Multi-definition power transmission channel real-time monitoring image compression method and device

Abstract

The invention discloses a multi-definition power transmission channel real-time monitoring image compression method and device, and the method comprises the steps: carrying out the edge calculation of a real-time image obtained based on an image collection device installed on a site, and recognizing the hidden danger in the image; image information high-definition transmission of the hidden danger area; singular values of the original image are decomposed and transformed, the singular values are ranked from large to small, and the sequence of a left singular vector and a right singular vector is adjusted; sequentially transmitting data according to the sequence; and the background receives data and reconstructs an image. According to the invention, high-efficiency compression of the power transmission channel image is realized, information integrity of the hidden danger area image concerned by operation and maintenance personnel is maintained, progressive reconstruction can be carried out at the background according to the data receiving sequence, the background image is gradually clear, and the method is suitable for scenes with weak communication signals and high image data transmission requirements. And the high-efficiency transmission requirement of real-time monitoring images of field power transmission lines can be met.

Description

Technical Field

[0001] The present invention relates to the technical field of image compression, and specifically to a real-time monitoring image compression method and device for a transmission line with multiple resolutions, which adopts different resolution compression methods for different regions of the transmission line monitoring image to reduce the difficulty of wireless image transmission. Background Art

[0002] With the rapid growth of the scale of overhead transmission lines in China, the visualization technology of transmission line channels has been widely applied. This technology realizes the real-time monitoring and intelligent analysis of transmission lines and channel environments through advanced technical means such as visualization image acquisition, image recognition, Internet of Things, and artificial intelligence based on edge devices, improves the safety and efficiency of transmission line operation and maintenance, reduces the cost of manual inspection, and timely discovers potential risks such as channel wildfires, mechanical construction, natural disasters, and foreign objects on conductors.

[0003] At present, the intelligent identification technology for potential risks in transmission line channels at the device end is becoming increasingly mature, and the recognition results are generally transmitted back to the background in the form of images through a wireless communication network for operation and maintenance personnel to judge and formulate targeted measures. Since a large number of transmission lines are located in the wild mountains and the communication network signal is weak, the image transmission time is long and the transmission efficiency is low, which cannot meet the timeliness requirements of backend judgment. On the other hand, compared with a large amount of background information in the image, operation and maintenance personnel are more concerned about identified dangerous sources such as construction machinery and wildfires. Therefore, the present invention proposes a real-time monitoring image compression method for a transmission line with multiple resolutions to solve the problems existing in the prior art. Summary of the Invention

[0004] The purpose of the present invention is to provide a real-time monitoring image compression method and device for a transmission line with multiple resolutions, which makes up for the deficiencies of low image transmission efficiency and large amount of invalid information in the case of weak field communication signals, and has the advantages of no distortion in key image areas, high compression efficiency, and progressive transmission.

[0005] To achieve the above purpose, the present invention includes the following steps:

[0006] A real-time monitoring image compression for a transmission line with multiple resolutions includes the following steps:

[0007] (1) Based on the real-time image obtained by the image acquisition device installed on site, edge computing identifies potential risk information in the real-time image;

[0008] (2) Without compressing the image information in the potential risk information, directly transmit it to the background;

[0009] (3) Perform singular value decomposition transformation on the original image to obtain singular values, left singular vectors, and right singular vectors;

[0010] (4) Sort the singular values from largest to smallest, and correspondingly adjust the orders of the left singular vectors and the right singular vectors;

[0011] (5) Transmit the corresponding floating-point number sequences to the background in the order of singular values, left singular vectors, and right singular vectors;

[0012] (6) The background receives the uncompressed image transmitted in step (2) and the floating-point number sequence data transmitted in step (5), calculates the RGB color values of each pixel point, and reconstructs the image.

[0013] Further, for the real-time image obtained by the image acquisition device installed on-site in step (1), the specific steps for edge computing to identify the hidden danger information in the real-time image are as follows: The image obtained by the on-site image acquisition device is a color image of m×n pixel points, and the RGB color values of each pixel point are stored, occupying 24 bits; A model trained based on deep learning is used to intelligently identify various hidden danger information in the image, marked with a rectangular box, and the coordinates of the diagonal points of the rectangular box (x1, y1) and (x2, y2) are recorded.

