CN-122027797-A - Unmanned aerial vehicle inspection image transmission method and system based on asymmetric coding and decoding image compression technology

CN122027797ACN 122027797 ACN122027797 ACN 122027797ACN-122027797-A

Abstract

The invention discloses an unmanned aerial vehicle inspection image transmission method and system based on asymmetric coding and decoding. The method comprises the steps of firstly identifying a device sensitive area and an environment non-sensitive area in an image through a semantic segmentation network, then carrying out encryption compression on the device sensitive area by adopting a compression sensing technology based on a public key mechanism, carrying out high-efficiency compression on the environment non-sensitive area by adopting a lightweight encoder, transmitting a processed code stream to a ground station through a wireless channel, carrying out differential reconstruction by adopting a high-performance decoder according to a user authority level by the ground station, carrying out joint optimization on a coder and a decoder by adopting a mixed objective function in a training stage, and finally completing inspection state analysis and decision by utilizing a reconstructed image. The invention realizes the safe and efficient transmission of the inspection image, improves the transmission efficiency while guaranteeing the safety of the sensitive data, and provides a complete solution for intelligent inspection of the unmanned aerial vehicle.

Inventors

LIU ZHUANG
FU JIAXING
LENG CONG
YANG QIANG
ZU XIANGDONG
ZHANG DAWEI
WANG XINGBO

Assignees

国网内蒙古东部电力有限公司
中国电力科学研究院有限公司
中科方寸知微(南京)科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260225

Claims (10)

