Over-the-horizon individual image transmission relay enhancement device and method for complex terrain

Abstract

The invention belongs to the technical field of image communication, and discloses a beyond-visual-range single-soldier image transmission relay enhancement device and method for complex terrains. The method comprises the steps of collecting target image data; acquiring three-dimensional point cloud data of an environment where an individual soldier is located; dividing the three-dimensional point cloud data to obtain N point cloud data subspaces; performing feature extraction on the N point cloud data subspaces to obtain environment feature data of the N point cloud data subspaces; collecting network operation data of the N point cloud data subspaces; performing feature extraction on the network operation data of the N point cloud data subspaces to obtain network feature data of the N point cloud data subspaces; based on the environment feature data and the network feature data of the N point cloud data subspaces, carrying out adaptive optimization on a transmission path of the individual relay equipment; according to the scheme, the transmission path of the single-soldier relay equipment is adaptively optimized, and the problem that a traditional transmission link is insufficient in adaptive capacity in complex terrains is effectively solved.

Description

Technical Field

[0001] The present invention relates to the field of image communication technology. More specifically, the present invention relates to an over-the-horizon individual image transmission relay enhancement device and method for complex terrains. Background Art

[0002] Existing over-the-horizon image transmission technologies have many deficiencies when facing complex terrains. In complex terrains, traditional transmission link designs lack the ability to adaptively adjust according to the terrain and real-time signal conditions. When encountering sudden situations such as terrain undulations and signal interferences, it is difficult to quickly and flexibly optimize the transmission path and parameters, which results in poor signal transmission stability, interruptions or delays in actual applications, affects the timeliness of information, and further leads to signal strength attenuation, making it difficult to meet the requirements of long-distance and high-quality image transmission, and restricting the wide application and effectiveness of individual image transmission relay technologies in complex environments.

[0003] Therefore, there is an urgent need for an optimized solution that can adaptively adjust the transmission link in real time according to complex terrains and signal changes to meet the urgent needs of military operations and complex environment operations for long-distance and high-quality image transmission, and ensure efficient and stable image transmission effects in various complex terrains. Summary of the Invention

[0004] To overcome the above-mentioned defects of the prior art and achieve the above object, the present invention provides the following technical solution: An over-the-horizon individual image transmission relay enhancement method for complex terrains, including:

[0005] Collecting target image data;

[0006] Collecting three-dimensional point cloud data of the environment where the individual is located;

[0007] Dividing the three-dimensional point cloud data to obtain N point cloud data subspaces;

[0008] Extracting feature data of the N point cloud data subspaces to obtain environmental feature data of the N point cloud data subspaces;

[0009] Collecting network operation data of the N point cloud data subspaces;

[0010] Extracting feature data of the network operation data of the N point cloud data subspaces to obtain network feature data of the N point cloud data subspaces;

[0011] Adaptive optimization of the transmission path of the individual relay device based on the environmental feature data and network feature data of the N point cloud data subspaces.

[0012] Furthermore, the method for adaptively optimizing the transmission path of the individual relay device includes:

[0013] Input the environmental feature data and network feature data of N point cloud data subspaces into the fitness evaluation model respectively to obtain the fitness evaluation scores corresponding to the N point cloud data subspaces;

[0014] Construct a set of nodes to be planned for transmission by using the point cloud data subspaces whose fitness evaluation scores are greater than or equal to the preset fitness evaluation score threshold;

[0015] Connect the nodes to be planned for transmission in the set of nodes to be planned for transmission to construct G sub-segments of transmission paths, where G is an integer greater than zero;

[0016] For each sub-segment of the transmission path, calculate the corresponding weight of the sub-segment of the transmission path according to the environmental feature data and network feature data at the start point and end point of the sub-segment of the transmission path; construct a set of weights of the sub-segments of the transmission path from the G weights of the sub-segments of the transmission path;

[0017] Optimize the G sub-segments of the transmission path based on the set of weights of the sub-segments of the transmission path to obtain the optimal transmission path; enable the single-soldier relay device to transmit the target image data according to the optimal transmission path.

[0018] Further, the method for obtaining the optimal transmission path includes:

[0019] Preset the number of sub-segments of the transmission path as R, where R is an integer greater than zero; initialize the optimal transmission path as empty;

[0020] Select the sub-segment of the transmission path with the largest weight from the set of weights of the sub-segments of the transmission path, denoted as the sub-segment of the transmission path to be planned, and remove the sub-segment of the transmission path to be planned from the set of weights of the sub-segments of the transmission path;

[0021] Add the sub-segment of the transmission path to be planned to the optimal transmission path, and determine whether there is a loop in the optimal transmission path. If there is a loop in the optimal transmission path, remove the sub-segment of the transmission path to be planned from the optimal transmission path; if there is no loop in the optimal transmission path, retain the sub-segment of the transmission path to be planned in the optimal transmission path;

[0022] Repeat the above process until the number of sub-segments of the transmission path in the optimal transmission path reaches R and then stop.

[0023] Further, the method for determining whether there is a loop in the optimal transmission path includes:

[0024] Randomly select a sub-segment of the transmission path from the optimal transmission path as the starting point, and traverse along the selected sub-segment of the transmission path; if the starting position is returned during the traversal process and no sub-segment of the transmission path is repeatedly passed through in the whole path, there is a loop in the optimal transmission path.