[0014] Further, the specific steps for directly transmitting the image information in the hidden danger information to the background without compression in step (2) are as follows: For the rectangular box located at (x1, y1) to (x2, y2) in the image described in step (1), transmit the coordinates of the corner points of the rectangular box and the RGB color values of each pixel point within the area, and complete the transmission of all rectangular box information in sequence.

[0015] Further, the specific steps for performing singular value decomposition transformation on the original image to obtain singular values, left singular vectors, and right singular vectors in step (3) are as follows: Decompose the original image into 3 matrices of size m×n according to the RGB color values, denoted as the R matrix, the G matrix, and the B matrix. The singular value decomposition form of each matrix is A = UΣV T , where Σ is a (m×n) diagonal matrix, and the diagonal elements are non-negative singular values; U is a (m×m) orthogonal matrix, and the column vectors are left singular vectors; V T is the transpose of an (n×n) orthogonal matrix, and the row vectors are right singular vectors.

[0016] Further, the specific steps for sorting the singular values from largest to smallest and correspondingly adjusting the orders of the left singular vectors and the right singular vectors in step (4) are as follows: Process the 3 matrices described in step (3) in sequence, and arrange the singular values in each matrix from largest to smallest: σ1≥σ2≥…≥σ r ≥0, where r is the rank of the matrix, and synchronously adjust the orders of the left singular vectors and the right singular vectors.

[0017] Further, the specific steps of sequentially transmitting data to the background in the order of singular values, left singular vectors, and right singular vectors in step (5) are as follows: The first k singular values in step (4) correspond to the main features of the image, where k takes 5%-10% of the smaller value of m and n. First, transmit a floating-point number sequence of length (m + n + 1) composed of the first singular value (1 floating-point number), left singular vector (m floating-point numbers), and right singular vector (n floating-point numbers) of the R matrix; then transmit the floating-point number sequence composed of the first singular value, left singular vector, and right singular vector of the G matrix; then transmit the floating-point number sequence composed of the first singular value, left singular vector, and right singular vector of the B matrix; then transmit the second set of data of the R, G, and B matrices, and so on, until the transmission of the preset kth set of data is completed.

[0018] Further, the specific steps of the background receiving data and reconstructing the image in step (6) are as follows: After the background receives the floating-point number sequence of length 3k(m + n + 1) transmitted in step (5), construct an orthogonal matrix U of (m×k), a diagonal matrix Σ of (k×k), and an orthogonal matrix V of (k×n), and calculate the R, G, and B values of each pixel point according to the formula A = UΣV T to draw the background image, and then sequentially replace the images within the hidden danger rectangular area with the uncompressed images transmitted in step (2).

[0019] Further, in step (6), progressive reconstruction is performed in the order of receiving data, so that the background in the image gradually becomes clear.

[0020] A real-time monitoring image compression device for transmission channels with multiple resolutions includes:

[0021] A hidden danger information acquisition module, which is used to acquire real-time images based on the image acquisition devices installed on site and identify the hidden danger information in the real-time images through edge computing;

[0022] An image information transmission module, which is used to directly transmit the image information in the hidden danger information without compression to the image reconstruction module;

[0023] A singular value decomposition transformation module, which is used to perform singular value decomposition transformation on the original image to obtain singular values, left singular vectors, and right singular vectors;

[0024] A singular value sorting module, which is used to sort the singular values from largest to smallest and correspondingly adjust the order of the left singular vectors and right singular vectors;

[0025] A floating-point number sequence transmission module, which is used to sequentially transmit the corresponding floating-point number sequences to the image reconstruction module in the order of singular values, left singular vectors, and right singular vectors;

[0026] An image reconstruction module, configured to receive the uncompressed image transmitted by the image information transmission module and the floating-point sequence data transmitted by the floating-point sequence transmission module, calculate the RGB color values of each pixel point, and reconstruct the image.

[0027] The advantages of the present invention are as follows:

[0028] 1. By directly controlling the compression ratio through the number (k) of singular values selected for retention, the power transmission operation and maintenance personnel can freely balance the quality and compression ratio. Generally, retaining the first 10% of the singular values can restore most of the background image information, greatly reducing the pressure of wireless image communication transmission. Since a complete image can be drawn without all the information, progressive reconstruction of the background image with gradually increasing clarity can be achieved, facilitating applications in scenarios with high real-time requirements such as wildfire monitoring and construction machinery monitoring.