1. An unmanned aerial vehicle inspection image transmission method based on an asymmetric coding and decoding image compression technology is characterized by comprising the following steps: s1, acquiring an original image of a patrol area by an unmanned aerial vehicle, and identifying and marking a sensitive area and an environmental non-sensitive area of equipment in the image; S2, performing asymmetric compression sampling and chaotic encryption processing on a device sensitive area by adopting a two-dimensional compression sensing technology based on a public key mechanism, and performing high-efficiency compression on an environment non-sensitive area through a lightweight encoder comprising a two-way feature extraction and detail enhancement module; S3, transmitting the compressed code stream subjected to the differentiation treatment to a ground station through a wireless channel; S4, the ground station carries out image reconstruction on the received code stream through a high-performance decoder, and according to the authority level of the user, completely reconstructs all area contents for authorized users and only reconstructs environment non-sensitive area contents for partial authorized users; s5, adopting a mixed objective function combining multi-scale image fidelity measurement and pixel level error measurement in a training stage to jointly optimize encoder and decoder parameters; and S6, the ground station completes inspection state analysis and decision by using the reconstructed image.
2. The unmanned aerial vehicle inspection image transmission method based on the asymmetric coding and decoding image compression technology according to claim 1, wherein the identifying and labeling the equipment sensitive area and the environment non-sensitive area in the image in S1 comprises: carrying out pixel-level region division on an original inspection image by adopting a semantic segmentation network based on deep learning, wherein: the equipment sensitive area comprises an electric equipment insulator, a transformer sleeve, a line fitting and a connecting part of the line fitting; the environmental non-sensitive areas include sky background, vegetation coverage and building outlines; The semantic segmentation network adopts an encoder-decoder architecture, wherein an encoder extracts multi-scale features by using ResNet-50 backbone networks, and a decoder fuses feature graphs of different scales through a feature pyramid network and enhances feature response of a key region by adopting a attention mechanism; In the region labeling stage, a hierarchical privacy label is automatically added to the identified equipment sensitive region, and the identified equipment sensitive region is divided into a primary sensitive region and a secondary sensitive region according to the equipment key degree, wherein the primary sensitive region corresponds to an equipment core component, an enhanced encryption strategy is adopted, and the secondary sensitive region corresponds to an equipment auxiliary component, and a standard encryption strategy is adopted.
3. The unmanned aerial vehicle inspection image transmission method based on the asymmetric coding and decoding image compression technology according to claim 2, wherein the step S2 of performing asymmetric compressed sampling and chaotic encryption processing on the device sensitive area by adopting a two-dimensional compressed sensing technology based on a public key mechanism comprises the following steps: Constructing a two-dimensional compressed sensing sampling matrix by adopting a public key mechanism based on elliptic curve cryptography, wherein the sampling matrix phi meets the finite equidistant property, and the sampling rate is dynamically configured according to the level of a sensitive area of the equipment, wherein the sampling rate is 0.4 for a first-level sensitive area and 0.5 for a second-level sensitive area; after compression sampling is completed, a random scrambling matrix is generated by adopting a three-dimensional hyper-chaotic system to carry out diffusion encryption on sampling data, wherein a state equation of the three-dimensional hyper-chaotic system is expressed as follows: , , , Wherein, the Representing the first The three-dimensional state variables of the system at the time of the iteration, Representing the first Three-dimensional state variables of the system during the next iteration; 、、 Deriving a system parameter from a hash value of a public key through a key expansion algorithm; and carrying out numerical normalization processing on the encrypted sampling data by adopting modulo 256 operation, and embedding initial parameters and regional position information of the chaotic system into a compressed sensing measured value through an orthogonal matrix to realize hidden transmission of key data.
4. The unmanned aerial vehicle inspection image transmission method based on the asymmetric coding and decoding image compression technology according to claim 2, wherein the efficient compression of the environment non-sensitive area in S2 by a lightweight encoder comprising a two-way feature extraction and detail enhancement module comprises the following steps: For an environmentally non-sensitive region, an efficient compression process is performed by a lightweight encoder comprising: a) The two-way feature extraction module is used for capturing space fine granularity features through partial window division and internal pixel association calculation in the first way, capturing semantic structure features through channel importance evaluation in the second way, and then aggregating the two ways of features; b) And the detail enhancement module is used for enhancing and screening key contour and texture information in the aggregated characteristics by adopting a feedforward network with a reverse structure and a characteristic selection switch.
5. The unmanned aerial vehicle inspection image transmission method based on the asymmetric coding and decoding image compression technology according to claim 4, wherein the efficient compression processing by the lightweight encoder comprises the following steps: In the two-way feature extraction module, a first way is divided by adopting non-overlapping local windows, and space fine granularity features are captured in each window by calculating covariance matrix feature values among pixels; the second path adopts the channel importance assessment based on the channel attention mechanism, calculates the importance weight of each channel through global average pooling and a full connection layer, and extracts the semantic structural features; The detail enhancement module adopts an inverse feed forward network structure, wherein the first layer expands the characteristic dimension, the middle layer adopts GELU activation functions to perform nonlinear transformation, and the last layer compresses the characteristic dimension back to the original dimension; At the output end of the encoder, the enhanced features are compressed by adopting an entropy coding technology based on arithmetic coding, wherein the probability model is adaptively updated according to the statistical characteristics of the environmental non-sensitive area features transmitted in a history mode.