[0025] Further, the method for obtaining the N point cloud data subspaces includes:

[0026] Step 1: Divide the three-dimensional point cloud data according to a preset sliding step length to obtain N central data points of the three-dimensional point cloud data; let the initial value of n be 1, and the value range of n is from 1 to N;

[0027] Step 2: Obtain the nth central data point, and select H point cloud data points adjacent to the nth central data point from the three-dimensional point cloud data; construct the nth point cloud data subspace with the nth central data point and the H point cloud data points;

[0028] Step 3: Let n = n + 1. If n is less than or equal to N, continue to execute Step 2. If n is greater than N, end the current process to obtain N point cloud data subspaces.

[0029] Furthermore, the method for obtaining the environmental feature data of the N point cloud data subspaces includes:

[0030] S100: Let the initial value of n be 1, and the value range of n is from 1 to N;

[0031] S101: Based on the three-dimensional coordinates of the central data point and the corresponding H point cloud data points in the nth point cloud data subspace, calculate the environmental slope and environmental surface roughness of the nth point cloud data subspace; based on the incident laser intensity and reflected laser intensity of the central data point and the corresponding H point cloud data points in the nth point cloud data subspace, calculate the electromagnetic wave reflectivity of the nth point cloud data subspace;

[0032] S102: Extract features from the environmental slope, environmental surface roughness, and electromagnetic wave reflectivity according to a preset method to obtain the electromagnetic wave refractive index of the nth point cloud data subspace;

[0033] S103: Construct the environmental feature data of the nth point cloud data point with the environmental slope, environmental surface roughness, electromagnetic wave reflectivity, and electromagnetic wave refractive index;

[0034] S104: Let n = n + 1. If n is less than or equal to N, continue to execute S101 to S103. If n is greater than N, obtain the corresponding environmental feature data of the N point cloud data subspaces and end the current process.

[0035] Furthermore, the method for obtaining the electromagnetic wave refractive index of the nth point cloud data subspace includes:

[0036] Input the environmental slope, environmental surface roughness, and electromagnetic wave reflectivity of the nth point cloud data subspace into the medium type diagnosis model to obtain the medium type of the nth point cloud data subspace;

[0037] Obtain the electromagnetic wave propagation speed corresponding to the medium type of the nth point cloud data subspace from the pre-built medium-propagation speed database; calculate the electromagnetic wave refractive index of the nth point cloud data point based on the speed of light and the electromagnetic wave propagation speed.

[0038] Furthermore, the method for obtaining the network feature data of N point cloud data subspaces includes:

[0039] S200: Let the initial value of n be 1, and the value range of n is from 1 to N;

[0040] S201: Calculate the corresponding channel capacity based on the network bandwidth, signal power, and noise power of the nth point cloud data subspace; calculate the corresponding link quality index based on the signal power, noise power, and network packet loss rate of the nth point cloud data subspace; calculate the link throughput of the nth point cloud data subspace based on the network packet loss rate and network bandwidth; calculate the link stability factor of the nth point cloud data subspace based on the link quality index and network latency;

[0041] S202: Construct the channel capacity, link quality index, link throughput, and link stability factor into the network feature data of the nth point cloud data subspace;

[0042] S203: Let n = n + 1. If n is less than or equal to N, continue to execute S201 to S102. If n is greater than N, construct the network feature data of N point cloud data subspaces into a network feature data set, and end the current process.

[0043] Furthermore, the training method of the fitness evaluation model includes:

[0044] Pre-collect a fitness evaluation data set, which includes K groups of fitness evaluation data and the corresponding fitness evaluation scores of K groups of fitness evaluation data. K is a positive integer greater than 0. The fitness evaluation data includes environmental feature data and network feature data; divide the fitness evaluation data set into a training set and a validation set, where the training set is used to train the fitness evaluation model, and the validation set is used to evaluate the generalization performance of the fitness evaluation model;

[0045] During the training process of the fitness evaluation model, minimize the cross-entropy loss function as the optimization goal, use the early stopping strategy to monitor the performance of the validation set, and optimize the model performance by continuously adjusting the network parameters; when the prediction accuracy on the validation set reaches the expected accuracy, it is considered that the fitness evaluation model has converged, and stop training; the fitness evaluation model is trained using a deep neural network based on a multi-layer perceptron.

[0046] Convert the fitness evaluation data into a feature vector. The input layer of the fitness evaluation model receives the high-dimensional feature vector, extracts the non-linear relationships in the data through multiple hidden layers, and finally the output layer of the fitness evaluation model calculates the probability distribution of the fitness evaluation scores through the softmax activation function, and outputs the fitness evaluation score corresponding to the maximum probability as the final prediction result.

[0047] An over-the-horizon individual soldier image transmission relay enhancement device for complex terrains, which is used to implement the over-the-horizon individual soldier image transmission relay enhancement method for complex terrains, includes:

[0048] An image acquisition module, which is used to acquire target image data;

[0049] A first acquisition module, which is used to acquire the three-dimensional point cloud data of the environment where the individual soldier is located;

[0050] A first processing module, which is used to divide the three-dimensional point cloud data to obtain N point cloud data subspaces;

[0051] A second processing module, which is used to extract features from the N point cloud data subspaces to obtain the environmental feature data of the N point cloud data subspaces;

[0052] A second acquisition module, which is used to acquire the network operation data of the N point cloud data subspaces;

[0053] A third processing module, which is used to extract features from the network operation data of the N point cloud data subspaces to obtain the network feature data of the N point cloud data subspaces;

[0054] A path optimization module, which adaptively optimizes the transmission path of the individual soldier relay device based on the environmental feature data and network feature data of the N point cloud data subspaces.

