Method for improving target observation capability of target range laser high-speed reflection sequence imaging

Abstract

The invention discloses a pulse laser monocular high-speed sequence image target extraction and inter-frame dodging method, which is mainly used for improving the target observation capability of target range laser reflection imaging. The method includes: acquiring a high-speed flight target laser sequence reflection image; converting each frame of image from an RGB space to an HSV space to obtain a difference image; obtaining a target initial miss distance, taking the target initial miss distance as a center, windowing to obtain a small image, segmenting a target from the small image by using OTSU, obtaining a target mask, and calculating a centroid coordinate of the target mask as a final output miss distance; performing point multiplication on the target mask and the image V component to obtain a target image, and calculating a brightness factor and an illumination compensation factor of the target image; the V component Retinex is decomposed into an illumination layer and a reflecting layer, illumination compensation is achieved on the illumination layer of each frame of image through a gamma transformation method, a new illumination layer is generated, and the index is an illumination compensation factor; and generating a new V component, and turning back to the RGB space.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing of high-speed moving targets, and in particular to a method for improving the target observation capability of a target range laser high-speed reflection sequence imaging. Background Art

[0002] With the continuous development of digital image processing technology, high-speed sequence image target extraction and enhancement technology is also constantly developing and improving. At present, the commonly used target segmentation method is OTSU. The OTSU algorithm, also known as the maximum inter-class variance method (Otsu method), is a method for determining the threshold. The OTSU algorithm is simple to calculate and is not affected by the brightness and contrast of the image, so it has been widely used in digital image processing. It divides the image into two parts, background and foreground, according to the grayscale characteristics of the image. However, the OTSU algorithm is affected by the number of target and background pixels, which will lead to inaccurate segmentation. In addition, due to the non-uniformity of the light field, the brightness of each frame of the sequence image is different. Commonly used image enhancement methods are histogram mapping method and model method. Both methods have a certain enhancement effect on a single image, but do not consider the brightness relationship between frames. It is difficult to ensure the brightness consistency between sequence image frames by enhancing each frame. For example, Chinese patent 2014107356017 discloses "High-speed motion target extraction method based on motion vector", and the authorization announcement number is CN104463910B. This method first directly extracts motion vectors from the video stream; then normalizes the motion vectors and estimates the global motion parameters of the scene; finally, by analyzing the statistical characteristics of the motion vectors, rapid detection of high-speed moving targets is achieved. This method can directly utilize the motion information contained in the motion vectors in the compressed video data, greatly improving the speed of extracting moving targets from compressed videos, and effectively improving the performance of high-speed moving target detection in a dynamic camera environment. However, this method still cannot solve the problem of inconsistent brightness between frames. When used for target extraction in pulsed laser monocular high-speed sequence images, it will lead to inaccurate target segmentation and uneven illumination of the target sequence. Summary of the invention

[0003] In order to solve the problem of accurate target extraction and inter-frame brightness consistency, the present invention proposes a method for improving the target observation capability of the target range laser high-speed reflection sequence imaging, accurately extracting the target area and improving the accuracy of the miss distance calculation. At the same time, it solves the problem of illumination changes to the target during flight and improves the brightness consistency between frames.

[0004] The present invention adopts the following technical solutions:

[0005] The method for improving the target observation capability of the target range laser high-speed reflection sequence imaging described in the present invention comprises the following steps:

[0006] 1) Pulse laser monocular high-speed photography to obtain the laser sequence reflection image F of the high-speed flying targeti (x,y)(0<i≤N);

[0007] 2) Convert each frame of image from RGB space to HSV space, and subtract the V component of the i-1 frame from the V component of the i-th frame to obtain the difference image D i (x,y)(1<i≤N);

[0008] 3) Using adaptive threshold segmentation to segment the difference image to obtain the initial off-target amount;

[0009] 4) In the V component, take the initial miss distance as the center, open a square window, and obtain a small image P i (x, y)(0<i≤N), use OTSU to segment the small image to obtain the image target mask, and calculate its centroid coordinates as the final miss distance;

[0010] 5) The target image K is obtained by multiplying the image target mask with the V component, and the brightness factor and illumination compensation factor of the target image K are calculated;

[0011] 6) Retinex decomposes the V component of each frame into an illumination layer and a reflection layer. The illumination layer is compensated for by gamma transformation to generate a new illumination layer, where the exponent is the illumination compensation factor. The new illumination layer and the reflection layer are multiplied to generate a new V component, which is then converted from the HSV space back to the RGB space.

