-readable storage medium is used to store executable instructions; The processor is configured to read the executable instructions stored in the computer-readable storage medium and execute the above method.

[0007] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to execute the above method.

[0008] The present invention also provides a computer program product, comprising a computer program or instructions, wherein the computer program or instructions implement the above method when executed by a processor.

[0009] Compared with the prior art, the above technical solutions conceived by the present invention can achieve the following beneficial effects: (1) The present invention proposes a lightweight adaptive computational region selection method architecture based on target positioning. This method dynamically locks the physical disturbance sensitive domain through the core ideas of lightweight flight target positioning, vertex-orientation geometric model, and linear segmentation of the frame. While ensuring the integrity of the shock wave morphology, it greatly reduces the computational load and significantly reduces the computing power requirements.

[0010] (2) This invention addresses the problem of high computational redundancy in full-frame two-dimensional traversal during target positioning. To avoid the resource consumption of the traditional sliding window method, this invention uses a background differential projection method to quickly locate the target frame. The sensitive area is located by using the pixel accumulation and difference between the image and the static background in the row and column directions, compressing the two-dimensional search problem into a one-dimensional signal analysis. This method is based on the density gradient perturbation characteristics caused by the movement of the aircraft. Through dynamic threshold segmentation, the projection differential extreme points are extracted and the aircraft boundary coordinates are quickly output. This method does not require sliding window traversal and the large amount of data collection, labeling, and model training required in deep learning methods.

[0011] (3) The present invention proposes a method for rapid vertex-orientation evaluation of high-speed flying targets based on fixed-size image input, with the goal of detecting the front vertex and instantaneous flight direction of the aircraft. By using a lightweight neural network in a normalized target frame, the vertices and orientation of the aircraft are efficiently predicted in combination with the lightweight neural network. The network input is an image in a normalized target frame that has been proportionally scaled to a smaller number of pixels (such as 64) on the long side and padded with zeros on the short side. By constructing an arbitrary neural network structure that can regress the vertex coordinates and orientation angles of the aircraft, the neural network is trained in combination with a simulated aircraft vertex and orientation data set. Since the input image size of the neural network is small (such as 64×64), the delay of this method is extremely small compared to traditional corner point detection and orientation detection methods. The vertex coordinates and orientation of the aircraft in the entire image can be obtained by combining the positioning coordinates of the flying target in the entire image.

[0012] (4) Determination of the area to be calculated based on simple linear segmentation. Based on the flight target vertex and orientation, the present invention constructs the equation of a line passing through the vertex that is consistent with the direction of the perpendicular line. Then, the area to be calculated is binary classified by determining whether the pixel coordinates in the image belong to the left or right side of the line. Compared with traditional image segmentation methods, this step requires minimal computation and greatly reduces computational latency. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 The present invention provides a flow chart of a lightweight method for reducing the amount of calculation in background schlieren imaging of high-speed aircraft.

[0014] Figure 2 This is Example 1 of the aircraft vertex, orientation, and mask of the area to be calculated in the first embodiment of the present invention.

[0015] Figure 3 This is Example 2 of the aircraft vertex, orientation, and mask of the area to be calculated in the first embodiment of the present invention. DETAILED DESCRIPTION

[0016] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0017] The specific process of this embodiment is a lightweight method for reducing the amount of calculation in background schlieren imaging of high-speed aircraft. Figure 1 As shown: Step 1: Based on the input background image Ib without aircraft and the background schlieren image I with aircraft, the background image without aircraft and the background schlieren image with aircraft are compressed into a one-dimensional signal by background differential projection, and a target frame is determined according to the one-dimensional signal; Step 2: Normalize the target frame to L×L, where L is the side length of the target frame (number of pixels). Preferably, L=64; Step 3: Use the trained lightweight neural network to determine the aircraft vertex (xp0, yp0) and heading angle (expressed as angle, the counterclockwise angle relative to the positive x-axis) within the normalized target box. Preferably, radians are used as units. Step 4: Linearly map the predicted aircraft vertex coordinates and heading angles to the original image space to obtain the original vertex coordinates (xp, yp) and heading angles of the aircraft; Step 5: Based on the original vertex coordinates (xp, yp) and orientation angle of the aircraft, construct a dividing line representing the maximum front of the shock wave. By calculating the coordinate value of the center point of the target frame relative to the position symbol Sc of the dividing line, all pixels with the same symbol Sc in the background schlieren image containing the aircraft are screened out, forming the area to be calculated.