[0055] Compared with the prior art, the technical effects and advantages of the over-the-horizon individual soldier image transmission relay enhancement device and method for complex terrains of the present invention are as follows:

[0056] First, based on lidar, millimeter-wave radar and multi-camera vision, the three-dimensional point cloud data of the environment where the individual soldier is located is collected in real time. By dividing the point cloud data subspaces and extracting key features such as environmental slope, surface roughness, electromagnetic wave reflectivity and refractive index, and combining with the medium type diagnosis model, the physical characteristics of terrain obstacles are accurately identified, providing high-precision environmental data support for signal propagation path optimization.

[0057] Secondly, by dynamically monitoring network operation parameters (such as signal power, noise power, delay, packet loss rate, etc.), the channel capacity, link quality index, throughput, and link stability factor are calculated in real time, a multi-dimensional network feature index system is constructed, and the link availability is comprehensively evaluated. On this basis, an adaptation degree evaluation model is adopted to comprehensively integrate environmental feature data and network feature data, dynamically generate an adaptation degree score, screen the optimal relay nodes, and select the transmission path through an adaptive multi-hop relay mechanism.

[0058] Finally, by dynamically adjusting the transmission path and relay node deployment, it is ensured that low-latency and highly reliable image data transmission is achieved in complex terrains, avoiding the signal interruption problem caused by fixed paths in traditional technologies. In addition, through loop-free transmission path planning, packet cyclic redundancy is effectively prevented, and the network resource utilization rate is improved. In military operations, it can help commanders obtain high-definition images of the battlefield situation in real time and improve decision-making efficiency; in scenarios such as emergency rescue and geological exploration, it can quickly establish long-distance communication links, ensure the efficient backhaul of on-site data, and the safe execution of tasks, and has broad military and civilian application values.

[0059] This solution innovatively integrates three-dimensional environmental feature dynamic modeling and network link intelligent optimization to adaptively optimize the transmission path of the single-soldier relay device, effectively solving the problems of insufficient adaptability, serious signal attenuation, and poor transmission stability of traditional transmission links in complex terrains, and improving the data transmission distance and quality in over-the-horizon scenarios. Brief Description of the Drawings

[0060] Figure 1 Structural diagram of the over-the-horizon single-soldier image transmission relay enhancement device for complex terrains in Embodiment 1 of the present invention;

[0061] Figure 2 Flowchart of the method for adaptively optimizing the transmission path of the single-soldier relay device in Embodiment 3 of the present invention;

[0062] Figure 3 Structural diagram of the over-the-horizon single-soldier image transmission relay enhancement device for complex terrains in Embodiment 2 of the present invention;

[0063] Figure 4 Flowchart of the method for adaptively optimizing the transmission path of the single-soldier relay device;

[0064] Figure 5 Flowchart of the method for obtaining the optimal transmission path;

[0065] Figure 6 Flowchart of the method for obtaining the environmental feature data of N point cloud data subspaces;

[0066] Figure 7Flowchart of the method for obtaining network feature data of N point cloud data subspaces. Detailed implementation manners

[0067] Next, the technical solutions in the embodiments of the present invention will be described in detail, clearly and completely with reference to the accompanying drawings in the embodiments of the present invention. It should be specifically noted that the following described specific embodiments are only used to better illustrate and explain the technical solutions of the present invention, aiming to enable those skilled in the art to better understand and implement the present invention, and should not be construed as a limitation on the protection scope of the present invention. Without departing from the spirit and essence of the present invention, those skilled in the art can modify, adjust or equivalently replace it according to the content disclosed in the present invention, and these should all be regarded as the protection scope of the present invention.

[0068] Embodiment 1

[0069] Please refer to Figure 1 As shown, the ultra-long-range single-soldier video transmission relay enhancement device for complex terrains in this embodiment includes an image acquisition module, a first acquisition module, a first processing module, a second processing module, a second acquisition module, a third processing module, and a path optimization module. Each module is connected by wire and / or wirelessly to achieve data transmission.

[0070] The image acquisition module is used to acquire target image data; the target image data is acquired by remote devices such as drones, surveillance cameras, and unmanned vehicles.

[0071] It should be noted that in the scenario of ultra-long-range video transmission, when the target image data is being transmitted, the signal is blocked by terrain, buildings, or obstacles, and the target image data cannot be directly transmitted to the terminal. In this case, a single-soldier relay device (a relay device carried by a single soldier) is required for relaying. The relay device carried by a single soldier is a lightweight, low-power, and intelligent wireless communication node, mainly used to enhance the data transmission ability in an ultra-long-range (NLOS) environment and ensure the communication stability between the front-end acquisition device and the back-end command center or server. The relay device can not only act as a data transfer station to forward information between different network nodes, but also optimize the data transmission path, thereby improving the quality and coverage of data transmission.

[0072] The first acquisition module is used to acquire three-dimensional point cloud data of the environment where the single soldier is located; the three-dimensional point cloud data is obtained by a lidar, a millimeter-wave radar, and a multi-camera vision system.

[0073] The first processing module is used to divide the three-dimensional point cloud data to obtain N point cloud data subspaces.

[0074] The method for obtaining N point cloud data subspaces includes:

[0075] Step 1: Divide the three-dimensional point cloud data according to a preset sliding step size to obtain N central data points of the three-dimensional point cloud data; Let the initial value of n be 1, and the value range of n is from 1 to N;

[0076] Step 2: Obtain the nth central data point, and select H point cloud data points adjacent to the nth central data point from the three-dimensional point cloud data; Construct the nth point cloud data subspace with the nth central data point and the H point cloud data points;

[0077] Step 3: Let n = n + 1. If n is less than or equal to N, continue to execute Step 2. If n is greater than N, end the current process to obtain N point cloud data subspaces.

