Real-time compression and rapid transmission method for inspection images of unmanned aerial vehicle

Abstract

The invention discloses a real-time compression and rapid transmission method for inspection images of an unmanned aerial vehicle. The method comprises the following steps: acquiring and preprocessing the images; image real-time compression: decomposing the preprocessed image by adopting wavelet transform, and compressing the decomposed image by adopting dynamic adjustment and multi-stage compression; data transmission optimization: transmitting the compressed image by adopting a block transmission strategy in combination with forward error correction and redundant data; ground control station data processing: the ground control station receives the compressed image transmitted by the unmanned aerial vehicle, performs defect identification by using a convolutional neural network, and performs defect analysis and early warning in combination with historical data and an expert system; and system performance evaluation and optimization: evaluating and optimizing the image compression efficiency, the transmission speed and the defect identification accuracy of the system by adopting a method of combining simulation and actual test. According to the scheme, efficient and reliable unmanned aerial vehicle image acquisition, compression, transmission and analysis are realized.

Description

Technical Field

[0001] The present invention relates to the field of power image processing, and in particular to a method for real-time compression and rapid transmission of unmanned aerial vehicle inspection images. Background Art

[0002] With the rapid development and widespread application of drone technology, drone inspections have been widely used in the fields of electricity, petroleum, transportation, agriculture, etc. Drone inspections can collect images of target areas in real time by carrying high-definition cameras, thereby realizing remote monitoring and detection of target areas. However, the real-time compression and rapid transmission of drone inspection images face many challenges, mainly including the following aspects: large amount of image data, limited transmission bandwidth, low data transmission reliability, and difficulty in defect identification.

[0003] Existing image compression algorithms have difficulty in achieving a balance between compression efficiency and image quality, especially in the compression process of high-resolution images, where the compression efficiency is low; existing transmission protocols perform poorly in dealing with the instability and signal interference of wireless transmission, and have low transmission reliability; existing defect recognition algorithms have low recognition accuracy in the case of complex backgrounds and changing lighting, and are difficult to meet actual application needs; existing systems have difficulties in the coordination and optimization of various modules, making it difficult to achieve efficient image acquisition, compression, transmission and analysis. Summary of the invention

[0004] The purpose of the invention is to provide a method for real-time compression and rapid transmission of UAV inspection images to solve the above-mentioned problems existing in the prior art.

[0005] Technical solution: A method for real-time compression and rapid transmission of UAV inspection images, comprising the following steps:

[0006] Step S1, image acquisition and preprocessing, including: image acquisition, acquiring images by using a high-definition camera mounted on a drone to obtain an original image; image preprocessing, denoising and enhancing the original image to obtain a preprocessed image;

[0007] Step S2, real-time image compression: decomposing the preprocessed image by wavelet transform to obtain a decomposed image; and compressing the decomposed image by dynamic adjustment and multi-level compression method to obtain a compressed image;

[0008] Step S3, data transmission optimization: adopting a block transmission strategy combined with forward error correction and redundant data methods, selecting the UDP protocol to transmit the compressed image to improve the reliability of data transmission;

[0009] Step S4, ground control station data processing: the ground control station receives the compressed image transmitted by the UAV, uses a convolutional neural network to identify defects, and combines historical data and expert systems to perform defect analysis and early warning;

[0010] Step S5, system performance evaluation and optimization: using a combination of simulation and actual testing methods, the system's image compression efficiency, transmission speed, and defect recognition accuracy are evaluated, and the image compression algorithm, transmission protocol, and defect recognition algorithm are optimized based on the evaluation results.

[0011] According to another aspect of the present application, the step S1 is further:

[0012] Step S11, image acquisition: The image acquisition frequency is set to 30fps, and the original image is obtained in the format of:

[0013] D = W × H × F × B;

[0014] Among them, 4K width W = 3840, 4K height H = 2160, frame rate F = 30, RGB format bytes per pixel B = 3;

[0015] Step S12, image denoising: Use a Gaussian filter to denoise the original image, smooth the image by weighted averaging the pixel values ​​in the neighborhood, and obtain a denoised image. The specific method is as follows:

[0016] I^(x,y)=∑ (i,j)∈Ω I(i,j)·G(i,j);

[0017] Where, Ω is the neighborhood of pixel (x, y), I(i, j) is the pixel value of the original image at (i, j), I^(x, y) is the pixel value of the denoised image at (x, y), and G(i, j) is the weight of the Gaussian kernel at (i, j);

[0018] Step S13, image enhancement: the denoised image is enhanced by using histogram equalization technology to obtain an enhanced image. The specific method is as follows:

[0019] I^(x,y)=round((I(x,y)-Imin / (Imax-Imin)·(L-1));

[0020] Where Imin and Imax are the minimum and maximum grayscale values ​​of the image, respectively, and L is the number of gray levels.