6. The unmanned aerial vehicle inspection image transmission method based on the asymmetric coding and decoding image compression technology according to claim 1, wherein the step S3 of transmitting the compressed code stream subjected to the differentiation processing to a ground station through a wireless channel comprises the following steps: and respectively carrying out channel coding and packaging treatment on the encrypted compressed code stream of the sensitive area of the equipment and the high-efficiency compressed code stream of the environment non-sensitive area, wherein: Adopting a first type of error correction coding for the code stream of the sensitive area of the equipment; adopting a second type of error correction coding for the environmental non-sensitive area code stream; The two types of code streams are packaged into transmission frames according to a preset protocol, wherein the transmission frame structure comprises a frame header, a frame type identifier, a data length field, a payload data area and a check code, and the frame type identifier is used for distinguishing a device sensitive area code stream from an environment non-sensitive area code stream; in the transmission process, the modulation mode is adaptively adjusted according to the quality of the wireless channel; and establishing a transmission quality monitoring mechanism, counting the error rate and the packet loss rate in real time, and automatically triggering a transmission rate degradation and retransmission mechanism when the error rate of a plurality of continuous transmission frames exceeds a set threshold.
7. The unmanned aerial vehicle inspection image transmission method based on the asymmetric coding and decoding image compression technology according to claim 1, wherein the ground station in S4 performs image reconstruction through a high-performance decoder, and the method comprises the following steps: The high-performance decoder adopts a multi-scale feature fusion architecture based on an attention mechanism, and comprises a feature analysis module, a region reconstruction module and an image fusion module, wherein: The characteristic analysis module performs channel decoding and protocol analysis on the received code stream, and separates a device sensitive area code stream and an environment non-sensitive area code stream according to the frame type identification; the regional reconstruction module implements a differential reconstruction strategy aiming at users with different authorities; The image fusion module adopts a self-adaptive weighted fusion algorithm to seamlessly splice the reconstructed sensitive area and the reconstructed non-sensitive area, wherein a weighted average algorithm is used for eliminating splicing marks at the boundary of the areas; In the reconstruction process, the decoder performs data integrity verification according to the check code in the transmission frame, and requests retransmission of the data frame with failed check.
8. The unmanned aerial vehicle inspection image transmission method based on the asymmetric codec image compression technology according to claim 1, wherein the step S5 of jointly optimizing encoder and decoder parameters by using a hybrid objective function comprises the steps of: The mixed objective function is formed by weighting four items of multi-scale structural similarity loss, pixel level mean square error loss, characteristic perception loss and region self-adaptive loss; A progressive training strategy is adopted in the optimization process: the first stage uses an optimizer with weight attenuation to mainly optimize the compression reconstruction quality of an environment non-sensitive area at a first learning rate; Introducing a reconstruction task of the sensitive area in the second stage, reducing the learning rate to a second learning rate, and adding an edge protection loss term to improve the holding capacity of the edge details of the sensitive area of the equipment; performing end-to-end joint fine tuning in the third stage, wherein the learning rate is attenuated by adopting a cosine annealing strategy, and simultaneously, a gradient clipping strategy is introduced to limit gradient norms; Training data enhancement includes targeted enhancement strategies based on unmanned aerial vehicle inspection scene characteristics, including random rotation, brightness adjustment, gaussian noise addition, atmospheric turbulence simulation, and sensor noise simulation.
9. The unmanned aerial vehicle inspection image transmission method based on the asymmetric coding and decoding image compression technology according to claim 1, wherein the ground station in S6 completes inspection state analysis and decision by using the reconstructed image, and the method comprises the following steps: Establishing a multi-task analysis model, and executing the following analysis tasks in parallel, namely detecting equipment defects, namely identifying part defects in a sensitive area of the equipment, evaluating environmental risks, namely identifying potential risks in a non-sensitive area of the environment and evaluating risk grades; constructing a hierarchical decision support system, and generating treatment suggestions according to analysis results, namely generating maintenance work orders for general defects and arranging a patrol plan; triggering real-time warning and pushing information for serious hidden danger; The method comprises the steps of establishing a visual feedback mechanism, superposing and displaying analysis results on a reconstructed image, marking equipment defect levels by using color codes, displaying environmental risk distribution by using thermodynamic diagrams, and displaying state change prediction through trend curves; and correlating the analysis result with the historical data, and establishing a full life cycle health file of the equipment.
10. An unmanned aerial vehicle inspection image transmission system based on an asymmetric coding and decoding image compression technology, which is applied to the unmanned aerial vehicle inspection image transmission method based on the asymmetric coding and decoding image compression technology as claimed in any one of claims 1 to 9, and is characterized in that the system comprises: The image acquisition and region identification module is used for acquiring an original image of the inspection region by the unmanned aerial vehicle, and identifying and marking a device sensitive region and an environment non-sensitive region in the image; The asymmetric coding processing module is used for carrying out asymmetric compression sampling and chaotic encryption processing on the equipment sensitive area by adopting a two-dimensional compression sensing technology based on a public key mechanism, and simultaneously carrying out high-efficiency compression on the environment non-sensitive area through a lightweight encoder comprising a two-way feature extraction and detail enhancement module; the secure transmission module is used for transmitting the compressed code stream subjected to the differentiation processing to the ground station through a wireless channel; the high-performance decoding and reconstructing module is used for reconstructing images of the received code stream by the ground station through the high-performance decoder, completely reconstructing all area contents of authorized users according to the user authority level, and reconstructing environment non-sensitive area contents only for partial authorized users; The model training optimization module is used for jointly optimizing the encoder and decoder parameters by adopting a mixed objective function combining the multi-scale image fidelity measurement and the pixel level error measurement in the training stage; and the inspection analysis and decision module is used for finishing inspection state analysis and decision by the ground station by using the reconstructed image.