[0078] The second processing module is used to extract features from the N point cloud data subspaces to obtain the environmental feature data of the N point cloud data subspaces.

[0079] As Figure 6 shown, the method for obtaining the environmental feature data of the N point cloud data subspaces includes:

[0080] S100: Let the initial value of n be 1, and the value range of n is from 1 to N;

[0081] S101: Based on the three-dimensional coordinates of the central data point and the corresponding H point cloud data points in the nth point cloud data subspace, calculate the environmental slope and environmental surface roughness of the nth point cloud data subspace; Based on the incident laser intensity and reflected laser intensity of the central data point and the corresponding H point cloud data points in the nth point cloud data subspace, calculate the electromagnetic wave reflectivity of the nth point cloud data subspace;

[0082] S102: Extract features from the environmental slope, environmental surface roughness, and electromagnetic wave reflectivity according to a preset method to obtain the electromagnetic wave refractive index of the nth point cloud data subspace;

[0083] S103: Construct the environmental feature data of the nth point cloud data point with the environmental slope, environmental surface roughness, electromagnetic wave reflectivity, and electromagnetic wave refractive index;

[0084] S104: Let n = n + 1. If n is less than or equal to N, continue to execute S101 to S103. If n is greater than N, obtain the corresponding environmental feature data of the N point cloud data subspaces and end the current process.

[0085] The method for obtaining the electromagnetic wave refractive index of the nth point cloud data subspace includes:

[0086] Input the environmental slope, environmental surface roughness, and electromagnetic wave reflectivity of the nth point cloud data subspace into the medium type diagnosis model to obtain the medium type of the nth point cloud data subspace; the medium types include metal, glass, water, concrete, and vegetation;

[0087] Obtain the electromagnetic wave propagation speed corresponding to the medium type of the nth point cloud data subspace from the pre-constructed medium-propagation speed database; calculate the electromagnetic wave refractive index of the nth point cloud data point based on the speed of light and the electromagnetic wave propagation speed.

[0088] The calculation method of the electromagnetic wave refractive index includes:

[0089]

[0090] where ZSL n is the electromagnetic wave refractive index of the nth point cloud data subspace, GS is the speed of light, and the value of the speed of light is 3×10 8 m / s, and CBSD n is the electromagnetic wave propagation speed of the medium type of the nth point cloud data subspace.

[0091] The calculation method of the environmental slope includes:

[0092]

[0093] where HJPD n is the environmental slope of the nth central data point, π is a constant, and the value of π is 3.14159, XFC n is the covariance matrix, H is the number of point cloud data points, h ≤ H, SWZB n is the three-dimensional coordinate of the nth central data point, is the transposed matrix corresponding to the calculation result of, |XFC n | is the value of the determinant corresponding to the covariance matrix; is the eigenvector of the covariance matrix.

[0094] For example, the calculation result of is (7, 5, 4), and the corresponding transposed matrix is

[0095] The calculation method of the environmental surface roughness includes:

[0096]

[0097] where BMCD n is the environmental surface roughness of the nth central data point, XZB n is the X-axis coordinate of the nth central data point, XZBh is the X-axis coordinate of the h-th point cloud data point; YZB n is the Y-axis coordinate of the n-th central data point, YZB h is the Y-axis coordinate of the h-th point cloud data point; ZZB n is the Z-axis coordinate of the n-th central data point, ZZB h is the Z-axis coordinate of the h-th point cloud data point.

[0098] The calculation method of the electromagnetic wave reflectivity includes:

[0099]

[0100] Among them, FSL n is the electromagnetic wave reflectivity of the n-th central data point, FS n is the reflected laser intensity of the n-th central data point, RS n is the incident laser intensity of the n-th central data point, FS h is the reflected laser intensity of the h-th point cloud data point, RS h is the incident laser intensity of the h-th point cloud data point.

[0101] The training method of the medium type diagnosis model includes:

[0102] Pre-construct a medium type diagnosis data set, which includes Y groups of medium type diagnosis data and the corresponding medium types of Y groups of medium type diagnosis data. Y is a positive integer greater than 0. The medium type diagnosis data includes environmental slope, environmental surface roughness, and electromagnetic wave reflectivity; divide the medium type diagnosis data set into a medium type diagnosis data training set and a medium type diagnosis data validation set. Among them, the medium type diagnosis data training set is used for parameter learning of the medium type diagnosis model, and the medium type diagnosis data validation set is used for real-time evaluation of the generalization ability of the medium type diagnosis model;

[0103] During the training process of the medium type diagnosis model, adopt a deep neural network structure based on a multi-layer perceptron. Convert the medium type diagnosis data into a feature vector as the input, extract the non-linear features in the data through the hidden layer, and finally generate the probability distribution of the medium type using the softmax activation function in the output layer. Output the medium type corresponding to the maximum probability as the final prediction result; the training process aims to minimize the cross-entropy loss function, and at the same time introduce an early stopping strategy to monitor the performance of the medium type diagnosis data validation set. When the prediction accuracy on the medium type diagnosis data validation set reaches the preset threshold, it is considered that the medium type diagnosis model has converged, and the training stops immediately.

[0104] The softmax activation function is:

[0105]

[0106] Among them, Softmax(JZLX num ) is the output probability corresponding to the num-th feature vector, JZLX num is the num-th feature vector, NUM is the total number of feature vectors, and e is a constant.