[0021] According to another aspect of the present application, step S2 is further:

[0022] Step S21, image decomposition: using wavelet transform to decompose the preprocessed image, including:

[0023] Step S211, wavelet transform: decompose the image into sub-bands of different frequencies, thereby achieving image compression. The specific method is as follows:

[0024] W(a,b)=∫ -∞ ∞ ∫ -∞ ∞ I(x,y)·ψ a,b (x,y)d x d y ;

[0025] Among them, W(a,b) is the wavelet coefficient, I(x,y) is the preprocessed image, ψ a,b (x,y) is the wavelet basis function, a and b are the scale and translation parameters respectively;

[0026] Step S212, decomposition level: According to the complexity of the image and the compression requirements, select 2 to 4 levels of decomposition, the method is as follows:

[0027] I = LL + LH + HL + HH;

[0028] Among them, LL is the low-frequency subband, i.e., the approximate image, and LH, HL, and HH are the high-frequency subbands, i.e., the detail images;

[0029] Step S22, dynamic adjustment: dynamically adjust the decomposed image to further reduce the amount of data, including:

[0030] Step S221, threshold processing: by setting the threshold, remove the noise and redundant information in the wavelet coefficients, the method is as follows:

[0031] W^(a,b)=W(a,b), if∣W(a,b)∣>T, otherwise W^(a,b)=0;

[0032] Where T is the threshold, W(a,b) is the wavelet coefficient, and W^(a,b) is the processed wavelet coefficient;

[0033] Step S222, adaptive threshold: dynamically adjust the threshold according to the local characteristics of the image to improve the compression effect, the method is as follows:

[0034] T(x,y)=median(∣W(a,b)∣)·σ;

[0035] Where σ is the adjustment factor, which is usually between 2.5 and 3.5;

[0036] Step S23, multi-level compression: further compressing the dynamically adjusted image using a multi-level compression method, including:

[0037] Step S231, quantization: quantize the wavelet coefficients to reduce the accuracy of the coefficients. The method is as follows:

[0038] W^(a,b)=round(W(a,b) / Q)·Q;

[0039] Where Q is the quantization step size;

[0040] Step S232, entropy coding: Encode the quantized coefficients using an entropy coding method to further reduce the amount of data. The method is as follows:

[0041] C = encode(W^(a,b));

[0042] Among them, C is the encoded data;

[0043] Step S233, block compression: Divide the image into multiple small blocks and compress them separately to improve compression efficiency and parallel processing capability. The method is as follows:

[0044] I=∪ i＝1 N Bi;

[0045] Where Bi is the i-th image block, N is the number of image blocks;

[0046] Step S24, compressed image reconstruction: reconstructing the compressed image to obtain a compressed image, including:

[0047] Step S241, inverse wavelet transform: perform inverse wavelet transform on the compressed wavelet coefficients to reconstruct the image, the method is:

[0048] I^(x,y)=∑ a,b W^(a,b)·ψ a,b (x,y);

[0049] Where I^(x,y) is the reconstructed image;

[0050] Step S242, image post-processing: perform post-processing on the reconstructed image, such as removing blocking effects, enhancing contrast, etc., to improve image quality, the method is as follows:

[0051] I^(x,y)=postprocess(I^(x,y)).

[0052] According to another aspect of the present application, step S3 is further:

[0053] Step S31, block transmission strategy: divide the compressed image into multiple small blocks and transmit them separately, including:

[0054] Step S311, block strategy: divide the image into multiple small blocks according to the image size and network bandwidth, the method is as follows:

[0055] I=∪ i＝ 1 N Bi;

[0056] Where I is the compressed image, Bi is the i-th image block, and N is the number of image blocks;

[0057] Step S312, parallel transmission: using multithreading or parallel processing technology to transmit multiple image blocks at the same time to increase the transmission speed, the method is:

[0058] Speed=∑ i＝1 N Speed i； ;

[0059] Among them, Speed i is the transmission speed of the i-th image block;

[0060] Step S32, forward error correction: adding forward error correction code during the transmission process to improve the reliability of data transmission, including:

[0061] Step S321, FEC encoding: adding redundant data to each image block so that the receiving end can detect and correct errors during transmission, the method is as follows:

[0062] Bi′=Bi+Ei;

[0063] Where Bi′ is the image block after adding FEC coding, and Ei is the redundant data generated by FEC coding;

[0064] Step S322, FEC decoding: FEC decoding is performed on the transmitted image block at the receiving end to detect and correct errors, the method is as follows:

[0065] B^i=decode(Bi′);

[0066] Where, B^i is the decoded image block;

[0067] Step S33, redundant data: adding redundant data during the transmission process to improve the reliability of data transmission, including:

[0068] Step S331, redundant data generation: a certain proportion of redundant data is added to each image block so that the receiving end can detect and correct errors in the transmission process, the method is as follows:

[0069] Bi′=Bi+Ri;

[0070] Among them, Ri is the generated redundant data;

[0071] Step S332, redundant data detection: redundant data detection is performed on the transmitted image block at the receiving end to detect and correct errors, the method is as follows:

[0072] B^i = detect(Bi′);

[0073] Where B^i is the detected image block;

[0074] Step S34, selecting UDP protocol: selecting UDP protocol to transmit the compressed image to improve transmission speed and real-time performance;

[0075] Step S35, transmission scheduling: using a transmission scheduling algorithm to optimize the transmission order and priority of the image blocks, including:

[0076] Priority scheduling: Dynamically adjust the transmission order and priority according to the importance and urgency of the image blocks;

[0077] Dynamic scheduling: Dynamically adjust the transmission strategy according to network conditions and transmission progress to improve transmission efficiency.