[0029] 2. The images are drawn with different clarity levels, with high clarity for parts that are of key concern to operation and maintenance personnel such as potential hazards in the power transmission corridor, and low clarity for the background, etc., which better balances the contradiction between image quality and image size, laying a foundation for efficient image data transmission in power transmission visual monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a flowchart of a method for compressing real-time monitoring images of a power transmission corridor with multiple clarity levels according to an embodiment of the present invention;

[0031] Figure 2 It is an attached drawing of an actual case in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0032] Next, the technical solutions in the present invention will be clearly and completely described in conjunction with the drawings in the present invention.

[0033] As Figure 1 shown, an embodiment of the present invention provides a method for compressing real-time monitoring images of a power transmission corridor with multiple clarity levels, including the following steps:

[0034] Step P101: Collect power transmission corridor images. The steps for identifying potential hazards are as follows: Visual monitoring devices installed on power transmission line towers capture images of the power transmission line corridor. For a color image with m×n pixel points, the 24-bit RGB color values of each pixel point are stored, and each color range is 0 - 255, requiring a storage space of 3mn bytes. A model trained based on deep learning is used to intelligently identify various potential hazard information in the image. If there are no potential hazards in the image, wait to process the next image; if there are potential hazards in the image, each potential hazard is marked with a rectangular box, and the coordinates (x1, y1) and (x2, y2) of the diagonal points of the rectangular box are respectively recorded.

[0035] Step P102: The specific steps for transmitting the hidden danger image information are as follows: Since the general hidden danger part accounts for a relatively small proportion in the transmission line channel image and is the key part of concern for transmission operation and maintenance personnel, the hidden danger image information is directly transmitted. The transmitted data includes the diagonal point coordinates (x1, y1), (x2, y2) of the rectangle, and the RGB color values of the pixel points within the area. The storage space is 3(x2 - x1)(y2 - y1) bytes.

[0036] Step P103: The specific steps for the singular value decomposition of the original image are as follows: The original image is decomposed into three matrices of size m×n according to the RGB color values, denoted as the R matrix, G matrix, and B matrix. The singular value decomposition form of each matrix group is:

[0037] A = UΣV T (Equation 1)

[0038] where the column vectors of U are the eigenvectors of A T A; the column vectors of V are the eigenvectors of AA T ; the diagonal elements of Σ are the square roots of the eigenvalues of A T A (or AA T ), that is, the singular values. Considering that generally m < n in the transmission line channel monitoring image, the singular values can be obtained by calculating the eigenvalues of matrix A T A to reduce the calculation amount.

[0039] Step P104: The specific steps for sorting the singular values and singular vectors are as follows: The information of the image is mainly concentrated on the components corresponding to the larger singular values. Therefore, the three matrices in Step P103 are processed in sequence, and the singular values in each matrix are arranged in descending order: σ1 ≥ σ2 ≥ … ≥ σ r ≥ 0 (r is the rank of the matrix), and the orders of the column vectors of U and V are adjusted synchronously.

[0040] Step P105: The specific steps for transmitting the image data are as follows: Transmit the first singular value, left singular vector, and right singular vector of the R matrix, then transmit a floating-point number sequence of length (m + n + 1) composed of the first singular value (1 floating-point number), left singular vector (m floating-point numbers), and right singular vector (m floating-point numbers) of the G matrix, and then transmit a floating-point number sequence of length (m + n + 1) composed of the first singular value, left singular vector, and right singular vector of the B matrix; then transmit the second set of data of the R, G, and B matrices, and so on, until the preset kth set (k ≤ r) of data transmission is completed. The singular values and singular vector elements are all floating-point numbers, and each value is stored in 2 bytes. Then the data compression ratio is:

[0041]

[0042] Step P106: The specific steps for image data reconstruction are as follows: Calculate the R, G, and B values of each pixel point according to Equation 1, where U is an (m×k) orthogonal matrix, Σ is a (k×k) diagonal matrix, and V T is the transpose of a (k×n) orthogonal matrix. Draw the background image, and then sequentially replace the images within the hidden danger rectangular area with the uncompressed images transmitted in Step P102. It is also possible to draw the images of m×n pixel points multiple times in the background according to the received data volume to achieve progressive reconstruction of the background image and gradually make it clearer.