[0107] The second acquisition module is used to acquire the network operation data of N point cloud data subspaces; the network operation data includes signal power, noise power, network delay, network packet loss rate, network bandwidth, etc.

[0108] The third processing module is used to extract features from the network operation data of N point cloud data subspaces to obtain the network feature data of N point cloud data subspaces.

[0109] As Figure 7 shown, the method for obtaining the network feature data of N point cloud data subspaces includes:

[0110] S200: Let the initial value of n be 1, and the value range of n is from 1 to N;

[0111] S201: Based on the network bandwidth, signal power, and noise power of the n-th point cloud data subspace, calculate the corresponding channel capacity; based on the signal power, noise power, and network packet loss rate of the n-th point cloud data subspace, calculate the corresponding link quality index; based on the network packet loss rate and network bandwidth, calculate the link throughput of the n-th point cloud data subspace; based on the link quality index and network delay, calculate the link stability factor of the n-th point cloud data subspace;

[0112] S202: Construct the channel capacity, link quality index, link throughput, and link stability factor into the network feature data of the n-th point cloud data subspace;

[0113] S203: Let n = n + 1. If n is less than or equal to N, then continue to execute S201 to S102. If n is greater than N, then construct the network feature data of N point cloud data subspaces into a network feature data set, and end the current process.

[0114] The calculation method of the channel capacity includes:

[0115] XDRL n = WLDK × log 2 (1 + (XHGL n - ZSGL n ));

[0116] Among them, XDRL nis the channel capacity of the nth point cloud data subspace. The larger the channel capacity, the stronger the data transmission ability; WLDK is the network bandwidth; XHGL n is the signal power of the nth point cloud data subspace, ZSGL n is the noise power of the nth point cloud data subspace; log 2 (·) is the logarithmic function with base 2.

[0117] The calculation method of the link quality index includes:

[0118]

[0119] Among them, ZLZS n is the link quality index of the nth point cloud data subspace, DBL n is the network packet loss rate of the nth point cloud data subspace. The higher the link quality index, the better the link quality and the more suitable for routing decisions in multi-hop networks.

[0120] The method for obtaining the link throughput includes:

[0121] TTL n = WLDK×(1 - DBL n )

[0122] Among them, TTL n is the link throughput of the nth point cloud data subspace, which is used to reflect the actual available bandwidth of the link, rather than just the theoretical bandwidth.

[0123] The method for obtaining the link stability factor includes:

[0124]

[0125] Among them, WDYZ n is the link stability factor of the nth point cloud data subspace, WLYC n is the network delay of the nth point cloud data subspace. The link stability factor reflects the balance between link quality and delay. The higher the link stability factor, the better the link quality and the more suitable for path planning.

[0126] The path optimization module adaptively optimizes the transmission path of the single-soldier relay device based on the environmental feature data and network feature data of N point cloud data subspaces.

[0127] As Figure 4 shown, the method for adaptively optimizing the transmission path of the single-soldier relay device includes:

[0128] Input the environmental feature data and network feature data of N point cloud data subspaces into the fitness evaluation model respectively to obtain the fitness evaluation scores corresponding to the N point cloud data subspaces; the fitness evaluation scores are used to measure the fitness degree of the point cloud data subspaces as relay nodes of the transmission path. The higher the fitness evaluation score, the more suitable the point cloud data subspace is as a relay node of the transmission path.

[0129] Construct a set of to-be-planned transmission nodes by using the point cloud data subspaces whose fitness evaluation scores are greater than or equal to the preset fitness evaluation score threshold.

[0130] Connect the to-be-planned transmission nodes in the set of to-be-planned transmission nodes to construct G sub-segments of transmission paths, where G is an integer greater than zero.

[0131] For each sub-segment of the transmission path, calculate the corresponding weight of the sub-segment of the transmission path according to the environmental feature data and network feature data at the start point and end point of the sub-segment of the transmission path; construct a set of weights of the sub-segments of the transmission path from the G weights of the sub-segments of the transmission path.

[0132] Based on the set of weights of the sub-segments of the transmission path, optimize the G sub-segments of the transmission path to obtain the optimal transmission path; enable the single-soldier relay device to transmit the target image data according to the optimal transmission path.

[0133] As Figure 5 shown, the method for obtaining the optimal transmission path includes:

[0134] Preset the number of sub-segments of the transmission path as R, where R is an integer greater than zero; initialize the optimal transmission path as empty.

[0135] Select the sub-segment of the transmission path with the largest weight from the set of weights of the sub-segments of the transmission path, denoted as the to-be-planned sub-segment of the transmission path, and remove the to-be-planned sub-segment of the transmission path from the set of weights of the sub-segments of the transmission path.

[0136] Add the to-be-planned sub-segment of the transmission path to the optimal transmission path, and judge whether there is a loop in the optimal transmission path. If there is a loop in the optimal transmission path, remove the to-be-planned sub-segment of the transmission path from the optimal transmission path; if there is no loop in the optimal transmission path, retain the to-be-planned sub-segment of the transmission path in the optimal transmission path.

[0137] Repeat the above process until the number of sub-segments of the transmission path in the optimal transmission path reaches R and then stop.

[0138] The method for judging whether there is a loop in the optimal transmission path includes:

[0139] Randomly select a sub - segment of the transmission path from the optimal transmission path as the starting point, and traverse along the selected sub - segment of the transmission path; if the starting position is reached during the traversal and no sub - segment of the transmission path is repeatedly passed through in the entire path, then there is a loop in the optimal transmission path.