[0078] According to another aspect of the present application, step S4 is further:

[0079] Step S41, image reception: the ground control station receives the compressed image transmitted by the UAV;

[0080] Step S42, image decompression: decompressing the received compressed image to restore the original image;

[0081] Step S43, defect recognition: using a convolutional neural network to perform defect recognition on the decompressed image, and using a target detection algorithm to locate the defect area in the image;

[0082] Step S44, defect analysis: combining historical data and expert system, analyzing the identified defects, comparing the currently identified defects with historical data, analyzing the defect change trend, using expert system to analyze the defects, and providing diagnosis suggestions and treatment solutions;

[0083] Step S45, defect warning: Based on the defect analysis results, a defect warning is performed, and according to the severity and change trend of the defect, a corresponding warning signal is triggered, and the warning information is notified to relevant personnel.

[0084] Beneficial effects: This solution achieves efficient and reliable real-time compression and rapid transmission of UAV inspection images through steps such as image acquisition and preprocessing, real-time image compression, data transmission optimization, ground control station data processing, and system performance evaluation and optimization. The beneficial effects of the solution include improving image acquisition quality, image compression efficiency, data transmission reliability, defect recognition accuracy, real-time defect warning, and continuous optimization of system performance. Through these measures, this solution can significantly improve the management efficiency and safety of inspection targets, and provide strong technical support for UAV inspection applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 It is an overall flow chart of the scheme of the present invention.

[0086] Figure 2 It is a flow chart of step S2 of the present invention.

[0087] Figure 3 It is a flow chart of step S3 of the present invention.

[0088] Figure 4 It is a flow chart of step S4 of the present invention. DETAILED DESCRIPTION

[0089] like Figure 1 As shown, according to one aspect of the present application, a method for real-time compression and rapid transmission of drone inspection images includes the following steps:

[0090] Step S1, image acquisition and preprocessing, including: image acquisition, acquiring images by using a high-definition camera mounted on a drone to obtain an original image; image preprocessing, denoising and enhancing the original image to obtain a preprocessed image;

[0091] Step S2, real-time image compression: decomposing the preprocessed image by wavelet transform to obtain a decomposed image; and compressing the decomposed image by dynamic adjustment and multi-level compression method to obtain a compressed image;

[0092] Step S3, data transmission optimization: adopting a block transmission strategy combined with forward error correction and redundant data methods, selecting the UDP protocol to transmit the compressed image to improve the reliability of data transmission;

[0093] Step S4, ground control station data processing: the ground control station receives the compressed image transmitted by the UAV, uses a convolutional neural network to identify defects, and combines historical data and expert systems to perform defect analysis and early warning;

[0094] Step S5, system performance evaluation and optimization: using a combination of simulation and actual testing methods, the system's image compression efficiency, transmission speed, and defect recognition accuracy are evaluated, and the image compression algorithm, transmission protocol, and defect recognition algorithm are optimized based on the evaluation results.

[0095] In this embodiment, this scheme proposes a method for real-time compression and rapid transmission of UAV inspection images, which realizes efficient and reliable image acquisition, compression, transmission and analysis through steps such as image acquisition and preprocessing, real-time image compression, data transmission optimization, ground control station data processing, and system performance evaluation and optimization. The image is acquired by a high-definition camera carried by the UAV to ensure the high resolution and detail richness of the image, and provide high-quality original data for subsequent processing; the original image is denoised and enhanced to effectively remove the noise in the image, enhance the details and contrast of the image, and improve the overall quality of the image. The preprocessed image is decomposed by wavelet transform, which can decompose the image into sub-bands of different frequencies, thereby realizing efficient image compression; the decomposed image is dynamically adjusted, and the noise and redundant information in the wavelet coefficients are removed by setting a threshold, further reducing the amount of data; the dynamically adjusted image is further compressed by a multi-level compression method, and the precision of the coefficients is reduced by quantization and entropy coding, which significantly improves the compression efficiency. Divide the compressed image into multiple small blocks and transmit them separately, using multi-threading or parallel processing technology to improve the transmission speed; add forward error correction code during the transmission process so that the receiving end can detect and correct errors during the transmission process and improve the reliability of data transmission; add redundant data during the transmission process to further improve the reliability of data transmission and ensure the integrity of the data; select the UDP protocol to transmit the compressed image, which has low latency and high transmission speed and is suitable for real-time transmission scenarios. Use the pre-trained CNN model for image classification and target detection, which can accurately identify defects in the image; compare the currently identified defects with historical data, analyze the changing trend of defects, and provide a more comprehensive defect analysis; use the expert system to analyze defects, provide diagnostic suggestions and treatment solutions, and improve the accuracy and practicality of defect identification. According to the severity and changing trend of the defects, trigger the corresponding warning signal and notify the relevant personnel in time to handle them; notify the relevant personnel of the warning information through SMS, email, APP push, etc. to ensure that the defects can be handled in time. Evaluate the performance of the system through simulation testing, simulate actual application scenarios, record various performance indicators, and provide data support for system optimization; conduct tests in actual environments to verify the accuracy of simulation test results and further evaluate the performance of the system; analyze the performance indicators obtained from simulation tests and actual tests to find out the performance bottlenecks of the system and provide a basis for system optimization; optimize the image compression algorithm, transmission protocol, and defect recognition algorithm based on the performance analysis results to continuously improve the overall performance of the system. The system can complete image acquisition, compression, transmission, and analysis in a short time to meet real-time requirements; by optimizing the image compression algorithm, transmission protocol, and defect recognition algorithm, the system can achieve efficient image processing and transmission while ensuring image quality.