[0043] A case of image compression for monitoring a 500 kV transmission line corridor: The transmission corridor image has 1955×3840 pixel points. The front-end device intelligently identifies that there is a construction machinery hidden danger in the area (801,1977)-(872,2158) of the image, and the area of this area accounts for about 0.17% of the total image area, as Figure 2 shown in (a). Compress images with different sharpness levels and redraw them in the background. Figure 2 (b)-(f) respectively correspond to the image compression effects when k takes values of 10, 50, 100, 200, and 500, and the corresponding compression ratios are shown in Table 1.

[0044] Table 1 Compression ratios corresponding to different k values

[0045]

[0046] As can be seen from Figure 2 , since the hidden danger area is not compressed, even in the (b) figure with a small k value, the construction machinery in the hidden danger area can be clearly displayed, but the background is relatively blurred; the compression ratio of the (c) figure is 12.95, which can already clearly reflect the background information in the image, and the sharpness meets the actual engineering requirements; when k takes the value of 500, the compression ratio is 1.30, and the corresponding (f) figure has very little difference from the uncompressed (a) figure. When the background completes receiving the 10th group of data, a complete image of 1955×3840 pixel points can be drawn, as shown in the (b) figure; when continuing to complete receiving the 50th group of data, the image of 1955×3840 pixel points can be updated, as shown in the (c) figure; and so on to draw the (d)-(f) figures to achieve progressive reconstruction of the background image and gradually make it clearer.

[0047] The embodiment of the present invention also provides a multi-sharpness real-time monitoring image compression device for a transmission corridor, including:

[0048] A hidden danger information acquisition module, which is used to acquire the hidden danger information in the real-time image by edge computing based on the real-time image acquired by the image acquisition device installed on site;

[0049] An image information transmission module, which is used to directly transmit the image information in the hidden danger information without compression to the image reconstruction module;

[0050] The singular value decomposition transformation module is used to perform singular value decomposition transformation on the original image to obtain singular values, left singular vectors, and right singular vectors;

[0051] The singular value sorting module is used to sort the singular values from largest to smallest and correspondingly adjust the orders of the left singular vectors and the right singular vectors;

[0052] The floating-point number sequence transmission module is used to sequentially transmit the corresponding floating-point number sequences to the image reconstruction module in the order of singular values, left singular vectors, and right singular vectors;

[0053] The image reconstruction module is used to receive the uncompressed image transmitted by the image information transmission module and the floating-point number sequence data transmitted by the floating-point number sequence transmission module, calculate the RGB color values of each pixel point, and reconstruct the image.

[0054] The multi-clarity transmission channel image compression method proposed by the present invention has high compression efficiency, while maintaining the integrity of the image information in the hidden danger areas concerned by the operation and maintenance personnel, and can realize progressive image reconstruction in the background. It is suitable for scenarios with weak communication signals and high requirements for image data transmission, and can meet the high-efficiency transmission requirements of real-time monitoring images of field transmission lines.

[0055] As mentioned above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A real-time monitoring image compression method for transmission channels with multiple resolutions, characterized in that It includes the following steps: (1) Based on the real-time image obtained by the image acquisition device installed on-site, edge computing identifies the potential hazard information in the real-time image; (2) Without compressing the image information in the potential hazard information, directly transmit it to the background; (3) Perform singular value decomposition transformation on the original image to obtain singular values, left singular vectors, and right singular vectors; (4) Sort the singular values from largest to smallest, and correspondingly adjust the order of the left singular vectors and right singular vectors; (5) Sequentially transmit the corresponding floating-point number sequences of singular values, left singular vectors, and right singular vectors to the background in order; (6) The background receives the uncompressed image transmitted in step (2) and the floating-point number sequence data transmitted in step (5), calculates the RGB color values of each pixel point, and reconstructs the image.

2. The real-time monitoring image compression method for a transmission channel with multiple resolutions as described in claim 1, characterized in that: The specific steps of the above step (1) for edge computing to identify the potential hazard information in the real-time image based on the real-time image obtained by the image acquisition device installed on-site are as follows: The image obtained by the on-site image acquisition device is a color image of m×n pixel points, and the RGB color values of each pixel point are stored, occupying 24 bits; A model trained based on deep learning is used to intelligently identify various potential hazard information in the image, marked with a rectangular box, and record the diagonal coordinates (x1, y1) and (x2, y2) of the rectangular box.