[0018] Specifically, the specific process of determining the target frame using the dimensionality reduction method in step 1 is as follows: Step 1.1: Calculate DifX = abs(sumX(I) - sumX(Ib)) and DifY = abs(sumY(I) -sumY(Ib)), where abs() is the absolute value function, sumX(I) and sumY(I) are the results of summing the pixel values ​​of the background schlieren image I containing the aircraft in the horizontal (X-axis direction) and vertical (Y-axis direction) directions, respectively; sumX(Ib) and sumY(Ib) are the results of summing the pixel values ​​of the background image Ib without the aircraft in the horizontal (X-axis direction) and vertical (Y-axis direction), respectively. Where the aircraft appears, there will be obvious disturbances in the density gradient, and the projected differential signals DifX and DifY will have peaks in the corresponding rows and columns. By simply finding the peaks and valleys in a one-dimensional array, the approximate left and right boundaries (xa, xb) and upper and lower boundaries (ya, yb) of the aircraft can be determined, transforming the two-dimensional traversal problem into a one-dimensional threshold search to reduce the amount of computation and avoid the high computational complexity generated by the traditional full-frame sliding window method.

[0019] Step 1.2: Find the first coordinate (xa) and the last coordinate (xb) that satisfy the condition DifX > Th1 × max(DifX), where Th1 is a threshold between 0 and 1, preferably Th1 = 0.5, that is, the first and last coordinates that satisfy the condition that the DifX value is greater than half of its maximum value; Step 1.3: Find the first coordinate (ya) and the last coordinate (yb) that satisfy the condition DifY > Th1 × max(DifY). As in step 1.2, preferably, Th1 = 0.5, that is, the first and last coordinates that satisfy the condition that the DifY value is greater than half of its maximum value. Step 1.4: From xa, ya, find the first one that satisfies DifX <Th2×max(DifX) 及 DifY<Th2×max(DifY) 条件,其中Th2为0到1间的阈值,优选地Th2=0.1,即满足DifX值小于其最大值的十分之一以及DifY值小于其最大值的十分之一条件的坐标,记为x1,y1; Step 1.5: Find the first one that satisfies DifX from xb, yb onwards <Th2×max(DifX) 及 DifY<Th2×max(DifY) 条件,同步骤1.4一样,优选地,Th2=0.1,即满足 DifX值小于其最大值的十分之一以及DifY值小于其最大值的十分之一条件的坐标,记为x2,y2。

[0020] Th1 and Th2 can be adjusted and determined according to actual conditions.

[0021] Specifically, the specific process of normalizing the target frame to L×L in step 2 is: Step 2.1: Scale the image to L pixels on the long side and record the scaling factor a1 = Len / L, where Len is the original length of the long side of the image. Step 2.2: Fill the short side with 0 to L pixels from the right and bottom.

[0022] Specifically, the specific process of using the trained neural network model in step 3 to determine the aircraft vertex (xp0, yp0) and the heading angle (angle, the counterclockwise angle with the positive X axis) within the normalized target frame is as follows: Step 3.1: Select or construct a neural network model that can regress the vertex and orientation, such as the lightweight MobileNet model, ShuffleNet model, or SqueezeNet model. These models have a simple structure, low computational effort, and fast inference speed, making them suitable for efficient prediction of the aircraft's vertex and orientation. Step 3.2: Use the normalized bounding box data to train the neural network model. The training dataset can be constructed by overlaying multiple aircraft models of arbitrary orientations and positions on arbitrary image backgrounds. The aircraft vertices and orientations in each image are recorded as data labels. Because the neural network input image size is fixed, combined with this large amount of data, the resulting regression model has good versatility.

[0023] Step 3.3: Use the neural network model trained with the normalized target box data as input data to predict the aircraft vertex (xp0, yp0) and heading angle (angle, the counterclockwise angle relative to the positive x-axis).

[0024] Specifically, the specific process of calculating the image-level vertex coordinates (xp, yp) in step 4 is: Step 4.1: xp = xp0×a1 +x1; Step 4.2: yp = yp0×a1 +y1.

[0025] Specifically, the specific process of filtering out the area to be calculated based on vertices and orientations in step 5 is as follows: Step 5.1: Using the perpendicular line as the slope, find the equation of the line passing through the vertex (xp, yp), which can be expressed as: y−yp=tan(angle+pi / 2)⋅(x−xp) Where angle is the counterclockwise angle between the aircraft and the positive X-axis, tan() is the tangent function, and pi represents 180 radians.

[0026] Step 5.2: Convert the line to the general form of Ax + By + C = 0; Step 5.3: Substitute the center point of the target box (0.5×(x1+x2), 0.5×(y1+y2)) into the equation of the line and record the sign of the result. Step 5.4: For the point whose displacement is to be calculated, substitute it into the straight line equation. If the positive or negative value is the same as the center point of the target box, keep it; otherwise, discard it.

[0027] Example 1: This paper provides a lightweight method for reducing the computational complexity of background schlieren imaging for high-speed aircraft. This method is implemented in the Python programming language, using the OpenCV library for image processing and the PyTorch library for building and training a neural network model. The experimental data, including schlieren detection data from a high-speed aircraft, consists of 100 background images (Ib) and 100 background schlieren images (I).