[0096] According to another aspect of the present application, step S1 is further:

[0097] Step S11, image acquisition: The image acquisition frequency is set to 30fps, and the original image is obtained in the format of:

[0098] D = W × H × F × B;

[0099] Among them, 4K width W = 3840, 4K height H = 2160, frame rate F = 30, RGB format bytes per pixel B = 3;

[0100] Step S12, image denoising: Use a Gaussian filter to denoise the original image, smooth the image by weighted averaging the pixel values ​​in the neighborhood, and obtain a denoised image. The specific method is as follows:

[0101] I^(x,y)=∑ (i,j)∈Ω I(i,j)·G(i,j);

[0102] Where, Ω is the neighborhood of pixel (x, y), I(i, j) is the pixel value of the original image at (i, j), I^(x, y) is the pixel value of the denoised image at (x, y), and G(i, j) is the weight of the Gaussian kernel at (i, j);

[0103] Step S13, image enhancement: the denoised image is enhanced by using histogram equalization technology to obtain an enhanced image. The specific method is as follows:

[0104] I^(x,y)=round((I(x,y)-Imin / (Imax-Imin)·(L-1));

[0105] Where Imin and Imax are the minimum and maximum grayscale values ​​of the image, respectively, and L is the number of gray levels.

[0106] According to another aspect of the present application, step S1 can be further optimized to include the following steps:

[0107] Step S14, image correction: correcting the enhanced image using geometric correction and color correction, including:

[0108] Step S141, geometric correction: correcting geometric distortions in the image, such as tilt, distortion, etc., through affine transformation or perspective transformation;

[0109] Step S142, color correction: correcting the color deviation in the image through a color correction matrix or white balance adjustment to ensure the consistency of the image color;

[0110] Step S15, image segmentation: using image segmentation technology to segment the corrected image, including:

[0111] Step S151, threshold segmentation: segment the image into foreground and background by setting a threshold;

[0112] Step S152, edge detection: extract edge information in the image by using methods such as Canny edge detection;

[0113] Step S153, region growing: dividing the image into different regions by growing seed points;

[0114] Step S16: Image normalization: normalizing the segmented image using image normalization technology, including:

[0115] Step S161, grayscale normalization: normalize the grayscale value of the image to a specific range, such as [0, 1] or [0, 255];

[0116] Step S162, size normalization: normalize the size of the image to a specific size to facilitate subsequent image processing and analysis.

[0117] Image preprocessing is an important part of the real-time compression and rapid transmission of UAV inspection images. Through denoising, enhancement, correction, segmentation and normalization, the image quality and subsequent processing effect can be significantly improved. The preprocessed image can not only better support image compression and transmission, but also improve the accuracy of defect recognition and analysis.

[0118] like Figure 2 As shown, according to another aspect of the present application, the step S2 is further:

[0119] Step S21, image decomposition: using wavelet transform to decompose the preprocessed image, including:

[0120] Step S211, wavelet transform: decompose the image into sub-bands of different frequencies, thereby achieving image compression. The specific method is as follows:

[0121] W(a,b)=∫ -∞ ∞ ∫ -∞ ∞ I(x,y)·ψ a,b (x,y)d x d y ;

[0122] Among them, W(a,b) is the wavelet coefficient, I(x,y) is the preprocessed image, ψ a,b (x,y) is the wavelet basis function, a and b are the scale and translation parameters respectively;

[0123] Step S212, decomposition level: According to the complexity of the image and the compression requirements, select 2 to 4 levels of decomposition, the method is as follows:

[0124] I = LL + LH + HL + HH;

[0125] Among them, LL is the low-frequency subband, i.e., the approximate image, and LH, HL, and HH are the high-frequency subbands, i.e., the detail images;

[0126] Step S22, dynamic adjustment: dynamically adjust the decomposed image to further reduce the amount of data, including:

[0127] Step S221, threshold processing: by setting the threshold, remove the noise and redundant information in the wavelet coefficients, the method is as follows:

[0128] W^(a,b)=W(a,b), if∣W(a,b)∣>T, otherwise W^(a,b)=0;

[0129] Where T is the threshold, W(a,b) is the wavelet coefficient, and W^(a,b) is the processed wavelet coefficient;

[0130] Step S222, adaptive threshold: dynamically adjust the threshold according to the local characteristics of the image to improve the compression effect, the method is as follows:

[0131] T(x,y)=median(∣W(a,b)∣)·σ;

[0132] Where σ is the adjustment factor, which is usually between 2.5 and 3.5;