Description

Unmanned aerial vehicle inspection image transmission method and system based on asymmetric coding and decoding image compression technology Technical Field The invention relates to the technical field of image processing and transmission, in particular to an unmanned aerial vehicle inspection image transmission method and system based on an asymmetric coding and decoding image compression technology. Background With the wide application of unmanned aerial vehicle technology in the fields of electric power inspection, pipeline monitoring and the like, how to efficiently and safely transmit and process inspection image data has become an important point of industry attention. The images collected by the unmanned aerial vehicle in the inspection process generally comprise equipment sensitive information (such as power equipment details) and environment non-sensitive information (such as sky and vegetation), and the traditional image transmission method usually adopts unified compression and transmission strategies, so that the safety and the transmission efficiency cannot be considered. The main defects of the current mainstream image transmission method are that firstly, the same compression standard is adopted for the whole image, the differentiation processing is not carried out according to the region importance, so that the sensitive information is not protected enough or the transmission efficiency is low, secondly, the encryption means is single and is difficult to resist complex network attacks, thirdly, the differentiation reconstruction mechanism based on the user authority is lacking, the multi-level security management requirement cannot be met, and fourthly, the adaptability and the reconstruction quality of the existing codec in a complex inspection environment are required to be improved. Furthermore, traditional methods are relatively weak in terms of image analysis and decision support, and lack a complete solution to effectively combine image transmission with intelligent analysis. Therefore, there is a need for an unmanned aerial vehicle inspection image transmission method that can realize safe and efficient transmission, support hierarchical reconstruction, and have intelligent analysis capability. Disclosure of Invention The invention aims to overcome the defects of the prior art and provides an unmanned aerial vehicle inspection image transmission scheme capable of considering both safety and transmission efficiency. In a first aspect, an embodiment of the present application provides a method for transmitting an inspection image of an unmanned aerial vehicle based on an asymmetric codec image compression technique, where the method includes: s1, acquiring an original image of a patrol area by an unmanned aerial vehicle, and identifying and marking a sensitive area and an environmental non-sensitive area of equipment in the image; S2, performing asymmetric compression sampling and chaotic encryption processing on a device sensitive area by adopting a two-dimensional compression sensing technology based on a public key mechanism, and performing high-efficiency compression on an environment non-sensitive area through a lightweight encoder comprising a two-way feature extraction and detail enhancement module; S3, transmitting the compressed code stream subjected to the differentiation treatment to a ground station through a wireless channel; S4, the ground station carries out image reconstruction on the received code stream through a high-performance decoder, and according to the authority level of the user, completely reconstructs all area contents for authorized users and only reconstructs environment non-sensitive area contents for partial authorized users; s5, adopting a mixed objective function combining multi-scale image fidelity measurement and pixel level error measurement in a training stage to jointly optimize encoder and decoder parameters; and S6, the ground station completes inspection state analysis and decision by using the reconstructed image. In a second aspect, an embodiment of the present application provides an unmanned aerial vehicle inspection image transmission system based on an asymmetric codec image compression technology, which is applied to the unmanned aerial vehicle inspection image transmission method based on the asymmetric codec image compression technology according to the first aspect, where the system includes: The image acquisition and region identification module is used for acquiring an original image of the inspection region by the unmanned aerial vehicle, and identifying and marking a device sensitive region and an environment non-sensitive region in the image; The asymmetric coding processing module is used for carrying out asymmetric compression sampling and chaotic encryption processing on the equipment sensitive area by adopting a two-dimensional compression sensing technology based on a public key mechanism, and simultaneously carrying out high