[0028] The experimental results show that the present invention can accurately determine the area to be calculated, and the amount of calculation is reduced by about 64% on average compared with the traditional method. Figure 2 and Figure 3 ,in Figure 2 The general form of the perpendicular bisector equation is -41x - 287y + 225582 = 0, which reduces the computational complexity by 70.61%; Figure 3 The general form of the perpendicular bisector equation is 189x -13y - 54489 = 0, which reduces the computational complexity by 57.55%.

[0029] It will be easily understood by those skilled in the art that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A lightweight method for reducing the amount of calculation in background schlieren imaging of high-speed aircraft, characterized in that: It includes the following steps: Input the background image without the aircraft and the background schlieren image with the aircraft; Compress the background image without the aircraft and the background schlieren image with the aircraft into one-dimensional signals through background differential projection, determine the target box according to the one-dimensional signals, and normalize the target box; Predict the vertex coordinates and orientation angle of the aircraft in the flight direction within the normalized target box using a trained lightweight neural network; Linearly map the predicted vertex coordinates and orientation angle of the aircraft to the original image space to obtain the original vertex coordinates and orientation angle of the aircraft; Based on the original vertex coordinates and orientation angle of the aircraft, construct a dividing line representing the maximum front of the shock wave. By calculating the position symbol Sc of the coordinate value of the center point of the target box relative to this dividing line, filter out all pixel points in the background schlieren image with the aircraft that have the same symbol as Sc, which constitutes the area to be calculated.

2. The method according to claim 1, characterized in that The compression of the background image without the aircraft and the background schlieren image with the aircraft into one-dimensional signals through background differential projection includes: summing the pixel values of the background image without the aircraft and the background schlieren image with the aircraft in the horizontal and vertical directions respectively, taking the difference between the summation results in the horizontal and vertical directions respectively, and taking the absolute value of the difference to obtain the differential signal DifX in the horizontal direction and the differential signal DifY in the vertical direction as the compressed one-dimensional signals.

3. The method according to claim 2, characterized in that The determination of the target box according to the one-dimensional signals includes: Find the first coordinate xa and the last coordinate xb that satisfy the condition DifX > Th1×max(DifX), where Th1 is a threshold between 0 and 1, and max(DifX) is the maximum value of DifX; Find the first coordinate ya and the last coordinate yb that satisfy the condition DifY > Th1×max(DifY), where max(DifY) is the maximum value of DifY; Find the first coordinate that satisfies DifX < Th2×max(DifX) and DifY < Th2×max(DifY) from xa and ya forward, denoted as x1 and y1, where Th2 is a threshold between 0 and 1; Find the first coordinate that satisfies DifX < Th2×max(DifX) and DifY < Th2×max(DifY) from xb and yb backward, denoted as x2 and y2; Determine the target box from x1 and y1, x2 and y2.

4. The method according to claim 3, characterized in that The prediction of the vertex coordinates and orientation angle of the aircraft in the flight direction within the normalized target box using a trained lightweight neural network includes: Select a lightweight neural network; Use the training data within the normalized target box to train the neural network model; the training dataset is constructed by superimposing various aircraft models with arbitrary orientations and positions on an arbitrary image background, and record the vertices and orientations of the aircraft in each image as data labels; Use the normalized target box data as input and input it into the trained lightweight neural network to predict the vertex (xp0, yp0) and orientation angle of the aircraft.

5. The method according to claim 4, characterized in that The original vertex coordinates (xp, yp) of the aircraft are expressed as: xp = xp0×a1 +x1 yp = yp0×a1 +y1 The original heading angle of the aircraft remains unchanged.

6. The method according to claim 5, characterized in that The original vertex coordinates and orientation angle of the aircraft are used to construct a boundary line representing the maximum front of the shock wave. By calculating the coordinate value of the center point of the target frame relative to the position symbol Sc of the boundary line, all pixels with the same symbol Sc in the background schlieren image containing the aircraft are screened out, forming the area to be calculated, including: Taking the perpendicular line toward the angle as the slope, we get the equation of the line passing through the vertex (xp, yp), which is expressed as: y−yp=tan(angle+pi / 2)⋅(x−xp) Where angle is the counterclockwise angle between the aircraft and the positive X axis, tan() is the tangent function, and pi represents 180 radians. Convert the line to the general form of Ax + By + C = 0; Substitute the center point of the target frame (0.5×(x1+x2), 0.5×(y1+y2)) into the straight line equation and record the sign of the result. For the point whose displacement is to be calculated, substitute it into the straight line equation. If the positive or negative value is the same as the center point of the target box, it is retained; otherwise, it is discarded.

7. An electronic device, characterized in that: include: Computer-readable storage medium and processor; The computer-readable storage medium is used to store executable instructions; The processor is configured to read the executable instructions stored in the computer-readable storage medium and execute the method according to any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to execute the method according to any one of claims 1 to 6.

9. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images