[0133] Step S23, multi-level compression: further compressing the dynamically adjusted image using a multi-level compression method, including:

[0134] Step S231, quantization: quantize the wavelet coefficients to reduce the accuracy of the coefficients. The method is as follows:

[0135] W^(a,b)=round(W(a,b) / Q)·Q;

[0136] Where Q is the quantization step size;

[0137] Step S232, entropy coding: Encode the quantized coefficients using an entropy coding method to further reduce the amount of data. The method is as follows:

[0138] C = encode(W^(a,b));

[0139] Among them, C is the encoded data;

[0140] Step S233, block compression: Divide the image into multiple small blocks and compress them separately to improve compression efficiency and parallel processing capability. The method is as follows:

[0141] I=∪ i＝1 N Bi;

[0142] Where Bi is the i-th image block, N is the number of image blocks;

[0143] Step S24, compressed image reconstruction: reconstructing the compressed image to obtain a compressed image, including:

[0144] Step S241, inverse wavelet transform: perform inverse wavelet transform on the compressed wavelet coefficients to reconstruct the image, the method is:

[0145] I^(x,y)=∑ a,b W^(a,b)·ψ a,b (x,y);

[0146] Where I^(x,y) is the reconstructed image;

[0147] Step S242, image post-processing: perform post-processing on the reconstructed image, such as removing blocking effects, enhancing contrast, etc., to improve image quality, the method is as follows:

[0148] I^(x,y)=postprocess(I^(x,y)).

[0149] Real-time image compression can significantly reduce the amount of image data through wavelet transform, dynamic adjustment and multi-level compression, thereby improving transmission efficiency and reducing storage costs. The compressed image can not only better support real-time transmission, but also improve the efficiency of subsequent image analysis and recognition.

[0150] like Figure 3 As shown, according to another aspect of the present application, the step S3 is further:

[0151] Step S31, block transmission strategy: divide the compressed image into multiple small blocks and transmit them separately, including:

[0152] Step S311, block strategy: divide the image into multiple small blocks according to the image size and network bandwidth, the method is as follows:

[0153] I=∪ i＝ 1 N Bi;

[0154] Where I is the compressed image, Bi is the i-th image block, and N is the number of image blocks;

[0155] Step S312, parallel transmission: using multithreading or parallel processing technology to transmit multiple image blocks at the same time to increase the transmission speed, the method is:

[0156] Speed=∑ i＝1 N Speed i； ;

[0157] Among them, Speed i is the transmission speed of the i-th image block;

[0158] Step S32, forward error correction: adding forward error correction code during the transmission process to improve the reliability of data transmission, including:

[0159] Step S321, FEC encoding: adding redundant data to each image block so that the receiving end can detect and correct errors during transmission, the method is as follows:

[0160] Bi′=Bi+Ei;

[0161] Where Bi′ is the image block after adding FEC coding, and Ei is the redundant data generated by FEC coding;

[0162] Step S322, FEC decoding: FEC decoding is performed on the transmitted image block at the receiving end to detect and correct errors, the method is as follows:

[0163] B^i=decode(Bi′);

[0164] Where, B^i is the decoded image block;

[0165] Step S33, redundant data: adding redundant data during the transmission process to improve the reliability of data transmission, including:

[0166] Step S331, redundant data generation: a certain proportion of redundant data is added to each image block so that the receiving end can detect and correct errors in the transmission process, the method is as follows:

[0167] Bi′=Bi+Ri;

[0168] Among them, Ri is the generated redundant data;

[0169] Step S332, redundant data detection: redundant data detection is performed on the transmitted image block at the receiving end to detect and correct errors, the method is as follows:

[0170] B ^ i = detect(Bi′);

[0171] Among them, B ^ i is the detected image block;

[0172] Step S34, selecting UDP protocol: selecting UDP protocol to transmit the compressed image to improve transmission speed and real-time performance;

[0173] Step S35, transmission scheduling: using a transmission scheduling algorithm to optimize the transmission order and priority of the image blocks, including:

[0174] Priority scheduling: Dynamically adjust the transmission order and priority according to the importance and urgency of the image blocks;

[0175] Dynamic scheduling: Dynamically adjust the transmission strategy according to network conditions and transmission progress to improve transmission efficiency.

[0176] According to another aspect of the present application, the priority scheduling and dynamic scheduling in step S35 are specifically:

[0177] Priority scheduling: Dynamically adjust the transmission order and priority according to the importance and urgency of the image block:

[0178] Priority(Bi)=importance(Bi)·urgency(Bi);

[0179] Among them, Priority(Bi) is the priority of the i-th image block, importance(Bi) is the importance of the image block, and urgency(Bi) is the urgency of the image block;

[0180] Dynamic scheduling: Dynamically adjust the transmission strategy according to network conditions and transmission progress to improve transmission efficiency:

[0181] Schedule(Bi)=optimal(Network Condition,Progress);

[0182] Among them, Schedule (Bi) is the transmission scheduling strategy of the i-th image block, Network Condition is the network condition, and Progress is the transmission progress.