[0012] 7) Repeat the above steps 2)-6), output the final miss distance of each frame image and change the brightness of each frame image.

[0013] In step 3 above)

[0014] The adaptive threshold T0 is calculated using the following formula:

[0015]

[0016] Among them, G t is the maximum grayscale value of the differential image, G b It is the gray value with the largest number of pixels in the differential image or the average gray value of the differential image.

[0017] In step 3 above)

[0018] Firstly, the initial target area is obtained by using adaptive threshold segmentation on the difference image, and the centroid coordinates of the initial target area are calculated, and the centroid of the initial target area is taken as the initial miss distance.

[0019] In step 4 above)

[0020] The size of the square window is M*M, where M is the value obtained by adding the pixel value of the longest side and the pixel value of the shortest side of the initial target area obtained by segmenting the differential image using an adaptive threshold.

[0021] In step 5 above)

[0022] Calculate the brightness factor of the target image K according to the following formula:

[0023]

[0024] Where Lf represents the brightness factor, X and Y are the pixels of the target area, A is the target area, and L(x,y) is the grayscale value at the coordinate (x,y);

[0025] The illumination compensation factor of the target image K is calculated according to the following formula

[0026]

[0027] Where Cf represents the illumination compensation factor.

[0028] Compared with the prior art, the present invention has the following beneficial effects: achieving brightness consistency between frames and enhancing visual effects. The target area can be extracted more accurately, and the accuracy of miss distance calculation is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 is a flow chart of an embodiment of the present invention. DETAILED DESCRIPTION

[0030] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0031] The present invention provides a method for improving the target observation capability of the target range laser high-speed reflection sequence imaging, which is used to enable the user to extract the flying target obtained by laser high-speed photography and enhance the brightness consistency of the image sequence. The process of the method is as follows Figure 1 As shown, the following steps are included:

[0032] 1) Pulse laser monocular high-speed photography to obtain the laser sequence reflection image F of the high-speed flying target i (x,y)(0<i≤N);

[0033] 2) Convert each frame of image from RGB space to HSV space, and subtract the V component of the i-1 frame from the V component of the i-th frame to obtain the difference image D i (x,y)(1<i≤N);

[0034] 3) Using adaptive threshold segmentation to segment the difference image to obtain the initial off-target amount;

[0035] In this step, the adaptive threshold T0 is calculated using the following formula:

[0036]

[0037] Among them, G t is the maximum grayscale value of the differential image, G b It is the gray value with the largest number of pixels in the differential image or the average gray value of the differential image.

[0038] Firstly, the initial target area is obtained by using adaptive threshold segmentation on the difference image, and the centroid coordinates of the initial target area are calculated, and the centroid of the initial target area is taken as the initial miss distance.

[0039] 4) In the V component, take the initial miss distance as the center, open a square window, and obtain a small image P i (x, y)(0<i≤N), use OTSU to segment the small image to obtain the image target mask, and calculate its centroid coordinates as the final miss distance;

[0040] The size of the square window is M*M, where M is the value obtained by adding the pixel value of the longest side and the pixel value of the shortest side of the initial target area obtained by segmenting the differential image using an adaptive threshold.

[0041] 5) The target image K is obtained by multiplying the image target mask with the V component, and the brightness factor and illumination compensation factor of the target image K are calculated;

[0042] Calculate the brightness factor of the target image K according to the following formula:

[0043]

[0044] Where Lf represents the brightness factor, X and Y are the pixels of the target area, A is the target area, and L

[0045] (x,y) is the grayscale value at coordinate (x,y);

[0046] The illumination compensation factor of the target image K is calculated according to the following formula

[0047]

[0048] Where Cf represents the illumination compensation factor.

[0049] 6) Retinex decomposes the V component of each frame into an illumination layer and a reflection layer. The illumination layer is compensated for by gamma transformation to generate a new illumination layer, where the exponent is the illumination compensation factor. The new illumination layer and the reflection layer are multiplied to generate a new V component, which is then converted from the HSV space back to the RGB space.

[0050] Specifically, the Retinex decomposition of the V component can be expressed by the following formula:

[0051]

[0052] The calculation of the new illumination layer can be expressed by the following formula;

[0053]

[0054] Finally, the new illumination layer and the reflection layer are dotted to generate a new V component, which is then converted from the HSV space back to the RGB space.