3. The real-time monitoring image compression method for transmission channels with multiple resolutions according to claim 2, characterized in that: The specific steps of the above step (2) for directly transmitting the image information in the potential hazard information to the background without compression are as follows: For the rectangular box located at (x1, y1) to (x2, y2) in the image described in step (1), transmit the corner coordinates of the rectangular box and the RGB color values of each pixel point within the area, and sequentially complete the transmission of all rectangular box information.

4. The real-time monitoring image compression method for transmission channels with multiple resolutions as claimed in claim 1, wherein: The specific steps of performing singular value decomposition transformation on the original image in step (3) to obtain singular values, left singular vectors, and right singular vectors are as follows: The original image is decomposed into three matrices with dimensions of m×n according to RGB color values, denoted as the R matrix, the G matrix, and the B matrix. The singular value decomposition form of each matrix is A = UΣA T , where Σ is a diagonal matrix of (m×n), and the diagonal elements are non-negative singular values; U is an orthogonal matrix of (m×m), and the column vectors are left singular vectors; V T is the transpose of an orthogonal matrix of (n×n), and the row vectors are right singular vectors.

5. The real-time monitoring image compression method for a transmission channel with multiple resolutions as claimed in claim 4, wherein: In step (4), the specific steps of sorting the singular values from large to small and correspondingly adjusting the orders of the left and right singular vectors are as follows: successively process the three matrices described in step (3), and arrange the singular values in each matrix from large to small: σ1≥σ2≥…≥σ r ≥0, where r is the rank of the matrix, and synchronously adjust the orders of the left and right singular vectors.

6. The real-time monitoring image compression method for transmission channels with multiple resolutions according to claim 5, characterized in that: The specific steps of the above step (5) for sequentially transmitting data to the background in the order of singular values, left singular vectors, and right singular vectors are as follows: The first k singular values in step (4) correspond to the main features of the image, where k takes 5%-10% of the smaller value of m and n. First, transmit a floating-point number sequence of length (m + n + 1) composed of the first singular value (1 floating-point number) of the R matrix, the left singular vector (m floating-point numbers), and the right singular vector (n floating-point numbers); Then transmit the floating-point number sequence composed of the first singular value, left singular vector, and right singular vector of the G matrix; Then transmit the floating-point number sequence composed of the first singular value, left singular vector, and right singular vector of the B matrix; Then transmit the second set of data of the R, G, and B matrices, and so on, until the transmission of the preset kth set of data is completed.

7. The real-time monitoring image compression method for a transmission channel with multiple resolutions according to claim 6, characterized in that: In the step (6), the specific steps for the background to receive data and reconstruct the image are as follows: after the background receives the floating-point number sequence with a length of 3k(m + n + 1) transmitted in step (5), construct an orthogonal matrix U of (m×k), a diagonal matrix Σ of (k×k), and an orthogonal matrix V of (k×n), and calculate according to the formula A = UΣV T Calculate the R, G, and B values of each pixel point, draw the background image, and then sequentially replace the image within the hidden danger rectangular area with the uncompressed image transmitted in step (2).

8. The real-time monitoring image compression method for transmission channels with multiple resolutions as claimed in claim 1, wherein: In step (6), progressive reconstruction is performed in the order of receiving data, and the background in the image is gradually clarified.

9. A real-time monitoring image compression device for a power transmission channel with multiple resolutions, characterized in that, It includes: A potential hazard information acquisition module, which is used to identify the potential hazard information in the real-time image through edge computing based on the real-time image obtained by the image acquisition device installed on-site; An image information transmission module, which is used to directly transmit the image information in the potential hazard information to the image reconstruction module without compression; A singular value decomposition transformation module, which is used to perform singular value decomposition transformation on the original image to obtain singular values, left singular vectors, and right singular vectors; A singular value sorting module, which is used to sort the singular values from largest to smallest and correspondingly adjust the order of the left singular vectors and right singular vectors; A floating-point number sequence transmission module is used to sequentially transmit the corresponding floating-point number sequences to the image reconstruction module in the order of singular values, left singular vectors, and right singular vectors; An image reconstruction module is used to receive the uncompressed image transmitted by the image information transmission module and the floating-point number sequence data transmitted by the floating-point number sequence transmission module, calculate the RGB color values of each pixel point, and reconstruct the image.

Images