[0183] Data transmission optimization can significantly improve the reliability and efficiency of data transmission through block transmission strategy, forward error correction, redundant data, selection of UDP protocol and transmission scheduling. The optimized transmission strategy can not only better support real-time transmission, but also improve the efficiency of subsequent image analysis and recognition.

[0184] like Figure 4 As shown, according to another aspect of the present application, the step S4 is further:

[0185] Step S41, image reception: the ground control station receives the compressed image transmitted by the UAV;

[0186] Step S42, image decompression: decompressing the received compressed image to restore the original image;

[0187] Step S43, defect recognition: using a convolutional neural network to perform defect recognition on the decompressed image, and using a target detection algorithm to locate the defect area in the image;

[0188] Step S44, defect analysis: combining historical data and expert system, analyzing the identified defects, comparing the currently identified defects with historical data, analyzing the defect change trend, using expert system to analyze the defects, and providing diagnosis suggestions and treatment solutions;

[0189] Step S45, defect warning: Based on the defect analysis results, a defect warning is performed, and according to the severity and change trend of the defect, a corresponding warning signal is triggered, and the warning information is notified to relevant personnel via SMS, email, or APP push.

[0190] The ground control station data processing can realize real-time processing and monitoring of UAV inspection images through steps such as image reception, decompression, defect identification, defect analysis and defect warning. By combining historical data and expert systems, the ground control station can not only accurately identify and analyze defects, but also issue warnings in a timely manner, thereby improving the management efficiency and safety of inspection targets.

[0191] According to another aspect of the present application, step S5 system performance evaluation and optimization is the last step in the real-time compression and rapid transmission method of drone inspection images, and is also a key link to ensure the efficient operation of the system. By evaluating the system's image compression efficiency, transmission speed, defect recognition accuracy, etc., and optimizing the image compression algorithm, transmission protocol, and defect recognition algorithm based on the evaluation results, the overall performance of the system can be further improved. The following are the specific steps and methods for system performance evaluation and optimization:

[0192] Step S51, performance evaluation indicators: determine key indicators for system performance evaluation, including:

[0193] Step S511, image compression efficiency: evaluate the ratio of the amount of data after image compression to the amount of original data, usually expressed as a compression ratio (CR):

[0194] CR=Original Size / Compressed Size;

[0195] Among them, Original Size is the data size of the original image, and Compressed Size is the data size of the compressed image;

[0196] Step S512, transmission speed: evaluate the speed at which image data is transmitted from the UAV to the ground control station, usually expressed as a transmission rate (TR):

[0197] TR=Data Size / Transmission Time;

[0198] Among them, Data Size is the amount of data transmitted, and Transmission Time is the transmission time;

[0199] Step S513, defect recognition accuracy: evaluate the accuracy of the defect recognition algorithm, usually expressed by accuracy, recall and F1-Score:

[0200] Accuracy=(TruePositives+TrueNegatives) / TotalSamples;

[0201] Recall=True Positives / (True Positives+False Negatives);

[0202] F1-Score=2·Precision·Recall / (Precision+Recall);

[0203] Among them, True Positives is the number of defects correctly identified, True Negatives is the number of non-defects correctly identified, False Negatives is the number of defects not identified, and False Positives is the number of defects incorrectly identified;

[0204] Step S52: Simulation test: Evaluate the performance of the system through simulation test, including:

[0205] Step S521, simulation environment construction: construct a simulation environment similar to the actual application scenario, including a drone, a ground control station, a network environment, etc.:

[0206] Simulated Environment=Setup(Drone,Ground Station,Network);

[0207] Among them, Simulated Environment is the simulation environment, Drone is the drone, Ground Station is the ground control station, and Network is the network environment;

[0208] Step S522, step simulation test execution: perform image acquisition, compression, transmission, decompression, defect identification and other steps in the simulation environment, and record various performance indicators:

[0209] Performance Metrics=Execute(Simulated Environment);

[0210] Among them, Performance Metrics is the performance index obtained by simulation test;

[0211] Step S53, actual test: evaluating the performance of the system through actual test, including:

[0212] Step S531, actual environment deployment: deploy the drone and ground control station in the actual application scenario to perform operations such as image acquisition, compression, transmission, decompression, and defect identification:

[0213] Actual Environment=Deploy(Drone,Ground Station);

[0214] Among them, Actual Environment is the actual environment;

[0215] Step S532, actual test execution: perform image acquisition, compression, transmission, decompression, defect identification and other steps in the actual environment, and record various performance indicators:

[0216] Performance Metrics=Execute(Actual Environment);

[0217] Among them, Performance Metrics is the performance indicator obtained from actual testing;

[0218] Step S54, performance analysis: analyzing the performance indicators obtained from the simulation test and the actual test, including:

[0219] Step S541, data comparison: compare the performance indicators obtained by the simulation test and the actual test, and analyze the reasons for the differences:

[0220] Difference=Compare(Simulated Performance, Actual Performance);

[0221] Among them, Difference is the performance difference between the simulated test and the actual test, Simulated Performance is the performance index of the simulated test, and Actual Performance is the performance index of the actual test;

[0222] Step S542, performance bottleneck analysis: Analyze the performance bottleneck of the system according to the performance indicators, such as low image compression efficiency, slow transmission speed, low defect recognition accuracy, etc.:

[0223] Bottleneck=Analyze(Performance Metrics);

[0224] Among them, Bottleneck is the performance bottleneck of the system;

[0225] Step S55, system optimization: according to the performance analysis results, the system is optimized, including:

[0226] Step S551, image compression algorithm optimization: adjust the parameters of wavelet transform, optimize the dynamic adjustment strategy, and improve the multi-level compression method to improve the image compression efficiency:

[0227] Optimized Compression=Optimize(Wavelet Transform,Dynamic Adjustment,Multi-level Compression);

[0228] Among them, Optimized Compression is the optimized image compression algorithm;

[0229] Step S552, transmission protocol optimization: adjust the parameters of the UDP protocol, optimize the block transmission strategy, improve the forward error correction and redundant data strategy to improve the transmission speed and reliability:

[0230] Optimized Transmission=Optimize(UDP Parameters,Block Transmission,FEC,Redundancy);

[0231] Among them, Optimized Transmission is the optimized transmission protocol;

[0232] Step S553, defect recognition algorithm optimization: adjust the structure of the CNN model, optimize the target detection algorithm, improve historical data and expert system analysis methods to improve the accuracy of defect recognition:

[0233] Optimized Detection=Optimize(CNN Model,Object Detection,HistoricalData,Expert Sy stem);

[0234] Among them, Optimized Detection is the optimized defect recognition algorithm.

[0235] System performance evaluation and optimization can comprehensively evaluate and improve the overall performance of the system through simulation testing, actual testing, performance analysis and system optimization. By continuously optimizing image compression algorithms, transmission protocols and defect recognition algorithms, the system can better support real-time image acquisition, compression, transmission and analysis, thereby improving the management efficiency and safety of inspection targets.

[0236] It should be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations.

Claims

1. A method for real-time compression and rapid transmission of drone inspection images, characterized in that: The steps include: Step S1, image acquisition and preprocessing, including: image acquisition, acquiring images by using a high-definition camera mounted on a drone to obtain an original image; image preprocessing, denoising and enhancing the original image to obtain a preprocessed image; Step S2, real-time image compression: decomposing the preprocessed image by wavelet transform to obtain a decomposed image; and compressing the decomposed image by dynamic adjustment and multi-level compression method to obtain a compressed image; Step S3, data transmission optimization: adopting a block transmission strategy combined with forward error correction and redundant data methods, selecting the UDP protocol to transmit the compressed image to improve the reliability of data transmission; Step S4, ground control station data processing: the ground control station receives the compressed image transmitted by the UAV, uses a convolutional neural network to identify defects, and combines historical data and expert systems to perform defect analysis and early warning; Step S5, system performance evaluation and optimization: using a combination of simulation and actual testing methods, the system's image compression efficiency, transmission speed, and defect recognition accuracy are evaluated, and the image compression algorithm, transmission protocol, and defect recognition algorithm are optimized based on the evaluation results.

2. The method for real-time compression and rapid transmission of unmanned aerial vehicle inspection images as claimed in claim 1, characterized in that: The step S1 is further as follows: Step S11, image acquisition: The image acquisition frequency is set to 30fps, and the original image is obtained in the format of: D = W × H × F × B; Among them, 4K width W = 3840, 4K height H = 2160, frame rate F = 30, RGB format bytes per pixel B = 3; Step S12, image denoising: Use a Gaussian filter to denoise the original image, smooth the image by weighted averaging the pixel values ​​in the neighborhood, and obtain a denoised image. The specific method is as follows: I ^ (x,y)=∑ (i,j)∈Ω I(i,j)·G(i,j); Where, Ω is the neighborhood of pixel (x, y), I(i, j) is the pixel value at (i, j) in the original image, and I ^ (x,y) is the pixel value of the denoised image at (x,y), and G(i,j) is the weight of the Gaussian kernel at (i,j); Step S13, image enhancement: the denoised image is enhanced by using histogram equalization technology to obtain an enhanced image. The specific method is as follows: I ^ (x,y)=round((I(x,y)-Imin / (Imax-Imin)·(L-1)); Where Imin and Imax are the minimum and maximum grayscale values ​​of the image, respectively, and L is the number of gray levels.