[0055] 7) Repeat the above steps 2)-6), output the final miss distance of each frame image and change the brightness of each frame image.

[0056] The present invention is specifically described in detail using a pulsed laser monocular high-speed sequence image with a W*H resolution.

[0057] Step 1: Pulse laser high-speed photography to obtain the laser sequence reflection image F of the high-speed flying target i (x,y)(0<i≤10);

[0058] Step 2: Convert each frame of the image from RGB space to HSV space to obtain an image frame expressed by three components: hue (H), saturation (S), and brightness (V). Subtract the V component of the i-1 frame from the V component of the i-th frame to obtain the differential image D. i (x,y)(1<i≤9);

[0059] Step 3: Generate an adaptive threshold segmentation target using the maximum grayscale value and average grayscale value of the differential image, obtain the initial target area, calculate the centroid coordinates of the initial target area, and use the centroid of the initial target area as the initial miss amount.

[0060] Step 4: In the V component, open a window of 81*81 with the initial miss amount as the center. Use OTSU to segment the target in this small window and generate a target area mask; and calculate its centroid coordinates as the final miss amount;

[0061] Step 5: Multiply the target area mask with the image frame to obtain the target image, and calculate the brightness factor and illumination compensation factor;

[0062] Step 6: Retinex the V component, perform gamma transformation on the illumination layer of each frame, with the exponent being the illumination compensation factor, to generate a new illumination layer; multiply the new illumination layer and the reflection layer by points to generate a new V component, which is then converted from the HSV space back to the RGB space.

[0063] Step 7: Repeat the above steps 2 to 6 to output the final miss distance of each frame image and change the brightness of each frame image.

Claims

1. A method for improving the target observation capability of target range laser high-speed reflection sequence imaging, characterized in that: The following steps are involved: 1) Pulse laser monocular high-speed photography to obtain the laser sequence reflection image F of the high-speed flying target i (x,y)(0<i≤N); 2) Convert each frame of image from RGB space to HSV space, and subtract the V component of the i-1 frame from the V component of the i-th frame to obtain the difference image D i (x,y)(1<i≤N); 3) Using adaptive threshold segmentation to segment the difference image to obtain the initial off-target amount; 4) In the V component, take the initial miss distance as the center, open a square window, and obtain a small image P i (x, y)(0<i≤N), use OTSU to segment the small image to obtain the image target mask, and calculate its centroid coordinates as the final miss distance; 5) The target image K is obtained by multiplying the image target mask with the V component, and the brightness factor and illumination compensation factor of the target image K are calculated; 6) Retinex decomposes the V component of each frame into an illumination layer and a reflection layer. The illumination layer is compensated for by gamma transformation to generate a new illumination layer, where the exponent is the illumination compensation factor. The new illumination layer and the reflection layer are multiplied to generate a new V component, which is then converted from the HSV space back to the RGB space. 7) Repeat the above steps 2)-6), output the final miss distance of each frame image and change the brightness of each frame image.

2. According to claim 1, a method for improving the target observation capability of target range laser high-speed reflection sequence imaging is characterized in that: In step 3) The adaptive threshold T0 is calculated using the following formula: Among them, G t is the maximum grayscale value of the differential image, G b It is the gray value with the largest number of pixels in the differential image or the average gray value of the differential image.

3. A method for improving the target observation capability of target range laser high-speed reflection sequence imaging according to claim 1 or 2, characterized in that: In step 3) Firstly, the initial target area is obtained by using adaptive threshold segmentation on the difference image, and the centroid coordinates of the initial target area are calculated, and the centroid of the initial target area is taken as the initial miss distance.

4. A method for improving the target observation capability of target range laser high-speed reflection sequence imaging according to claim 1 or 2, characterized in that: In step 4) The size of the square window is M*M, where M is the value obtained by adding the pixel value of the longest side and the pixel value of the shortest side of the initial target area obtained by segmenting the differential image using an adaptive threshold.

5. A method for improving the target observation capability of target range laser high-speed reflection sequence imaging according to claim 1 or 2, characterized in that: In step 5) Calculate the brightness factor of the target image K according to the following formula: Where Lf represents the brightness factor, X and Y are the pixels of the target area, A is the target area, and L(x,y) is the grayscale value at the coordinate (x,y); Calculate the illumination compensation factor of the V component according to the following formula Where Cf represents the illumination compensation factor.

Images