[0140] It should be noted that during the process of the optimal transmission path, the goal is to construct a connected but loop - free network transmission path. The loop - free property is to avoid the infinite loop of data packets in the loop and reduce redundant paths. The definition of a loop is: if starting from a certain sub - segment of the transmission path, after passing through several sub - segments of the transmission path and finally returning to the starting node, and there are no repeated sub - segments of the transmission path in the entire transmission path, then this transmission path forms a loop.

[0141] The method for obtaining the weight of the transmission path sub - segment includes:

[0142] ZDQZ g =HJZH g +WLZH g ;

[0143]

[0144] Among them, ZDQZ g is the weight of the g - th sub - segment of the transmission path, HJZH g is the environmental feature weight of the g - th sub - segment of the transmission path; ZSL (g，s) is the refractive index of electromagnetic waves at the starting point of the g - th sub - segment of the transmission path, and ZSL (g，e) is the refractive index of electromagnetic waves at the end point of the g - th sub - segment of the transmission path; FSL (g，s) is the reflectivity of electromagnetic waves at the starting point of the g - th sub - segment of the transmission path, and FSL (g，e) is the reflectivity of electromagnetic waves at the end point of the g - th sub - segment of the transmission path; HJPD (g，e) is the environmental slope at the end point of the g - th sub - segment of the transmission path, and HJPD (g，s) is the environmental slope at the starting point of the g - th sub - segment of the transmission path; BMCD (g，e) is the environmental surface roughness at the end point of the g - th sub - segment of the transmission path, and BMCD (g，s) is the environmental surface roughness at the starting point of the g - th sub - segment of the transmission path, sec(·) is the secant function, and csc(·) is the cosecant function.

[0145] WLZH g is the network feature weight of the g - th sub - segment of the transmission path, XDRL (g，s) is the channel capacity at the starting point of the g - th sub - segment of the transmission path, and XDRL (g，e) is the channel capacity at the end point of the g - th sub - segment of the transmission path; ZLZS (g，e)The link quality index at the end of the g-th sub-segment of the transmission path, ZLZS (g，s) The link quality index at the start of the g-th sub-segment of the transmission path; WDYZ (g，s) The link stability factor at the start of the g-th sub-segment of the transmission path, WDYZ (g，e) The link stability factor at the end of the g-th sub-segment of the transmission path; TTL (g，s) The link throughput at the start of the g-th sub-segment of the transmission path, TTL (g，e) The link throughput at the end of the g-th sub-segment of the transmission path; Ln(·) is the logarithmic function, and e is a constant.

[0146] It should be noted that the refraction and reflection characteristics of electromagnetic waves will significantly affect the propagation path and intensity of signals. The use of secant and cosecant functions can reflect the comprehensive impact of different combinations of refraction and reflection rates on transmission. In complex terrains, the slope and surface roughness will affect the scattering of signals. By calculating the difference ratio of environmental slopes and the difference ratio of environmental surface roughness, the impact of terrain changes on signal transmission can be reflected. The channel capacity determines the amount of data that can be transmitted, the link quality index reflects the transmission quality, and logarithmic operations can balance the magnitude differences between these two important indicators of channel capacity and link quality index, and reasonably reflect their comprehensive impact on the transmission path. Link stability is directly related to the reliability of transmission, throughput affects transmission efficiency, and exponential functions can reflect the positive impact on the weight of the transmission path as the stability factor increases and the throughput increases.

[0147] The training method of the fitness evaluation model includes:

[0148] Pre-collect a fitness evaluation data set, which includes K groups of fitness evaluation data and the corresponding fitness evaluation scores for the K groups of fitness evaluation data. K is a positive integer greater than 0. The fitness evaluation data includes environmental feature data and network feature data; divide the fitness evaluation data set into a training set and a validation set, where the training set is used to train the fitness evaluation model, and the validation set is used to evaluate the generalization performance of the fitness evaluation model;

[0149] During the training process of the fitness evaluation model, minimize the cross-entropy loss function as the optimization goal, use the early stopping strategy to monitor the performance of the validation set, and optimize the model performance by continuously adjusting the network parameters; when the prediction accuracy rate on the validation set reaches the expected accuracy rate, it is considered that the fitness evaluation model has converged and stop training; the fitness evaluation model is trained using a deep neural network based on a multi-layer perceptron.

[0150] Convert the fitness evaluation data into feature vectors; the input layer of the fitness evaluation model receives high-dimensional feature vectors, extracts the non-linear relationships in the data through multiple hidden layers, and finally the output layer of the fitness evaluation model calculates the probability distribution of the fitness evaluation scores through the softmax activation function, and outputs the fitness evaluation score corresponding to the maximum probability as the final prediction result.

[0151] Embodiment 2

[0152] Please refer to Figure 3 As shown, this embodiment provides an over-the-horizon single-soldier video transmission relay enhancement device for complex terrains, further including:

[0153] A forwarding optimization module that adaptively optimizes the forwarding of the target image data of the single-soldier relay device based on the fitness evaluation scores of N point cloud data subspaces.

[0154] The method for adaptively optimizing the forwarding of the target image data of the single-soldier relay device includes:

[0155] Sum the fitness evaluation scores of the N point cloud data subspaces to obtain the total fitness evaluation score;

[0156] If the total fitness evaluation score is greater than or equal to the preset total score threshold, there is no need to further forward the target image data, and the target image data is transmitted to the terminal device;

[0157] If the total fitness evaluation score is less than the preset total score threshold, it is necessary to further forward the target image data, forward the data to the next-hop single-soldier relay device, and repeat the above determination process until the target image data is successfully transmitted to the terminal device.