3. The method for real-time compression and rapid transmission of unmanned aerial vehicle inspection images as claimed in claim 2, characterized in that: The real-time image compression in step S2 is further as follows: Step S21, image decomposition: using wavelet transform to decompose the preprocessed image, including: Step S211, wavelet transform: decompose the image into sub-bands of different frequencies, thereby achieving image compression. The specific method is as follows: W(a,b)=∫ -∞ ∞ ∫ -∞ ∞ I(x,y)·ψ a,b (x,y)d x d y ; Among them, W(a,b) is the wavelet coefficient, I(x,y) is the preprocessed image, ψ a,b (x,y) is the wavelet basis function, a and b are the scale and translation parameters respectively; Step S212, decomposition level: According to the complexity of the image and the compression requirements, select 2 to 4 levels of decomposition, the method is as follows: I = LL + LH + HL + HH; Among them, LL is the low-frequency subband, i.e., the approximate image, and LH, HL, and HH are the high-frequency subbands, i.e., the detail images; Step S22, dynamic adjustment: dynamically adjust the decomposed image to further reduce the amount of data, including: Step S221, threshold processing: by setting the threshold, remove the noise and redundant information in the wavelet coefficients, the method is as follows: W^(a,b)=W(a,b), if∣W(a,b)∣>T, otherwise W^(a,b)=0; Where T is the threshold, W(a,b) is the wavelet coefficient, and W^(a,b) is the processed wavelet coefficient; Step S222, adaptive threshold: dynamically adjust the threshold according to the local characteristics of the image to improve the compression effect, the method is as follows: T(x,y)=median(∣W(a,b)∣)·σ; Where σ is the adjustment factor, which is usually between 2.5 and 3.5; Step S23, multi-level compression: further compressing the dynamically adjusted image using a multi-level compression method, including: Step S231, quantization: quantize the wavelet coefficients to reduce the accuracy of the coefficients. The method is as follows: W^(a,b)=round(W(a,b) / Q)·Q; Where Q is the quantization step size; Step S232, entropy coding: Encode the quantized coefficients using an entropy coding method to further reduce the amount of data. The method is as follows: C = encode(W^(a,b)); Among them, C is the encoded data; Step S233, block compression: Divide the image into multiple small blocks and compress them separately to improve compression efficiency and parallel processing capability. The method is as follows: I=∪ i＝1 N Ball; Where Bi is the i-th image block, N is the number of image blocks; Step S24, compressed image reconstruction: reconstructing the compressed image to obtain a compressed image, including: Step S241, inverse wavelet transform: perform inverse wavelet transform on the compressed wavelet coefficients to reconstruct the image, the method is: I^(x,y)=∑ a,b W^(a,b)·ψ a,b (x,y); Where I^(x,y) is the reconstructed image; Step S242, image post-processing: post-process the reconstructed image, such as removing blocking effects, enhancing contrast, etc., to improve image quality, the method is as follows: I^(x,y)=postprocess(I^(x,y)).

4. The method for real-time compression and rapid transmission of unmanned aerial vehicle inspection images as claimed in claim 3, characterized in that: The step S3 is further as follows: Step S31, block transmission strategy: divide the compressed image into multiple small blocks and transmit them separately, including: Step S311, block strategy: divide the image into multiple small blocks according to the image size and network bandwidth, the method is as follows: I=∪ i =1 N Ball; Where I is the compressed image, Bi is the i-th image block, and N is the number of image blocks; Step S312, parallel transmission: using multithreading or parallel processing technology to transmit multiple image blocks at the same time to increase the transmission speed, the method is: Speed=∑ i＝1 N Speed i； ; Among them, Speed i is the transmission speed of the i-th image block; Step S32, forward error correction: adding forward error correction code during the transmission process to improve the reliability of data transmission, including: Step S321, FEC encoding: adding redundant data to each image block so that the receiving end can detect and correct errors during transmission, the method is as follows: Bi ′ =Bi+Ei; Among them, Bi ′ is the image block after adding FEC coding, and Ei is the redundant data generated by FEC coding; Step S322, FEC decoding: FEC decoding is performed on the transmitted image block at the receiving end to detect and correct errors, the method is as follows: B^i=decode(Bi′); Where, B^i is the decoded image block; Step S33, redundant data: adding redundant data during the transmission process to improve the reliability of data transmission, including: Step S331, redundant data generation: a certain proportion of redundant data is added to each image block so that the receiving end can detect and correct errors in the transmission process, the method is as follows: Bi ′ =Bi+Ri; Among them, Ri is the generated redundant data; Step S332, redundant data detection: redundant data detection is performed on the transmitted image block at the receiving end to detect and correct errors, the method is as follows: B ^ i=detect(Bi ′ ); Where B^i is the detected image block; Step S34, selecting UDP protocol: selecting UDP protocol to transmit the compressed image to improve transmission speed and real-time performance; Step S35, transmission scheduling: using a transmission scheduling algorithm to optimize the transmission order and priority of the image blocks, including: Priority scheduling: Dynamically adjust the transmission order and priority according to the importance and urgency of the image blocks; Dynamic scheduling: Dynamically adjust the transmission strategy according to network conditions and transmission progress to improve transmission efficiency.

5. The method for real-time compression and rapid transmission of unmanned aerial vehicle inspection images as claimed in claim 4, characterized in that: The step S4 is further as follows: Step S41, image reception: the ground control station receives the compressed image transmitted by the UAV; Step S42, image decompression: decompressing the received compressed image to restore the original image; Step S43, defect recognition: using a convolutional neural network to perform defect recognition on the decompressed image, and using a target detection algorithm to locate the defect area in the image; Step S44, defect analysis: combining historical data and expert system, analyzing the identified defects, comparing the currently identified defects with historical data, analyzing the defect change trend, using expert system to analyze the defects, and providing diagnosis suggestions and treatment solutions; Step S45, defect warning: Based on the defect analysis results, a defect warning is performed, and according to the severity and change trend of the defect, a corresponding warning signal is triggered, and the warning information is notified to relevant personnel.

Images