[0158] It should be noted that this solution takes the image data transmission in the single-soldier relay mode as an example. The work process is as follows:

[0159] During the task execution, the relay device carried by the single soldier (hereinafter referred to as the "single-soldier relay device") continuously monitors and collects the wireless channel state information of the surrounding environment.

[0160] After the remote data acquisition device (such as a drone, surveillance camera, or driverless vehicle) obtains the target image data, if the remote data acquisition device cannot directly establish a stable communication link with the terminal device (such as a command center or a backend server), the target image data is first transmitted via a wireless link to the nearest single-soldier relay device. The single-soldier relay device establishes a wireless connection with other nearby single-soldier relay devices or fixed base stations through an adaptive multi-hop relay link mechanism and dynamically adjusts the transmission path. The single-soldier relay device determines whether to further forward the target image data based on the fitness evaluation scores of N point cloud data subspaces; if it does not need to further forward the target image data, it transmits the target image data to the terminal device; if it does not need to further forward the target image data, the single-soldier relay device forwards the target image data to the next-hop single-soldier relay device until the target image data is successfully transmitted to the terminal device.

[0161] Embodiment 3

[0162] Please refer to Figure 2 As shown, this embodiment provides an ultra-long-range single-soldier image transmission relay enhancement method for complex terrains, including:

[0163] Collect target image data;

[0164] Collect three-dimensional point cloud data of the environment where the single soldier is located;

[0165] Divide the three-dimensional point cloud data to obtain N point cloud data subspaces;

[0166] Extract the environmental feature data of the N point cloud data subspaces to obtain the environmental feature data of the N point cloud data subspaces;

[0167] Collect the network operation data of the N point cloud data subspaces;

[0168] Extract the network feature data of the N point cloud data subspaces to obtain the network feature data of the N point cloud data subspaces;

[0169] Adaptive optimization of the transmission path of the single-soldier relay device based on the environmental feature data and network feature data of the N point cloud data subspaces.

[0170] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

[0171] Finally, the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A beyond-line-of-sight image transmission relay enhancement method for complex terrain, characterized in that: include: Collect target image data; Collect 3D point cloud data of the soldier’s environment; Divide the three-dimensional point cloud data to obtain N point cloud data subspaces; Perform feature extraction on N point cloud data subspaces to obtain environmental feature data of the N point cloud data subspaces; Collect network operation data of N point cloud data subspaces; Perform feature extraction on the network operation data of N point cloud data subspaces to obtain network feature data of N point cloud data subspaces; The transmission path of the individual relay device is adaptively optimized based on the environmental feature data and network feature data of N point cloud data subspaces.

2. The method for enhancing the beyond-visual-range image transmission relay for individual soldiers in complex terrain according to claim 1 is characterized in that: The method for adaptively optimizing the transmission path of a single-soldier relay device includes: The environmental feature data and network feature data of N point cloud data subspaces are respectively input into the fitness evaluation model to obtain the fitness evaluation scores corresponding to the N point cloud data subspaces; The point cloud data subspace whose fitness evaluation score is greater than or equal to a preset fitness evaluation score threshold is constructed into a set of transmission nodes to be planned; The transmission nodes to be planned in the set of transmission nodes to be planned are connected to each other to form G transmission path sub-segments, where G is an integer greater than zero; For each transmission path sub-segment, the corresponding transmission path sub-segment weight is calculated based on the environmental characteristic data and network characteristic data of the starting point and the end point of the transmission path sub-segment; the G transmission path sub-segment weights are constructed into a transmission path sub-segment weight set; Based on the transmission path sub-segment weight set, G transmission path sub-segments are optimized to obtain the optimal transmission path; so that the individual relay device transmits the target image data according to the optimal transmission path.

3. The method for enhancing the beyond-visual-range image transmission relay for individual soldiers in complex terrain according to claim 2 is characterized in that: The method for acquiring the optimal transmission path includes: The number of transmission path sub-segments is preset to R, where R is an integer greater than zero; the optimal transmission path is initialized to empty; Selecting a transmission path subsegment with the largest transmission path subsegment weight from the transmission path subsegment weight set, recording it as the transmission path subsegment to be planned, and removing the transmission path subsegment to be planned from the transmission path subsegment weight set; Add the transmission path sub-segment to be planned to the optimal transmission path, and determine whether there is a loop in the optimal transmission path. If there is a loop in the optimal transmission path, remove the transmission path sub-segment to be planned from the optimal transmission path; if there is no loop in the optimal transmission path, keep the transmission path sub-segment to be planned in the optimal transmission path; The above process is repeated until the number of transmission path sub-segments in the optimal transmission path reaches R.

4. The method for enhancing beyond-visual-range individual soldier image transmission relay in complex terrain according to claim 3 is characterized in that: Methods for determining whether a loop exists in the optimal transmission path include: A transmission path subsegment is randomly selected from the optimal transmission path as the starting point, and traversal is performed along the selected transmission path subsegment; if the traversal returns to the starting position and no transmission path subsegment is repeated in the entire path, there is a loop in the optimal transmission path.

5. The method for enhancing beyond-visual-range individual soldier image transmission relay in complex terrain according to claim 1 is characterized in that: The method for obtaining N point cloud data subspaces includes: Step 1: Divide the 3D point cloud data according to a preset sliding step length to obtain N center data points of the 3D point cloud data; let the initial value of n be 1, and the value range of n be 1 to N; Step 2: Obtain the nth central data point, select H point cloud data points adjacent to the nth central data point from the three-dimensional point cloud data; construct the nth point cloud data subspace with the nth central data point and the H point cloud data points; Step 3: Let n=n+1. If n is less than or equal to N, continue to execute step 2. If n is greater than N, end the current process and obtain N point cloud data subspaces.

6. The method for enhancing beyond-visual-range individual soldier image transmission relay in complex terrain according to claim 1 is characterized in that: The method for obtaining the environmental feature data of N point cloud data subspaces includes: S100: let the initial value of n be 1, and the value range of n be 1 to N; S101: based on the three-dimensional coordinates of the central data point in the nth point cloud data subspace and the corresponding H point cloud data points, calculate the environmental slope and environmental surface roughness of the nth point cloud data subspace; based on the incident laser intensity and reflected laser intensity of the central data point in the nth point cloud data subspace and the corresponding H point cloud data points, calculate the electromagnetic wave reflectivity of the nth point cloud data subspace; S102: extracting features of the environment slope, environment surface roughness and electromagnetic wave reflectivity according to a preset method to obtain the electromagnetic wave refractive index of the nth point cloud data subspace; S103: constructing environmental characteristic data of the nth point cloud data point by using environmental slope, environmental surface roughness, electromagnetic wave reflectivity and electromagnetic wave refractive index; S104: Let n=n+1. If n is less than or equal to N, continue to execute S101 to S103. If n is greater than N, obtain the corresponding environmental feature data of N point cloud data subspaces and end the current process.

7. The method for enhancing the beyond-visual-range image transmission relay for individual soldiers in complex terrain according to claim 6 is characterized in that: The method for obtaining the electromagnetic wave refractive index of the nth point cloud data subspace includes: The environmental slope, environmental surface roughness and electromagnetic wave reflectivity of the nth point cloud data subspace are input into the medium type diagnosis model to obtain the medium type of the nth point cloud data subspace; The electromagnetic wave propagation velocity corresponding to the medium type of the nth point cloud data subspace is obtained from the pre-constructed medium-propagation velocity database; the electromagnetic wave refractive index of the nth point cloud data point is calculated based on the speed of light and the electromagnetic wave propagation velocity.

8. The method for enhancing beyond-visual-range individual soldier image transmission relay in complex terrain according to claim 1 is characterized in that: The method for obtaining network feature data of N point cloud data subspaces includes: S200: Let the initial value of n be 1, and the value range of n be 1 to N; S201: Based on the network bandwidth, signal power and noise power of the nth point cloud data subspace, the corresponding channel capacity is calculated; based on the signal power, noise power and network packet loss rate of the nth point cloud data subspace, the corresponding link quality index is calculated; based on the network packet loss rate and network bandwidth, the link throughput of the nth point cloud data subspace is calculated; based on the link quality index and network delay, the link stability factor of the nth point cloud data subspace is calculated; S202: constructing the channel capacity, link quality index, link throughput and link stability factor into network feature data of the nth point cloud data subspace; S203: Let n=n+1. If n is less than or equal to N, continue to execute S201 to S102. If n is greater than N, the network feature data of the N point cloud data subspaces are constructed into a network feature data set, and the current process ends.

9. The method for enhancing the beyond-visual-range image transmission relay for individual soldiers in complex terrain according to claim 2 is characterized in that: The training method of the fitness evaluation model includes: Pre-collecting a fitness evaluation data set, wherein the fitness evaluation data set includes K groups of fitness evaluation data and fitness evaluation scores corresponding to the K groups of fitness evaluation data, where K is a positive integer greater than 0, and the fitness evaluation data includes environmental feature data and network feature data; dividing the fitness evaluation data set into a training set and a validation set, wherein the training set is used to train a fitness evaluation model, and the validation set is used to evaluate the generalization performance of the fitness evaluation model; During the training process of the fitness evaluation model, the minimization of the cross entropy loss function is used as the optimization goal, and the early stopping strategy is used to monitor the performance of the validation set. The model performance is optimized by continuously adjusting the network parameters. When the prediction accuracy on the validation set reaches the expected accuracy, it is considered that the fitness evaluation model has converged and the training is stopped. The fitness evaluation model is trained using a deep neural network based on a multi-layer perceptron. The fitness evaluation data is converted into feature vectors; the input layer of the fitness evaluation model receives the feature vectors, and the nonlinear relationship in the data is extracted through multiple hidden layers. Finally, the output layer of the fitness evaluation model calculates the probability distribution of the fitness evaluation score through the softmax activation function, and outputs the fitness evaluation score corresponding to the maximum probability as the final prediction result.

10. A beyond-visual-range individual image transmission relay enhancement device for complex terrain, used to implement the beyond-visual-range individual image transmission relay enhancement method for complex terrain as described in any one of claims 1 to 9, characterized in that: include: An image acquisition module, used for acquiring target image data; The first acquisition module is used to collect three-dimensional point cloud data of the environment where the individual soldier is located; The first processing module is used to divide the three-dimensional point cloud data to obtain N point cloud data subspaces; The second processing module is used to extract features from the N point cloud data subspaces to obtain environmental feature data of the N point cloud data subspaces; The second acquisition module is used to collect network operation data of N point cloud data subspaces; The third processing module is used to extract features from the network operation data of the N point cloud data subspaces to obtain network feature data of the N point cloud data subspaces; The path optimization module adaptively optimizes the transmission path of the individual relay device based on the environmental feature data and network feature data of N point cloud data subspaces.

Images